UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Application of computer-aided drug discovery methods for targeting the oncogenic activity of Myc in prostate… Carabet, Lavinia Arielle 2019

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2019_september_carabet_lavinia_arielle.pdf [ 7.91MB ]
Metadata
JSON: 24-1.0380506.json
JSON-LD: 24-1.0380506-ld.json
RDF/XML (Pretty): 24-1.0380506-rdf.xml
RDF/JSON: 24-1.0380506-rdf.json
Turtle: 24-1.0380506-turtle.txt
N-Triples: 24-1.0380506-rdf-ntriples.txt
Original Record: 24-1.0380506-source.json
Full Text
24-1.0380506-fulltext.txt
Citation
24-1.0380506.ris

Full Text

APPLICATION OF COMPUTER-AIDED DRUG DISCOVERY METHODS FOR TARGETING THE ONCOGENIC ACTIVITY OF MYC IN PROSTATE CANCER   by  Lavinia Arielle Carabet MSc., Johns Hopkins University, USA, 2009 MSc., Concordia University, Canada, 2014  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Experimental Medicine)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   August 2019  © Lavinia Arielle Carabet, 2019 ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled: Application of Computer-Aided Drug Discovery Methods for Targeting the Oncogenic Activity of Myc in Prostate Cancer  submitted by Lavinia Arielle Carabet in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Experimental Medicine  Examining Committee: Dr. Artem Cherkasov, Urologic Sciences Supervisor  Dr. Robert N. Young, Chemistry, Simon Fraser University Supervisory Committee Member  Dr. Joerg Gsponer, Biochemistry and Molecular Biology University Examiner Dr. Caigan Du, Urologic Sciences University Examiner  Additional Supervisory Committee Members: Dr. Paul S. Rennie, Urologic Sciences Supervisory Committee Member Dr. Xuesen Dong, Urologic Sciences Supervisory Committee Member  iii  Abstract To date, castration-resistant prostate cancer (CRPC) remains an incurable disease, as conventional therapeutic inhibition of androgen receptor (AR) signaling with anti-androgens inevitably leads to treatment-resistance, further progression to final stage neuroendocrine prostate cancer (NEPC), and rapid demise. Therefore, development of novel targeted therapies for CRPC and NEPC is of paramount importance. Targeting the oncogenic activity of Myc family of transcription factors has long been and currently is a major topic of cancer research. While Myc is an essential regulator of normal growth, its exacerbated expression is a hallmark of human cancer. Amplifications of Myc family members play critical roles in prostate cancer progression and therapy-resistance. c-Myc is amplified across its full-spectrum and has special relevance in CRPC as it positively regulates the expression and activity of AR itself as well as of ligand-independent AR-V truncated variants, such as AR-V7 that confers resistance to anti-androgens. Moreover, N-Myc amplifications induce NEPC phenotype. Although Myc proteins are high-value targets for therapeutic intervention, clinically viable Myc-directed inhibitors await discovery. Intrinsically disordered, Myc lacks effective binding pockets. Therefore, the use of conventional methods of structure-based drug discovery is an inherent challenge. Moreover, the oncogenic function of Myc is dependent on its dimerization with the obligate partner Max, which together form a functional transcriptional complex capable of activating critical genomic targets and eliciting oncogenic effects. This dissertation describes the discovery and development of novel small-molecule inhibitors targeting the oncogenic activity of Myc-Max complexes. Specifically, we utilized methods of computer-aided drug discovery (CADD) to target directly complexes of c-Myc and iv  N-Myc with Max, as well as Myc-upregulated hnRNP A1 splicing factor. The use of CADD enabled us to identify small-molecule drug candidates, which selectively disrupt critical protein-nucleic acid interactions – a therapeutic approach that has not been previously exercised for these targets. The CADD techniques encompassed large-scale structure-based docking and molecular dynamics simulations along with ligand-based approaches including pharmacophore modeling, chemical similarity searches and ADMET profiling, complemented by experimental validation. On the outlook, the identified lead inhibitors lay the foundation for development of safe and effective clinical candidates that may serve as prospective therapeutics for CRPC and NEPC.   v  Lay Summary  Prostate cancer (PCa) is the most frequently diagnosed cancer in men. To date, therapeutic options for advanced stages of disease are limited. Currently approved drugs targeting the androgen receptor (major PCa driver) become ineffective as treatment-resistance inevitably develops. Consequently, there is an urgent unmet clinical need for innovative PCa therapeutics with novel mechanisms of action. PCa progression to its resistant and lethal forms has been associated with elevated activity of Myc oncoproteins. Development of clinically viable small-molecules targeting Myc is a major objective of modern cancer research, not yet achieved due to significant Myc-associated challenges and a general lack of rational efforts. This study describes the application of computer-assisted drug design technologies towards the discovery of novel inhibitors targeting the transforming activity of Myc in PCa. The identified lead compounds lay the groundwork for developing safe and effective clinical candidates that may serve as promising therapeutics for treatment of PCa.  vi  Preface  1. A version of Chapter 1 has been published [Carabet LA, Rennie PS, Cherkasov A. Therapeutic Inhibition of Myc in Cancer. Structural Bases and Computer-Aided Drug Discovery Approaches. Int J Mol Sci 2018, 20 (1)]. Drs. Cherkasov and Rennie are senior authors and they supervised this project from concept formation to manuscript revision. I wrote the original manuscript, revised and finalized the manuscript.   Chapter 1 includes excerpts from published work [Paul N, Carabet LA, Lallous N, Yamazaki T, Gleave ME, Rennie PS, Cherkasov A. Cheminformatics Modeling of Adverse Drug Responses by Clinically Relevant Mutants of Human Androgen Receptor. J Chem Inf Model 2016, 56, 2507-2516]. I developed novel 4D inductive descriptors implemented in the Scientific Vector Language and contributed in writing the original manuscript.  2. Part of the work described in Chapter 3 (section 3.2.1) has been published [Carabet LA, Lallous N, Leblanc E, Ban F, Morin H, Lawn S, Ghaidi F, Lee J, Mills IG, Gleave ME, Rennie PS, Cherkasov A. Computer-aided drug discovery of Myc-Max inhibitors as potential therapeutics for prostate cancer. Eur J Med Chem 2018, 160, 108-119]. Drs. Cherkasov and Rennie are senior authors and they supervised this project from concept formation to manuscript revision. I performed all the computational drug discovery work, as well as drafted and revised the manuscript. Drs. Lallous N and Leblanc E designed the wet lab experiments and edited the manuscript. Members of Dr. Rennie’s group, Drs. Lallous N, Leblanc E, Lawn S and Dalal K, as well as Ms. Morin H, Ghaidi F and Mr. Lee J contributed to the biological evaluation of the inhibitors. Dr. Ban F helped with the initial computations.  vii  I performed all the computational work in section 3.2.2 of Chapter 3 as well as that in Chapter 4. Dr. Singh K and Ms. Kalyta A contributed to the biological evaluation of the inhibitors.  3. Work described in Chapter 5 (except section 5.2.2.1) has been published [Carabet LA, Leblanc E, Lallous N, Morin H, Ghaidi F, Lee J, Rennie PS, Cherkasov A. Computer-aided discovery of small molecules targeting the RNA splicing activity of hnRNP A1 in castration-resistant prostate cancer. Molecules 2019, 24(4)]. Drs. Cherkasov and Rennie are senior authors and they supervised this project from concept formation to manuscript revision. I performed all CADD work and wrote the original manuscript, revised and finalized the manuscript. Drs. Leblanc E and Lallous N designed the wet lab experiments. Members of Dr. Rennie’s group, Drs. Leblanc E and Lallous N, as well as Ms. Morin H, Ghaidi F and Mr. Lee J contributed to the biological evaluation of the inhibitors. Ms. Kalyta A helped with wet lab technical assistance.  viii  Table of Contents  Abstract ......................................................................................................................................... iii Lay Summary .................................................................................................................................v Preface ........................................................................................................................................... vi Table of Contents ....................................................................................................................... viii List of Tables .............................................................................................................................. xiii List of Figures ............................................................................................................................. xiv List of Abbreviations ................................................................................................................. xxi Acknowledgements ................................................................................................................. xxvii Dedication ............................................................................................................................... xxviii Chapter 1: Introduction ................................................................................................................1 1.1 Prostate Cancer ............................................................................................................... 1 1.1.1 Diagnosis..................................................................................................................... 2 1.1.2 Clinical States and Treatment Options ....................................................................... 4 1.1.3 Androgen Receptor ..................................................................................................... 6 1.1.4 Drug Resistance .......................................................................................................... 8 1.1.5 AR Independence ...................................................................................................... 12 1.2 MYC ............................................................................................................................. 13 1.2.1 Discovery of MYC Oncogene .................................................................................. 13 1.2.2 MYC Family of Transcription Factors ..................................................................... 13 1.2.3 Physiological Function of MYC ............................................................................... 14 1.2.4 MYC and Cancer ...................................................................................................... 14 ix  1.2.4.1 MYC and Prostate Cancer ................................................................................ 15 1.2.5 MYC Deregulation.................................................................................................... 16 1.2.6 MYC Structure-Function Relationships ................................................................... 17 1.2.6.1 MYC Gene Organization and Transcriptional Control ..................................... 17 1.2.6.2 MYC Protein Organization and Translational Control ..................................... 21 1.2.6.3 MYC/MAX Extended Network ........................................................................ 25 1.2.6.4 Structural Aspects of Protein-Protein and Protein-DNA Interactions within the Myc/Max Extended Network and Beyond ....................................................................... 29 1.2.7 Drugging MYC ......................................................................................................... 35 1.2.7.1 Challenges and Opportunities ........................................................................... 36 1.2.7.2 MYC Targeting Strategies ................................................................................ 37 1.2.7.3 Indirect Therapeutic Inhibition ......................................................................... 37 1.2.7.4 Synthetic Lethality ............................................................................................ 38 1.2.7.5 Direct Therapeutic Inhibition ............................................................................ 39 1.3 Current Advances in the Development of Direct MYC Inhibitors ............................... 39 1.3.1 Direct Inhibitors of MYC Expression ....................................................................... 40 1.3.2 Direct Inhibitors of MYC-MAX Function ................................................................ 53 1.3.2.1 Inhibitors of MYC-MAX Heterodimerization .................................................. 54 1.3.2.2 Inhibitors of MYC-MAX Binding to DNA ...................................................... 65 1.3.2.3 Computational Approaches toward MYC-MAX Inhibition ............................. 70 1.3.2.4 Binding sites for MYC-MAX Small Molecule Inhibitors ................................ 78 1.3.2.5 The Need for MYC-MAX Small Molecule Inhibitors ..................................... 89 1.4 Drug Discovery and Development................................................................................ 90 x  1.4.1 Computer-Aided Drug Discovery ............................................................................. 96 1.5 Objective and Rationale of the Study ......................................................................... 108 1.6 Thesis Layout .............................................................................................................. 110 Chapter 2: Materials and Methods ..........................................................................................112 2.1 In silico modeling ....................................................................................................... 112 2.1.1 Protein Modeling .................................................................................................... 112 2.1.2 Binding Sites Identification .................................................................................... 113 2.1.3 Protein Structure Preparation .................................................................................. 113 2.1.4 Ligand Preparation .................................................................................................. 114 2.1.5 Virtual Screening .................................................................................................... 114 2.1.6 Consensus Scoring and Voting ............................................................................... 116 2.1.7 Pharmacophore Modeling ....................................................................................... 117 2.1.8 Chemical Similarity Searches ................................................................................. 117 2.1.9 MM/GBSA Simulations.......................................................................................... 118 2.1.10 Molecular Dynamics Simulations ....................................................................... 119 2.1.11 Free Energy Perturbations Molecular Dynamics Simulations ............................ 120 2.1.12 ADMET Prediction ............................................................................................. 122 2.2 Experimental Validation ............................................................................................. 122 2.2.1 Cell Culture and Reagents ...................................................................................... 122 2.2.2 Transfection and Reporter Assays .......................................................................... 122 2.2.3 Cell Viability Assay ................................................................................................ 123 2.2.4 Protein Purification ................................................................................................. 123 2.2.5 Bio-layer Interferometry (BLI) Assay .................................................................... 124 xi  2.2.6 Chromatin Fractionation Assay .............................................................................. 125 2.2.7 Western Blotting ..................................................................................................... 125 2.2.8 qRT-PCR................................................................................................................. 126 2.2.9 Metabolic Stability .................................................................................................. 127 Chapter 3: Development of Novel c-Myc-Max Inhibitors using Methods of CADD ...........128 3.1 Background ................................................................................................................. 128 3.2 Results ......................................................................................................................... 128 3.2.1 Discovery of c-Myc-Max Inhibitors ....................................................................... 128 3.2.1.1 Binding Sites Identification ............................................................................ 128 3.2.1.2 In silico Identification of Hit Compounds ...................................................... 131 3.2.1.3 In silico Binding Mode of VPC-70063 ........................................................... 136 3.2.1.4 In vitro Characterization of Hit Compounds .................................................. 139 3.2.2 Optimization of c-Myc-Max Inhibitors .................................................................. 145 3.2.2.1 VPC-70063 Derivatives .................................................................................. 147 3.2.2.2 Development of Lead Compound VPC-70551 ............................................... 149 3.3 Discussion ................................................................................................................... 167 Chapter 4: Targeting N-Myc-Max with c-Myc-Max Inhibitors ............................................173 4.1 Background ................................................................................................................. 173 4.2 Results ......................................................................................................................... 175 4.3 Discussion ................................................................................................................... 181 Chapter 5: Development of hnRNP A1 Inhibitors..................................................................183 5.1 Background ................................................................................................................. 183 5.2 Results ......................................................................................................................... 186 xii  5.2.1 Binding site identification on hnRNP A1 RBD domain ......................................... 186 5.2.2 In silico identification of hit compounds targeting the hnRNP A1 binding site ..... 189 5.2.2.1 VPC-80021 ..................................................................................................... 190 5.2.2.2 VPC-80051 ..................................................................................................... 198 5.3 Discussion ................................................................................................................... 205 Chapter 6: Conclusion ...............................................................................................................208 6.1 Summary of the Study ................................................................................................ 208 6.2 Future Directions ........................................................................................................ 212 Bibliography ...............................................................................................................................216 Appendix .....................................................................................................................................237  xiii  List of Tables  Table 1.1 Risk stratification schema for prostate cancer ................................................................ 4 Table 1.2 Activating AR mutations recurrent in CRPC ................................................................. 9 Table 1.3 Gain-of-function AR splice isoforms recurrent in CRPC ............................................. 10 Table 3.1 Chemical structures and activities of initial hit compounds that bind the ordered Myc-Max DBD. ................................................................................................................................... 134 Table 3.2 Chemical structures and activities of best analogs of VPC-70063. ............................ 148 Table 3.3 Activities of best on-demand synthesizable ENAMINE-REAL derivatives of VPC-70127........................................................................................................................................... 158 Table 5.1 In silico scores and synthesis cost of representative VPC-80021 medchem derivatives...................................................................................................................................................... 197  xiv  List of Figures  Figure 1.1 Structure of AR-FL and AR-V truncated variants transcripts and proteins. ................. 7 Figure 1.2 Transcriptional activation by (A) AR full-length and (B) constitutively active AR-V splice isoforms. ............................................................................................................................... 8 Figure 1.3 40 years of Myc research. ............................................................................................ 13 Figure 1.4 Distribution of Myc amplification frequencies among various cancers. ..................... 14 Figure 1.5 Myc signaling pathway................................................................................................ 17 Figure 1.6 Structure of MYC gene and protein: functional domains and interactors. ................... 18 Figure 1.7 The extended Myc-Max network. The players and their regulatory functions. .......... 25 Figure 1.8 Structure-function schematic of the members of the extended Myc/Max network. ... 26 Figure 1.9 Structural Aspects of Protein-Protein and Protein-DNA Interactions within the Myc/Max Extended Network and Beyond. .................................................................................. 30 Figure 1.10 Myc-Max/DNA 5’-CACGTG-3’ E-box recognition interface. ................................. 33 Figure 1.11 Chemical structures of TMPyP4 and Se2SAP cationic porphyrin G-quadruplex (G4) stabilizers. ..................................................................................................................................... 42 Figure 1.12 Chemical structure of SYUIO-05 quindoline derivative G4 stabilizer. .................... 43 Figure 1.13 Chemical structures of CX-3543 (quarfloxin) and optimized CX-5461 G4 stabilizers in early phase clinical trials........................................................................................................... 44 Figure 1.14 Chemical structures of selective Myc G4 stabilizers, Compound 1 and optimized derivative DC-34. .......................................................................................................................... 47 Figure 1.15 Chemical structure of IZCZ-3 G4 stabilizer. ............................................................. 48 Figure 1.16 Chemical structure of anthracenyl isoaxole amides (AIMs) G4 ligands. .................. 49 xv  Figure 1.17 Chemical structure of natural compound carbamide 1 identified as G4 stabilizer through computational approaches. .............................................................................................. 50 Figure 1.18 Chemical structure of VS10 G4 stabilizer identified through computational approaches..................................................................................................................................... 51 Figure 1.19 Chemical structure of compound 56 G4 stabilizers identified computationally ....... 52 Figure 1.20 Chemical structure of TPP Myc G4 stabilizer identified using a rational drug design platform ......................................................................................................................................... 52 Figure 1.21 Chemical structures of 10058-F4, 10074-A4, and 10074-G5 inhibitors of Myc-Max dimerization identified through high-throughput screening of a finite combinatorial library. ..... 56 Figure 1.22 Chemical structures of JY-3-094 and 3jc48-3, improved analogs of compound 10074-G5. ..................................................................................................................................... 58 Figure 1.23 Chemical structures of Mycro3, KJ-Pyr-9, sAJM589, and MYCMI-6 inhibitors of Myc-Max dimerization identified through high-throughput screening of diverse small-sized compound libraries........................................................................................................................ 61 Figure 1.24 Chemical structures of MYRA-A, NSC308848, JKY-2-169, and KSI-3716 inhibitors of Myc-Max binding to DNA identified through high-throughput screening or specifically engineered to disrupt protein-DNA interactions ........................................................................... 67 Figure 1.25 Chemical structures of PKUMDL-YC-1201 to -1205 compounds that disrupt Myc-Max dimerization identified through virtual screening (VS) multi-conformational docking against a reference ensemble of Myc disordered conformations generated through molecular dynamics (MD) simulations .......................................................................................................... 72 xvi  Figure 1.26 Chemical structure of 7594-0035 Myc-Max inhibitor identified in silico utilizing the Myc-Max 1NKP X-ray structure, yet targeting a previously reported disordered binding region for 10074-G5 but in ordered form ................................................................................................ 75 Figure 1.27 Chemical structures of representative Max-Max stabilizers that alter Myc-Max function by blocking binding to equivalent DNA binding sites ................................................... 77 Figure 1.28 Binding sites for structurally diverse Myc-Max inhibitors mapped along Myc and Max protein sequences .................................................................................................................. 79 Figure 1.29 Drug discovery and development pipeline. ............................................................... 91 Figure 1.30 General workflow for virtual screening large chemical libraries. ............................. 93 Figure 1.31 Schematic of hit-to-lead optimization process. ......................................................... 94 Figure 3.1 Three independent binding sites identified in silico at the Myc-Max/DNA and Myc-Max dimerization interfaces. ...................................................................................................... 131 Figure 3.2 In silico model of the basic/helix-loop-helix/leucine zipper (bHLHZ) domain of Myc-Max bound to the 5’-CACGTG-3’ DNA recognition sequence, the Myc-Max predicted binding site at the DNA interface and the pharmacophore utilized for subsequent virtual screening and filtering of potential binders........................................................................................................ 132 Figure 3.3 (A) Predicted binding pose of VPC-70063 in space-filling representation within the Myc-Max DBD pocket; (B) Pocket residues predicted to interact with VPC-70063 at the Myc-Max DNA interface..................................................................................................................... 137 Figure 3.4 Myc-Max/VPC-70063 interactions obtained from MM/GBSA simulations with implicit solvent in two-dimensional representation. ................................................................... 139 Figure 3.5 Effects of VPC-70063 on Myc-Max transcriptional activity, AR-V7 levels and viability of selected cell lines. (A) Dose response effect in LNCaP PCa cells on the xvii  transcriptional activity of c-Myc by using a c-Myc mediated luciferase reporter. (B) Inhibition of Myc-Max reduces the levels of AR-V7 in 22Rv1 cells. (C) The effect of VPC-70063 in comparison with 10074-G5 on cell viability of Myc positive (LNCaP) and Myc negative (HO15.19) cell lines .................................................................................................................... 141 Figure 3.6 Inhibition of Myc-Max with VPC-70063 induces apoptosis of LNCaP cells as indicated by cleavage of PARP in Western blot. ........................................................................ 142 Figure 3.7 BLI quantification of disruption of Myc-Max interactions with DNA upon treatment with VPC-70063. (A) Purification of GST-Myc and His-Max using size exclusion chromatography. (B) Inhibition of Myc-Max interaction with the biotinylated E-box quantified by BLI in presence of 500 μM of studied compounds. (C) Dose response inhibition of Myc-Max binding to DNA in presence of VPC-70063. .............................................................................. 143 Figure 3.8 VPC-70063 strongly dissociates Myc from chromatin and significantly reduces Myc protein levels in a dose-dependent manner at single-digit μM range ......................................... 144 Figure 3.9 Fraction of original concentration of compounds 10074-G5 (left) and VPC-70063 (right) versus time in mouse liver microsomes ........................................................................... 145 Figure 3.10 In silico drug discovery pipeline ............................................................................. 147 Figure 3.11 Myc-Max/VPC-70033 interactions obtained from MM/GBSA simulations with implicit solvent............................................................................................................................ 150 Figure 3.12 VPC-70127 a novel scaffold obtained through ROCS chemical similarity searches using VPC-70033 as a template. ................................................................................................. 151 Figure 3.13 (A) Docking pose of VPC-70127 in space-filling representation within the Myc-Max DBD pocket. (B) Myc-Max/VPC-70127 interactions within the DBD site ............................... 152 xviii  Figure 3.14 Myc-Max/VPC-70127 interactions obtained from MM/GBSA simulations with continuum solvent. ...................................................................................................................... 153 Figure 3.15 Dynamics of Myc-Max/VPC-70127 complex. VPC-70127 interacts favorably with Myc-Max DBD residues in 120-ns explicit solvent MD simulations without restraints. ........... 155 Figure 3.16 Predicted metabolism of VPC-70127. ..................................................................... 156 Figure 3.17 Strong effect of VPC-70127 on Myc-Max transcriptional activity, IC50 = 1 μM (A) and apoptosis (B) in LNCaP cells. .............................................................................................. 157 Figure 3.18 Chemical structure of three active congeners of VPC-70127 ................................. 159 Figure 3.19 (A) Effect of VPC-70551 on Myc-Max transcriptional activity in LNCaP cells, IC50 = 4 μM. (B). Minimal effect on viability of HO15.19 Myc null cell line. .................................. 160 Figure 3.20 VPC-70551 improved chemical scaffold. ............................................................... 160 Figure 3.21 Fraction of original concentration of lead compound VPC-70551 versus time in mouse liver microsomes (t1/2 = 140 min). ................................................................................. 161 Figure 3.22 (A) Docking pose of VPC-70551 in space-filling representation within the Myc-Max DBD pocket. (B) Rigid Myc-Max/VPC-70551 interactions within the DBD site ..................... 162 Figure 3.23 Myc-Max/VPC-70551 interactions obtained from MM/GBSA simulations in implicit solvent. ........................................................................................................................................ 163 Figure 3.24 Dynamics of Myc-Max/VPC-70551 complex. VPC-70551 interacts favorably with Myc-Max DBD residues in 120-ns explicit solvent MD simulations......................................... 164 Figure 3.25 VPC-70551 competes with DNA by disrupting critical Myc-Max/DNA interactions...................................................................................................................................................... 166 Figure 4.1 High-resolution N-Myc-Max homology model using c-Myc-Max X-ray structure as a template. ...................................................................................................................................... 175 xix  Figure 4.2 Binding poses and interactions of VPC-70551 within the c-Myc-Max (A) and N-Myc-Max (B) DBD pockets. ............................................................................................................... 177 Figure 4.3 VPC-70551 interactions with c-Myc-Max (A) and N-Myc-Max (B) obtained from MM/GBSA simulations in implicit solvent ................................................................................ 178 Figure 4.4 Effects of lead compound VPC-70551 on N-Myc-Max transcriptional activity in IMR32 N-Myc amplified cell line (A), on apoptosis (B) and downstream target genes (C) in the same cell line............................................................................................................................... 179 Figure 4.5 Effect of lead compound VPC-70551 on viability of N-Myc-driven IMR32 cells (A), and N-Myc independent SK-N-AS and NB-16 neuroblastoma cell lines (B). ........................... 180 Figure 4.6 Effect of lead compound VPC-70551 on viability of N-Myc-driven LASCPC-01 NEPC cell line............................................................................................................................. 181 Figure 5.1 hnRNP A1 protein domains organization.................................................................. 185 Figure 5.2 Chemical structure of quercetin, literature hnRNP A1 inhibitor. .............................. 185 Figure 5.3 In silico model of the UP1 domain of the hnRNP A1 splicing factor bound to the 5’-AG-3’ recognition sequence and the predicted binding site on hnRNP A1 at the RNA recognition interface....................................................................................................................................... 187 Figure 5.4 hnRNP A1 interactions with 5’-AG-3’ RNA recognition sequence within the pocket...................................................................................................................................................... 188 Figure 5.5 (Left) Predicted binding pose of VPC-80021 within the hnRNP A1 RBD pocket. (Right) Significant structural overlap between VPC-80021 and nucleobases of the 5′-AG-3′ RNA within the hnRNP A1 RBD pocket. ............................................................................................ 191 Figure 5.6 hnRNP A1/VPC-80021 interactions obtained from MM/GBSA simulations with implicit solvent............................................................................................................................ 192 xx  Figure 5.7 Dynamics of hnRNP A1-80021 complex during 350 ns MD simulations with explicit solvent. ........................................................................................................................................ 194 Figure 5.8 Chemical structures of representative in silico designed derivatives of VPC-80021...................................................................................................................................................... 196 Figure 5.9 hnRNP A1-ligand interactions diagram obtained from FEP+ MD simulations of derivative 80021-A6 morphed from 80021-A1 intermediate analog of parental VPC-80021 ... 197 Figure 5.10 (a) Left. Predicted binding pose of VPC-80051 within the hnRNP A1 RBD pocket. Right. Significant structural overlap between VPC-80051 and nucleobases of the 5′-AG-3′ RNA within the hnRNP A1 RBD pocket; (b) Left. Predicted binding pose of quercetin. Right. Structural overlap between quercetin and the 5′-AG-3′ RNA .................................................... 200 Figure 5.11 hnRNP A1-ligand interactions obtained from MM/GBSA simulations with implicit solvent for: (a) VPC-80051 and (b) quercetin ............................................................................ 202 Figure 5.12 Dynamics of hnRNP A1-80051 complex during 100 ns MD simulations with explicit solvent. ........................................................................................................................................ 204 Figure 5.13 In vitro characterization of VPC-80051 hnRNP A1 inhibitor. (a) Dose-dependent direct binding of VPC-80051 and control quercetin (QRCT) to purified UP1 domain of hnRNP A1 protein quantified by bio-layer interferometry (BLI). (b) Reduction of levels of AR-V7 splice variant upon treatment with 10 and 25 μM of VPC-80051 and QRCT as analyzed by Western blotting. (c) Reduction of 22Rv1 cell viability upon treatment with VPC-80051 and QRCT at greater than 10 μM doses where decrease in AR-V7 levels was observed. ............................... 205  xxi  List of Abbreviations ABS  Androgen Binding Site ADMET Absorption, Distribution, Metabolism, Excretion, and Toxicity AMBER Assisted Model Building and Energy Refinement AR  Androgen Receptor AR-FL  Androgen Receptor Full-Length AR-V  Androgen Receptor Variant AURKA Aurora Kinase A BCL2  BCL2 Apoptosis Regulator BET  Bromodomain and Extra Terminal bHLHLZ Basic Helix-Loop-Helix Leucine Zipper BLI  Bio-Layer Interferometry BRCA2 Breast Cancer Type 2 Susceptibility Protein BRD4  Bromodomain-Containing Protein 4 BRN2  Brain-Specific Homeobox/POU Domain Protein 2 CADD  Computer-Aided Drug Discovery CD  Circular Dichroism CHARMM Chemistry at Harvard Macromolecular Mechanics c-KIT  KIT Proto-oncogene Receptor Tyrosine Kinase CRPC  Castration-Resistant Prostate Cancer CYP  Cytochrome P450 CYP17A1 Cytochrome P450, Subfamily XVII (Steroid 17-Alpha-Hydroxylase) DBD  DNA Binding Domain xxii  DHEA  Dehydroepiandrosterone DHT  5α-Dihydrotestosterone DOPE  Discrete Optimized Protein Energy E-box  Enhancer box eHiTS  Electronic High-Throughput Screening EMSA  Electrophoretic Mobility Shift Assay ER   Estrogen Receptor ERG  ETS Transcription Factor EZH2  Enhancer of Zeste 2 Polycomb Repressive Complex 2 Subunit FBW7  F-Box and WD Repeat Domain Containing 7 FDA  U.S. Food and Drug Administration FEP  Free Energy Perturbations FF  Force Field FOXO3A Forkhead Box O3 FRET  Fluorescence Resonance Energy Transfer FUSE  Far Upstream Sequence Element GBSA  Generalized Born Surface Area Glide  Grid-based Ligand Docking with Energetics GPU  Graphics Processing Units GSK3  Glycogen Synthase Kinase 3 HAT  Histone Acetyltransferase HDAC  Histone Deacetylase HIF1α  Hypoxia-Inducible Factor 1 Subunit Alpha xxiii  HK2  Hexokinase 2 hnRNP A1 Heterogeneous Nuclear Ribonucleoprotein A1 HSP90  Heat Shock Protein 90 h-Telo  Telomere Maintenance hTERT Telomerase Reverse Transcriptase ICM  Internal Coordinate Mechanics IDP  Intrinsically Disordered Protein IL-6  Interleukin 6 LBD  Ligand Binding Domain LHRH  Luteinizing Hormone Releasing Hormone LMO3  LIM Domain Only Protein 3 MAD  MAX Dimerization Protein 1 MAPK  Mitogen-Activated Protein Kinase MAX  MYC Associated Factor X MB  MYC Boxes mCRPC Metastatic Castration-Resistant Prostate Cancer  MD  Molecular Dynamics MGA  MAX Dimerization Protein MGA Miz1   Myc-interacting Zn-finger protein 1 MLX  MAX Dimerization Protein MLX MLXIP MLX Interacting Protein MLXIPL MLX Interacting Protein Like MM  Molecular Mechanics xxiv  MMFF  Merck Molecular Force Field MNT  MAX Network Transcriptional Repressor mTORC1 Mechanistic Target of Rapamycin Complex 1 MXD MAX Dimerization Protein 1 MYC  V-Myc Avian Myelocytomatosis Viral Oncogene Homolog MYCL V-Myc Avian Myelocytomatosis Viral Oncogene Homolog, Lung Carcinoma Derived MYCN V-Myc Avian Myelocytomatosis Viral Oncogene Homolog, Neuroblastoma Derived NEPC  Neuroendocrine Prostate Cancer NHEIII1 Nuclease Hypersensitivity Element III 1 NKX3.1 NK3 Homeobox 1 NLS   Nuclear Localization Signal NM23-H2 Nucleoside Diphosphate Kinase 2 NMR  Nuclear Magnetic Resonance NOESY Nuclear Overhauser Effect Spectroscopy NOTCH Notch Receptor NTD   N-terminal Transactivation Domain OPLS  Optimized Potential for Liquid Simulations PARP  Poly (ADP-Ribose) Polymerase PCa  Prostate Cancer PCR  Polymerase Chain Reaction PDGFA Platelet Derived Growth Factor Subunit A xxv  PI3K  Phosphoinositide 3-kinase PIN  Prostatic Intraepithelial Neoplasia POLII   RNA polymerase II PP2A  Protein Phosphatase 2 PPI  Protein-Protein Interactions PSA  Prostate-Specific Antigen P-TEFb  Positive Transcription Elongation Factor b  PTEN  Phosphatase and Tensin Homolog QSAR  Quantitative Structure-Activity Relationships QSPR  Quantitative Structure-Property Relationships RAS  Rat Sarcoma RB1  Retinoblastoma Protein RBD  RNA Binding Domain RMSD  Root Mean Square Deviation ROCS  Rapid Overlay of Chemical Structures RRM  RNA Recognition Motif RTK  Receptor Tyrosine Kinase SAR  Structure-Activity Relationships SKP2  S-phase Kinase-Associated Protein 2 SMILES Simplified Molecular-Input Line-Entry System SOX2  Sex Determining Region Y-Box 2 SPC  Simple Point-Charge SPR  Surface Plasmon Resonance xxvi  SRRM4 Serine/Arginine Repetitive Matrix Protein 4 TAD  Transactivation Domain TCEP  Tris(2-carboxyethyl)phosphine TCF-4  Transcription Factor 4 TF  Transcription Factor TFAP4 Transcription Factor Activating Enhancer Binding Protein 4 TGF-β             Transforming Growth Factor-β TIP3P  Transferable Intermolecular Potential with 3 Points TMPRSS2 Transmembrane Serine Protease 2 TOCSY Total Correlation Spectroscopy TP53  Tumor Protein P53 TRRAP  Transformation/Transcription Domain Associated Protein UBE2C Ubiquitin Conjugating Enzyme E2 C UGT  Uridine Diphosphate-Glucuronyl Transferase UP1  Unwinding Protein 1 VEGFA Vascular Endothelial Growth Factor A VS  Virtual Screening WDR5  WD Repeat-Containing Protein 5  xxvii  Acknowledgements  I offer my gratitude to Dr. Artem Cherkasov for giving me the opportunity, supervisorship and resources that made this computational drug discovery research possible. I thank my supervisory committee members, Dr. Paul S. Rennie, Dr. Xuesen Dong and Dr. Robert N. Young for guidance and for imparting their expertise on prostate cancer biology, alternative splicing and modern medicinal chemistry. I want to thank Dr. Nada Lallous, Dr. Eric Leblanc and Dr. Kriti Singh for assays design and experimental validation. I am most grateful to Prostate Cancer Canada for financial support as an awardee of a 2 years NextGen 2017 Philip Feldberg Graduate Studentship.  xxviii  Dedication  To the memory of my beloved mother who succumbed to leukemia at an early age with no treatment available.  To my sister, a survivor of a rare life-threatening congenital disease.  To all committed to make a difference by advancing the knowledge and technology to find cures for devastating diseases.  To dance, for health.  To Alex, for love and care.1 Chapter 1: Introduction  1.1 Prostate Cancer Prostate cancer (PCa) is currently the most commonly diagnosed male cancer and is the second leading cause of death in men. The burden of PCa in Canada is substantial, with an estimated annual incidence of over 21,000 new cases and a PCa-related mortality exceeding 4,000 males yearly.1  PCa is a cancerous tumor that starts in the cells of the prostate, a walnut-sized gland, part of the male reproductive and urinary systems, which is highly prone to tumorigenesis in elderly men. The etiology of PCa has been linked to a variety of risk factors that include age (i.e. 75% of patients diagnosed with prostate cancer are older than 65 years), race/ethnicity, diet (i.e. environmental factors), family history (i.e. heritage of BRCA2 mutations), additive genetic alterations, hormonal and sexual factors, and chemical exposure among others.2 A normal prostate epithelial cell can develop into premalignant tumors that can give rise to prostatic intraepithelial neoplasia (PIN) characterized by atypical proliferating luminal epithelium, the precursor of invasive adenocarcinoma and metastatic prostate cancer.3 High-grade PIN and prostate cancer (PCa) share a spectrum of genetic and molecular abnormalities, including specific chromosomal alterations such as TMPRSS2-ERG fusion, loss of PTEN on 10q, amplification of MYC on 8q, along with changes in cell cycle and proliferative status. Phenotypically, the most relevant common alterations of immunohistochemical biomarkers in high-grade PIN and PCa include the above-mentioned MYC amplification as well as the decrease in prostate-specific differentiation markers, such as PSA and NKX3.1.3 2 The seminal work of Huggins and Hodges (1941)4,5 demonstrated that prostate cancer (PCa) is a hormone dependent disease resulting from activation of the androgen receptor (AR) protein by male steroid hormones (androgens), ablation of which leads to clinical benefits and represents the mainstay of PCa therapy even nowadays.   1.1.1 Diagnosis The prostate gland controls the flow of urine and production of seminal fluid. It surrounds the base of the male bladder and is located in front of the rectum. The prostate is normally rubbery, pliable and smooth, but its enlargement and hardness may indicate a diseased condition easily assessable by a physician with a gloved finger during a digital rectal exam (DRE).6 Besides well-established prognostic factors such as age, general health and personal preferences, the stage and grade of PCa dictate which treatment options, from those currently available, are most appropriate for each men. Staging is a way of describing where the cancer is located, how large it is and whether it has spread to the lymph nodes and/or metastasized affecting other parts of the body.  There are two types of staging for PCa. Clinical staging which combines the results of preoperative diagnostic tests such as DRE, PSA (prostate-specific antigen) testing and Gleason score. These results help inform a physician’s decision to order a biopsy and additional tests when concerns about possible metastatic spread arise. Additional investigations may include nuclear bone scans, X-rays of the skeleton, computerized tomography (CT) and magnetic resonance imaging (MRI) scans. Pathologic staging relies on information found during surgery and pathology reports of the prostate tissue removed during a biopsy. The surgery often includes the removal of the entire prostate and some surrounding lymph nodes.  3 Clinicians use two different systems to define stages of PCa. The most common is the TNM system, which provides specific tumor stage information by combining the letters “T” for tumor, “N” for node and “M” for metastasis with numbers and other letters. The other is the Whitmore-Jewett system, which classifies tumor stages by letters (A to D).6  In addition to staging, clinicians utilize grading systems to forecast disease outcome and make informed decisions regarding individual treatment selection. Determining the grade (or aggressiveness) of PCa involves examination of a biopsy specimen under a microscope to assess the appearance of prostate tumor cells as compared to healthy prostate cells.  Similar to staging, there are two systems for grading, where the first general system classifies PCa cells as low-, intermediate- or high-grade. The low-grade type of cancer cell is slow growing and looks similar to normal cells. The intermediate-grade tumor cells are more aggressive and look abnormal relative to low-grade cells. High-grade cancer cells are extremely aggressive. They grow and spread very quickly and do not resemble healthy prostate cells at all. The second, Gleason grading system is the most commonly used in PCa and consists of a cumulative Gleason score, ranging from 2 to 10, highly predictive of the growth and progression of the disease. To determine the overall Gleason score, a pathologist looks at cancer cells arrangement in the prostate and assigns primary and secondary Gleason grades in the range of 1 to 5 each corresponding to two different locations. For primary Gleason grade assessment, the pathologist looks for the main pattern of cell growth where PCa is most obvious. The secondary Gleason grade determination is identical except that the pathologist looks at the second most common tumor pattern. The higher the difference in appearances between normal and cancerous cells the higher the Gleason grade (up to 5). Cumulative Gleason scores of 5 or lower are not used in practice. The lowest Gleason score is 6 representing a low-grade cancer with well-4 differentiated cells. A Gleason score of 7 is medium-grade cancer, and a score from 8 to 10 indicates high-grade cancer with poorly differentiated or undifferentiated cells. The higher the number, the more aggressive the cancer is, rapidly proliferating and invading the entire organism.6 For instance, grade 1 cancer has a Gleason score of 6 (3+3), while the most aggressive grade 5 cancer has Gleason scores of 9 and 10 (4+5, 5+4, and 5+5).7 Although the PSA blood test by itself does not diagnose PCa, it is still significant and widely used in the clinic to detect and quantify prostate cancer from its early to advanced stages. For instance, a level of PSA greater than 10 ng/mL of blood may indicate cancer spread to lymph nodes, bone and other organs. The stage, grade, and PSA level represent the critical information that help both an educated patient and physicians to plan accordingly for optimal available treatment. Overall, PCa patients are stratified in five risk groups7,8 (Table 1.1). Table 1.1 Risk stratification schema for prostate cancer Stratification Description Clinical Stage Gleason Score PSA level Very low risk T1c, N0, M0 <= 6 < 10 ng/mL Low risk T1 to T2a <=6  Intermediate risk T2b to T2c 7 10 to 20 ng/mL High risk T3a 8 to 10 > 20 ng/mL Very high risk T3b to T4, N1, M1 8 to 10, primary 5 > 20 ng/mL  1.1.2 Clinical States and Treatment Options PCa is a heterogeneous disease with various states and subtypes for which adequacy of treatment is carefully considered in clinical practice based on the severity of the disease, hormonal dependency, and metastatic spread. From diagnosis to death, PCa progresses through a sequence of clinical states that include clinically localized disease, rising post-treatment PSA, non-castrate clinical metastases, and castrate-resistant metastatic state.9 Subpopulations of 5 castrate-resistant patients can further progress to end-state neuroendocrine prostate cancer phenotype.10 Stopping the progression and containing the disease to an early clinical state is tantamount to cure. When confined to the prostate gland, localized adenocarcinoma is potentially curable by radical prostatectomy or radiation therapy. These treatment options can be nonetheless problematic, as many patients may not develop symptomatic disease while others may have undetected metastatic spread. Moreover, side effects associated with these treatment options such as urinary symptoms and sexual dysfunction can negatively affect quality of life.6 For locally advanced or metastatic disease, androgen deprivation therapy (ADT) is the standard first-line treatment.11 ADT aims to deplete the levels of androgens by interfering with either their synthesis or their binding to the androgen receptor (AR). Bilateral orchiectomy (surgical castration) is a low cost, low morbidity and direct ADT modality to block androgen stimulation of the prostate by removal of the testes that produce ~90% of androgens. Due to obvious physical and psychological limitations of surgical castration, chemical castration is the preferred ADT in current clinical practice. Chemical castration involves the administration of luteinizing hormone releasing hormone (LHRH) agonists/antagonists that inhibit testosterone production by the testes, or antiandrogens (i.e. first-generation AR antagonists: flutamide, nilutamide and bicalutamide) that block AR activation by effectively competing with testicular or adrenal androgens.11  Approximately 80% of metastatic prostate cancer cases are initially responsive to ADT, as indicated by a decline in PSA levels, palliative effects that reduce pain caused by bone metastases, and treatment-associated progression-free survival of 1 to 2 years. Unfortunately, the effectiveness of this therapeutic option is only temporary due to progression of surviving tumor 6 cells to the castration-resistant phenotype (CRPC), primarily characterized by rise in PSA levels, AR reactivation, and aggressive proliferation.12 Back in 2004, docetaxel was the first approved chemotherapeutic drug for CRPC as it demonstrated significant survival benefits. The treatment of CRPC has since evolved to include novel agents with diverse mechanisms of action in pre- or post-chemotherapy settings. These include immunotherapy Sipuleucel-T (2010), radioisotope Radium-223 (2013) targeting bone metastasis, as well as cabazitaxel (2010) chemotherapy effective in docetaxel-resistant cancer that can minimally extend survival by an average of 2 to 4 months.13,14 The second-generation of more potent and selective AR antagonists, such as apalutamide and enzalutamide (2012), as well as the CYP17A1 inhibitor abiraterone (2011), are the most recently approved drugs by U.S. Food and Drug Administration (FDA) for treatment of metastatic CRPC (mCRPC).14 Abiraterone is an inhibitor of androgen synthesis approved for the treatment of men with docetaxel-resistant mCRPC. Enzalutamide is the most potent second-generation AR antagonist with demonstrated affinity for AR 8-fold higher than that of first-generation bicalutamide.15 In post-chemotherapy mCRPC setting, enzalutamide prolonged overall survival and improved the quality of life of treated patients.16  1.1.3 Androgen Receptor The human androgen receptor (AR), a ligand-activated transcription factor, is the driving force in the development and progression of PCa and constitutes the main drug target for treatment of PCa. Androgen steroids, such as 5α-dihydrotestosterone (DHT) and testosterone, activate AR by binding to the androgen-binding site (ABS) of the receptor triggering its nuclear translocation, binding to DNA and transcriptional activation of AR-responsive target genes (such 7 as NKX3.1, PSA, TMPRSS2 and UBE2C).17 The ABS serves as the primary target of AR with antiandrogens. The AR gene is located on chromosome X and contains eight exons that encode the AR protein (Figure 1.1). Structurally, the AR protein consists of an intrinsically disordered N-terminal transactivation domain (NTD) which is encoded by exon 1, a DNA-binding domain (DBD) encoded by exons 2 and 3, a hinge region that contains the nuclear localization signal (NLS) encoded by the 5’ region of exon 4, and a C-terminal ABS-containing ligand-binding domain (LBD) encoded by exons 5 through 8.18  Figure 1.1 Structure of AR-FL and AR-V truncated variants transcripts and proteins. (A) AR gene structure with canonical exons and cryptic exons (CE and 9a-d). (B) AR-FL and major AR-V truncated variants mRNA and protein structures. NTD (N-terminal domain; exon 1), DBD (DNA-binding domain; exons 2 and 3), H (hinge region; exon 4), LBD (ligand-binding domain; exons 5-8). Inverted triangle represents translational stop. 8 Binding of testosterone and its ~30 times more potent metabolite, DHT to the LBD induces a conformational change of the receptor that leads to dissociation of chaperone proteins and exposes the NLS. AR dimerizes and translocates to the nucleus where it interacts with transcriptional co-regulators, binds to androgen response elements (ARE) and regulates the transcription of a large number of genes involved in cell growth and differentiation in normal prostate and during PCa progression.19-21 Enzalutamide, the most potent among AR antagonists that all bind to the ABS, blocks AR nuclear translocation (a mechanism that differs from that of first-generation non-steroidal antiandrogens) and consequently effectively inhibits AR signaling activation (Figure 1.2).  Figure 1.2 Transcriptional activation by (A) AR full-length and (B) constitutively active AR-V splice isoforms. See text for details.  1.1.4 Drug Resistance Despite the initial efficacy of ADT, tumors recur resulting in the presently incurable CRPC, which often remains dependent on AR activity. Second-generation of more potent and selective 9 AR-directed small molecule drugs, enzalutamide and abiraterone face clinical limitations as drug resistance rapidly emerges and patients die within 2 years.22 Multiple therapy-induced mechanisms of resistance in CRPC have been identified, including AR overexpression, intracellular testosterone production, and imbalance of AR co-regulators, gain-of-function AR mutations and production of constitutively active AR splice variants.23  Activating AR point mutations while rare in untreated PCa are present in ~20% of CRPC patients and in up to 40% of CRPC patients given treatment with AR antagonists.24,25 Over 150 mutations have been identified that generally affect the AR LBD and to a lesser extent the AR NTD26, thus enabling AR activation by weak adrenal androgens and other steroid hormones, including DHEA, progesterone, estrogen and glucocorticoids, and importantly turning AR antagonists into agonists (Table 1.2).27-29 The T878A mutation, for instance, detected in circulating free DNA from patients with CRPC has been associated with resistance to abiraterone and enzalutamide.30 Novel clinically relevant mutations activated by enzalutamide may arise as anticipated by cheminformatics modeling and confirmed by experiment (i.e. T878G).31   Table 1.2 Activating AR mutations recurrent in CRPC Mutation Activation Aberrant effect T878A Activated by enzalutamide, 1st generation antiandrogens, progesterone and estrogen Alters ABS stereochemistry Converts antagonist to agonist  H875Y Activated by enzalutamide, 1st generation antiandrogens, glucocorticoids, estrogen, progesterone Converts antagonist to agonist AR by-pass F877L Activated by enzalutamide, apalutamide and flutamide Reverses anti-androgenic effect of enzalutamide W742C Activated by bicalutamide, flutamide Converts antagonist to agonist after treatment L702H Activated by glucocorticoids AR by-pass 10 A well-described mechanism of resistance to treatment with second-generation AR-directed therapies is aberrant splicing of AR pre-mRNA that yields highly active AR splice variants (AR-Vs) which lack a fully functional AR LBD and thus are unresponsive to the antiandrogen treatments.32 Over 20 AR truncated variants have been identified in PCa cell models and clinical specimens that, with few exceptions, retain an intact AR NTD domain and the AR DBD thus are capable to function as transcription factors. Some of these variants are constitutively active (i.e. able to translocate into the nucleus and initiate transcription without ligand binding; Figure 1.2B), such as AR-V3, AR-V7, and AR-V567es, whereas others are conditionally active dependent on the cellular context, such as AR-V1 and AR-V9. AR variants including AR-V1, AR-V3, AR-V7, and AR-V9 result from splicing to cryptic exons found near the canonical exon 3 of AR gene, inclusion of intronic sequences (AR-V23), exon skipping (AR-V567es), additional exon (AR-V16) or alternative promoter usage (AR-45) (Figure 1.1). Table 1.3 summarizes gain-of-function AR variants recurrently identified in CRPC.33  Table 1.3 Gain-of-function AR splice isoforms recurrent in CRPC Isoform Transcriptional activity Clinical relevance AR-V7 Constitutive Resistance to ADT, enzalutamide and abiraterone. Short time to relapse after surgery and rapid progression to CRPC. Short CRPC-specific survival. AR-V567es Constitutive Resistance to ADT. Enriched in metastases and CRPC. AR-V3 Constitutive Resistance to ADT and abiraterone. Short progression-free survival. AR-V1 Conditional Enriched in metastases and CRPC. AR-V9 Conditional Resistance to ADT and abiraterone. Short progression-free survival.  11 The constitutively active ligand-independent AR-V7 splice isoform is the variant with the highest clinical relevance as it confers primary resistance to enzalutamide and abiraterone, with poor PSA response and short progression-free, overall and cancer-specific survival of CRPC patients.34,35 AR variants may arise by multiple mechanisms including structural rearrangements of the AR gene, aberrant expression of specific splicing factors and involvement of other proteins such as the molecular chaperone HSP90.29,33 Given the clinical importance of gain-of-function AR splice isoforms, various therapeutic approaches aimed at disrupting AR-V signaling have been explored, including targeting the AR NTD or the AR DBD domains, targeting AR-V dimerization with full-length AR (AR-FL), reducing AR-V expression or inducing their degradation, inhibiting AR-V binding to chromatin as well as antisense oligonucleotides. Targeting the AR NTD is one promising approach as it could overcome not only the AR LBD mutational burden but also the drug-resistance of AR LBD-truncated isoforms, in particular that of AR-V7 to enzalutamide. A small number of emerging drug candidates that target AR and its active variants reached clinical trials. Among these, EPI-506 is an orally bioavailable AR inhibitor covalently binding the AR NTD that entered Phase I/II trial in CRPC patients (currently discontinued). Niclosamide, an FDA-approved anti-helminthic drug, which inhibits AR-V7 transcriptional activity, promotes its degradation and in preclinical studies overcomes enzalutamide and abiraterone resistance, entered Phase I trial in combination with enzalutamide in AR-V-positive mCRPC patients. Onalespib is an HSP90 inhibitor that blocks AR-V7 mRNA splicing currently under investigation.29,33 Despite their potential to overcome resistance in CRPC, it remains to be determined whether these inhibitors will advance to late-stage clinical trials and ultimately receive FDA-approval as CRPC drugs. Therefore, development of novel targeted therapies for treatment of CRPC is imperative. 12 1.1.5 AR Independence  AR-independent mechanisms of cancer resistance to continued abiraterone acetate or enzalutamide administration have been described, including glucocorticoid receptor overexpression (capable of taking over the AR transcriptome36), activation of other oncogenic pathways (e.g. TMPRSS2-ERG translocation, PTEN loss or TP53 deficiency, and amplifications of driver oncogenes such as PI3K and MYC)37 and neuroendocrine transdifferentiation (e.g. amplifications of AURKA and MYCN)38,39.  An emerging mechanism of resistance to targeted therapies recently observed in subsets of patients who relapse following AR-directed therapy is lineage plasticity whereby tumors switch from an AR-dependent to AR-independent state and emerge with neuroendocrine features. Neuroendocrine prostate cancer (NEPC) induced by AR-targeted therapy is highly aggressive and has an extremely poor prognosis.40 The incidence of de novo neuroendocrine phenotype in primary PCa is ~1%, whereas in lethal mCRPC rises to 25-30%.41 During NEPC transdifferentiation, cells lose their granular structure and acquire a dominant neuronal morphology, positive for specific neuroendocrine markers including chromogranin A, synaptophysin and neuron-specific enolase but negative for PSA.42 Available treatment options for NEPC rely on chemotherapy with docetaxel and/or cisplatin-based agents, which carry a short-lived response at the cost of significant toxicity.43 Thus, development of novel targeted therapies for emerging oncogenic drivers of NEPC is of paramount importance.    13 1.2 MYC  1.2.1 Discovery of MYC Oncogene Over 40 years have passed since the discovery of MYC, a major oncogene estimated to contribute to most (if not all) human cancers (Figure 1.3).44,45   Figure 1.3 40 years of Myc research. Increase in the number of publications available in NCBI PubMed on Myc in general (blue), Myc in cancer (orange), and Myc inhibition (green) from 1979 to 2018.  1.2.2 MYC Family of Transcription Factors MYC gene encodes the c-Myc protein (hereafter Myc), a nuclear transcription factor frequently amplified in cancers. Myc was originally discovered in Burkitt’s lymphoma, as a homolog to a viral protein that causes avian leukemia.46,47 Its two paralogs, N-Myc and L-Myc encoded by MYCN and MYCL genes, similarly transforming but more tissue-specific factors, were subsequently identified in neuroblastoma and lung cancer, respectively.48-50 14 1.2.3 Physiological Function of MYC Myc functions as a central downstream hub inside the nucleus, integrating signals from numerous upstream pathways to direct gene expression programs and regulate many biological functions, including promoting cell growth, proliferation, apoptosis, metabolism and transformation while blocking differentiation.51-56  1.2.4 MYC and Cancer Sustained proliferative signaling, evasion of tumor growth suppressors, cell death resistance, unleashed replicative immortality, induced angiogenesis, activated invasion and metastasis, genomic instability, metabolism reprogramming, and evasion of the immune system are well-established hallmarks of cancer in which Myc is consistently involved.57   Figure 1.4 Distribution of Myc amplification frequencies among various cancers. Data from cBioPortal for CancerGenomics.58,59 15 A wealth of research documents the critical role of Myc in cancer initiation, maintenance and progression to drug-resistant phenotypes. It has been estimated that Myc contributes to at least 75% of human malignancies, including prostate, breast, colon and cervical cancers, myeloid leukemia, lymphomas, small-cell lung carcinomas and neuroblastoma among others, most of which are aggressive and respond poorly to the current therapies (Figure 1.4).45  1.2.4.1 MYC and Prostate Cancer In prostate cancer (PCa), Myc family members – L-Myc, c-Myc and N-Myc – are implicated in pathogenesis and progression across the full spectrum of PCa, from localized adenocarcinoma to the most advanced and treatment-resistant forms – castration-resistant (CRPC) and its neuroendocrine (NEPC) phenotype. Amplifications of Myc family members are the most frequently observed genomic alterations associated with specific clinical stages and subtypes of PCa.38,42,60-64 L-Myc is amplified in ~27% of localized PCa, in a mutually exclusive manner to c-Myc61, whereas c-Myc is commonly amplified in all PCa stages and subtypes, in particular CRPC.65 Notably, a positive role has been attributed to c-Myc in regulating AR and its splice variants in CRPC. Analysis of gene-expression data from 159 metastatic CRPC samples and 2142 primary prostate tumors showed that the level of c-Myc positively correlates with that of individual AR isoforms.66 shRNA knockdown of c-Myc demonstrated that it positively regulates the coordinated expression of AR and AR-Vs in cell lines and patient-derived xenograft models of CRPC. Furthermore, c-Myc inhibition with a recently reported small-molecule prototype inhibitor sensitized enzalutamide-resistant cells to enzalutamide.66 In addition, it has been demonstrated that c-Myc overexpression upregulates the expression of the hnRNP A1 splicing factor, a direct target gene, that results in increased levels of AR-V7 - the 16 constitutively active, ligand-independent AR splice variant that promotes CRPC67,68 and is also observed in NEPC.63 Importantly, it is established that N-Myc amplifications are strongly associated with the NEPC phenotype induction.38,43,63  1.2.5 MYC Deregulation Myc activity is tightly regulated in normal cells, but in many cancers this control is lost leading to anomalous expression.45 Myc deregulation can occur at any stage of its short molecular life cycle, spanning from replication to transcription, translation and degradation. Mechanisms that account for Myc deregulation include amplifications or chromosomal translocations of the Myc locus, Myc mRNA destabilization and alteration in Myc protein turnover rate. The latter is due to either alterations in Myc protein stability (normally dependent on Myc’s phosphorylation status but caused by mutations in key phosphorylation sites) or to alterations of expression of proteins involved in Myc’s post-translational modifications, such as altered signaling from important ubiquitin-ligase cofactors (that engage the ubiquitin-proteasome system and lead to Myc protein degradation).45,69,70 In addition, Myc’s aberrant expression can occur as a consequence of upstream oncogenic signals (e.g. Ras/MAPK, PI3K, Notch, Wnt) all converging on Myc (Figure 1.5).45,69,70 Myc function is also highly dependent on chromatin context as well as binding partners and effector complexes that modulate various actions of Myc on gene expression.45,69,70 17  Figure 1.5 Myc signaling pathway. Various oncogenic signaling pathways converge on and regulate the expression of nuclear Myc. Myc and partner protein Max form a transcriptional complex that binds DNA and activates transcription of many target genes involved in cell cycle, apoptosis, proliferation and metabolism.  1.2.6 MYC Structure-Function Relationships 1.2.6.1 MYC Gene Organization and Transcriptional Control Human MYC gene, approximately 6 kbases long, found at locus 8q24.21 on chromosome 8, possesses a rather unusual topography. It contains three exons – a large non-coding exon 1, followed by coding exons 2 and 3, four distinct promoters – P0, P1, P2 and P3 – that drive Myc transcription, two major translation start codons (CTG, and ATG) from which two universally expressed Myc proteins arise, two polyadenylation signals and several DNAse 1-hypersensitive 18 sites (Figure 1.6).70,71 P0 transcripts start at multiple initiation sites. P1 and P2 are the two major classical TATA-containing promoter start sites located at the 5’ end of exon 1, with greater than three quarters of Myc transcripts originating from the P2 promoter.71   Figure 1.6 Structure of MYC gene and protein: functional domains and interactors. (Top) MYC locus. (Middle) MYC gene organization. (Bottom) Myc protein domains organization. A variety of proteins that regulate Myc activity and stability interact with these domains. The major Myc protein product, 439 amino acids long, is shown. Not drawn to scale. See text for details.   The Myc promoter region is a key convergence node intricately regulated by a variety of signaling pathways, transcription factors, cis-regulatory elements, chromatin remodeling and by its auto-suppression. The complexity of the Myc promoter control has been comprehensively described in a recent review71. Herein we present only the main highlights. 19 Nearly every major signal transduction pathway that controls cell proliferation or quiescence affects the Myc promoter and regulates Myc transcription, either directly or indirectly. In turn, these pathways are activated by a broad range of signaling molecules, including mitogens, growth factors, hormones, cytokines, oncogenes and tumor suppressors.71 No single regulatory pathway accounts for the activation of the Myc promoter. On one hand, the pathways can cross talk, are somewhat redundant and show variability dependent on cell type and cellular context. While essential for normal Myc regulation in minimizing undesired Myc expression, the integrated biological output from the multiple-input signaling networks poses the risks of driving pathological conditions.71 On the other hand, more than 30 transcription factors (such as E2F, SP1, β-catenin/TCF-4, Smad3, NF-kB, STAT3, ER and AR) bind to the Myc promoter at distinct cis-regulatory sequences. These transcription factors act as integrative nodes at the Myc promoter and as effectors from the different signaling pathways, thus mediating the regulation of Myc transcription in response to various proliferative and anti-proliferative signals.71  Two important cis-elements lie upstream of the P1 and P2 promoters – the far upstream sequence element (FUSE) and nuclease hypersensitivity element III 1 (NHEIII1) – that can form non-canonical DNA structures (such as repressive G4 configurations) and control Myc transcription in enthralling ways.71 A well characterized nucleic acid regulatory sequence, located -142 to -115 base pairs upstream of MYC’s P1 promoter, is the NHEIII1 element that controls 80 to 90% of Myc expression.72 Its guanine-rich (G-rich) strand contains a 27 base pair sequence, 5′-TGGGGAGGGTGGGGAGGGTGGGGAAGG-3′, termed Pu2773, which is comprised of five consecutive G-tracts that are known to adopt G-quadruplex structures (G4), alternative non double-stranded-B-form DNA (dsDNA) configurations.74 The typical G4 20 globular fold consists of three or four stacked guanine tetrads (G-tetrad), three in MYC G4 connected by interceding loops formed by up to three adenine or thymine nucleotides. The G-tetrad, the basic unit of a G4, is in turn comprised of four in-plane G bases paired via Hoogsteen-type hydrogen bonding and stabilized by a central monovalent cation, commonly K+ or Na+, of which potassium is favored due to a stronger coordination at the interface of two G-tetrads.75 Three different G4 topologies have been observed: parallel, antiparallel, and hybrid or mixed backbone. G4s can arrange in intramolecular (monomeric) or intermolecular (multimeric) structures that are dependent on the number of nucleic acid strands involved in their formation.75 In physiological relevant solutions containing K+ ions, the wild-type Pu27 of MYC has been shown to form a dominant intramolecular “propeller-type” parallel-stranded G4 structure arranged from G-tracts 2-5 at the 3′ end of the sequence.74 The NMR study of Ambrus et al76 further demonstrated that the biological relevant conformation of MYC G4 consists of a truncated 22 nucleotide sequence with 3 G→T mutations, 5′-TGAGGGTGGGTAGGGTGGGTAA-3′, termed Pu22 (PBD ID: 1XAV). An alternative G4 formed from G-tracts 1-4 at the 5′ end has also been resolved (PDB ID: 2LBY).77 MYC G4 formation occurs due to negative supercoiling78, preventing the binding of trans-regulatory proteins: the Sp1 transcription factor that binds dsDNA, and hnRNP K and CNBP single-stranded DNA (ssDNA) binding proteins, thus effectively switching off the MYC promoter and attenuating MYC transcription.73 The formation and stabilization of MYC G4 structures has been shown to be facilitated and modulated by two additional critical regulatory proteins, nucleolin and NM23-H2.73 Moreover, Myc transcription is extensively regulated via chromatin remodeling and depends on the presence or absence of particular nucleosomes, their histone acetylation or methylation patterns, as well as the DNA methylation status 71.  21 Feedback loops to most (if not all) systems regulated by Myc, including Myc auto-suppression, provide important mechanisms for control of Myc expression.71 The Myc protein directly or indirectly affects its own expression level. The direct regulation involves a negative feedback loop where Myc represses its own major P2 promoter at the level of transcription initiation and concordantly the Myc promoter is occupied by Myc itself. Myc auto-suppression requires its heterodimerization with its obligate partner Max, but does not occur via binding of Myc-Max to the specific E-box recognition sequence, as the targeted promoter region lacks canonical 5’-CACGTG-3’ sequences, but instead occurs via binding to the Inr (initiator) element mediated by Inr-binding and E2F transcription factors.71 Indirectly, Myc acts as both an activator and repressor of its own activators and repressors.71 Additional factors that trigger the Myc-driven exacerbated cellular proliferation and transformation evidenced in cancer have been described in recent publications70,71. Post-transcriptional deregulatory events further contribute to transforming Myc70 and two mechanisms of Myc mRNA turnover have been reported. The first is translation-independent, involving poly(A) tail shortening regulated by AU-rich sequences in the 3’ untranslated region.79,80 The second is a translation-dependent mechanism regulated by a region of mRNA corresponding to the C-terminal domain of the coding region determinant-binding protein (CRD-BP).81 Importantly, stabilization of Myc mRNA by CRD-BP accounts for the increased mRNA stability observed in human cancers.82  1.2.6.2 MYC Protein Organization and Translational Control The major Myc protein product, a protein sequence of 439 amino acids, that migrates as p64 (i.e. at 64 kDa) starts with the ATG codon at the 5’ end of exon 2, while the minor protein 22 product p67 having 14 additional N-terminal amino acids starts with the CTG initiation codon at the 3’ end of exon 1.70,71 Human Myc contains several highly conserved regions that are functionally important and organized in the same fashion among the three paralogs (c-Myc, N-Myc and L-Myc).69 All Myc proteins possess a largely disordered N-terminal transactivation domain (TAD), a 143 amino acid domain that is required for transcriptional and cell-transforming activity and contains conserved functional modules termed Myc boxes (MBI, MBII) (Figure 1.6 bottom). MBI serves as a phosphodegron and is involved in the ubiquitination and proteasomal degradation of Myc.69  Myc family members are very unstable with half-lives of 20-30 min in normal cells; however, in many tumors stabilization of Myc contributes to its deregulation. Multiple ubiquitin ligases control Myc stability. The E3 ubiquitin ligase FBW7 (F-Box and WD repeat domain containing 7) binds to MBI and regulates c-Myc and N-Myc in response to phosphorylation of Ser62 (stabilizes Myc) and Thr58 (destabilizes Myc) residues (these two major phosphorylation sites are indicated in Figure 1.3 above the MBI box). Mutation of Thr58 has been reported to occur in ~1/2 of Burkitt’s lymphoma cases, leading to Myc stabilization due to impaired proteasomal degradation and evasion of apoptosis.83-85 Loss of FBW7 has also been reported to result in decreased Myc turnover in a large number of tumors.86 MBII, the most studied region within Myc TAD, is important for most Myc activities and functions as a hub for binding to multiple key interactors including TRRAP (transformation/transcription domain associated protein) involved in chromatin-dependent Myc transcriptional signaling.87 TRRAP recruits chromatin-remodeling complexes including histone acetyltransferases (HAT), such as GCN5 and Tip60, to promote gene activation. Acetylated, open chromatin is bound by bromodomain-containing proteins, such as BRD4, and other coactivators of the bromodomain and extra 23 terminal (BET) family, which recruit the positive transcription elongation factor b (P-TEFb) complex at target genes promoters, activate the kinase activity of P-TEFb which phosphorylates the C-terminal domain of RNA polymerase II (POLII), causing pause release and leading to transcriptional elongation.88 Moreover, MBII is involved in Myc protein turnover, as is a docking site for SKP2, the E3 ubiquitin ligase component S-phase kinase-associated protein 2 that, in addition to FBW7, is involved in the degradation of Myc.89,90 SKP2 effects on Myc activity are briefly discussed below. The mechanisms underlying MYC degradation have been reviewed in detail by Farrell et al.91  Myc proteins also contain a central segment, enriched in proline, glutamic acid, serine and threonine (PEST) residues, which is necessary for rapid Myc degradation but not ubiquitination.84 Two additional conserved Myc boxes are also present: MBIII is important for transcriptional repression as it provides docking sites for components of histone deacetylase repressor complexes, such as SIN3 and HDAC392,93, and MBIV involved in Myc transcriptional activity and Myc-induced apoptosis. These boxes have been reported to interact with additional effectors including CREB-binding p300/CBP transcriptional co-activators, and WDR5 (WD repeat-containing protein 5) which stabilizes Myc interactions with chromatin to promote target gene recognition and Myc-driven tumorigenesis.94 Moreover, MBIV interacts with p27 (cyclin-dependent kinase inhibitor p27KIP1), one of the gatekeepers of G1-S transition of the cell cycle. p27 represses Myc and blocks Myc’s phosphorylation at Ser62 thus decreasing its activity. p27 is an essential target of SKP2 which triggers its ubiquitination and proteasomal-mediated degradation, hence relieving p27 repression of Myc and leading to transcriptional activation.95 The calpain cleavage site (CAPN in Figure 1.6) is involved in cytosolic Myc partial cleavage of the C-terminus, resulting in “Myc-nick”, a 298 amino acid N-terminal segment that 24 inactivates Myc transcriptional activity.96-98 The nuclear localization sequence (NLS) is also implicated in MYC cellular-transforming activity, transcription, and apoptosis.99 The C-terminal region of Myc proteins, ~100 amino acids in length, comprises the basic, helix-loop-helix, leucine zipper (bHLHLZ) dimerization and DNA-binding (DBD) domains. The most important interactor at the C-terminal region is Max and as already mentioned, is mandatory for Myc transcriptional activation. An additional interactor with Myc’s C-terminus is Miz1 (Myc-interacting Zn-finger protein 1) with role in Myc transcriptional repression. Myc repression involves the loss of Myc binding to E-box by displacement of Max by other bHLHLZ proteins (see section 1.2.6.3) allowing Myc to associate with and sequester Miz1, a POZ domain-containing zinc finger protein that induces G1 cell cycle arrest.100,101  Other important interactors with Myc C-terminus are ARF and SKP2. The ARF tumor suppressor antagonizes SKP2-mediated ubiquitination, inhibits Myc transactivation, proliferation, and transformation, and promotes Myc-induced, p53-independent apoptosis.102 SKP2 recognizes Myc through both MBII and bHLHLZ motifs to promote Myc poly-ubiquitination and degradation. While SKP2 decreases Myc protein stability and stimulates its degradation, it has an opposite effect on Myc transcriptional activity, promoting it instead of inhibiting it as does FBW7.91 SKP2 is a direct target gene of Myc, which augments its expression. As such, SKP2 may contribute to oncogenesis by both enhancing Myc transcriptional activity and regulating its protein level.91  Of note, Myc cofactors and their Myc-associated functions have been extensively covered in recent reviews103,104.  25 1.2.6.3 MYC/MAX Extended Network Myc and Max belong to an extended network of related bHLHLZ transcription factors that function as regulators of different aspects of cell behavior as they mediate a broad transcriptional response to diverse signals, including mitogenic, growth arrest, and metabolic stimuli.69,100,105,106   Figure 1.7 The extended Myc-Max network. The players and their regulatory functions. The network consists of three layers from left to right – Myc family; Mxd family/Mnt/Mga; and MondoA/MondoB centered on the network two main nodes: Max and Mlx. Double-headed arrows indicate heterodimerization interactions between network players. The resulting heterodimers either activate or repress transcription by binding to E-box DNA sequences.  The transcriptional network is centered on two nodes with the Max binding proteins forming one node, whereas Mlx binding proteins forming a second node. Beside the Myc family members, the Max-centered network includes members of the Mxd (originally called Mad) family of bHLHLZ proteins, Mxd1-4, and the more distantly related Mnt (Figure 1.7). In addition, Max binds to Mga, the largest protein in the Max network and a “dual-specificity” 26 transcription factor in that possesses not only the bHLHLZ Max DNA binding motif but also a T-domain DNA binding motif.101   Figure 1.8 Structure-function schematic of the members of the extended Myc/Max network. Known functional domains of different network members are summarized. TAD: transactivation domain; DBD: DNA binding domain; P: position of known phosphorylation sites; SID: mSin3 interaction domain; TRD: trans-repression domain; T-domain: T-box DNA binding domain; DCD: dimerization and cytoplasmic localization domain. MCR: MondoA conserved region. Not drawn to scale. Details provided in the text.  The Mxd family members act as antagonists of Myc function. Mxd proteins like Myc do not homodimerize nor bind to DNA by themselves. Instead, Mxd and Mnt form heterodimers with Max and act as transcriptional repressors competing with Myc-Max complex for binding at same promoter-proximal E-boxes. The transcriptional suppression by Mxd and Mnt stems from their ability to bind to the large corepressor Sin3 histone deacetylase complex (Figure 1.8).69,100,105,106 All Mxd and Mnt proteins possess a conserved amino acid sequence near their N-27 terminus, termed the mSin3 interaction domain (SID) that directly interacts with one amphipathic α–helical (PAH) domain within Sin3.107 Sin3 interacts with class I histone deacetylases (HDAC1 and HDAC2) leading to transcriptional silencing.107 While the Myc triad associates with the TRRAP-GCN5 coactivator complex, the Mxd family members recruit the Sin3-HDAC corepressor complex, suggesting their antagonistic behavior.69 The opposed transcriptional activities of Myc and Mxd families have been attributed to three overlapping mechanisms: competition for available Max to form heterodimers, competition of heterodimers for E-box DNA binding sites, and activation or repression of bound target genes.69,100,105,106  The Mga member of the extended network has been suggested to act as a tumor suppressor, with inactivating mutations detected in leukemia.108 The Mnt factor antagonizes both Myc-stimulated proliferation and apoptosis, its pro-survival function being critical for Myc-driven tumorigenesis consistent with Mnt’s dominant physiological activity of opposing the pro-apoptotic activity elicited by Myc 54. It has been demonstrated, that the loss of Mnt induces similar effects as Myc overexpression, such as enhanced transcription of Myc target genes resulting in accelerated proliferation, apoptosis, and transformation 109,110. It has been proposed that Myc functions by relief of Mnt repression.110,111 Max can form unstable homodimers and unlike Myc lacks a transactivation domain (Figure 1.8). At physiological levels, Max homodimers fail to regulate transcription, but Max overexpression can lead to transcriptional repression.112,113 Overexpressed Max has been shown to reduce Myc-induced carcinogenesis114,115 and in human cancer, higher Max levels have been associated with better prognosis.116 Max protein has several isoforms generated by alternative splicing. Besides the dominantly expressed Max isoform (i.e. 160 amino acids in length), ΔMax lacks the C-terminal 61 amino acids which are replaced by five residues before ending with an 28 alternative exon (Figure 1.8).117,118 Max phosphorylation (Figure 1.8) blocks Max homodimerization, but not heterodimerization with Myc. ΔMax is not phosphorylated and dimerizes with Myc augmenting its transforming activity.118 The network extended further with the discovery of a Max-like bHLHLZ protein, Mlx, as a dimerization partner for a subset of Mxd family members including Mxd1, Mxd4, and Mnt (Figure 1.7). Mxd-Mlx heterodimers interact with Sin3, bind E-box DNA sequences, and repress transcription similarly to Mxd-Max dimers (Figure 1.8).100,105,106 While Mlx does not associate with Max or Myc family members, it dimerizes with two others partners, MondoA (MLXIP) and ChREBP (MondoB or MLXIPL), which are cytoplasmic-nuclear shuttling proteins whose accumulation in the nucleus is triggered by glucose-derived metabolites.106,119 MondoA-Mlx and ChREBP-Mlx heterodimers bind E-boxes, act as nutrient-sensing transcription factors and regulate genes involved in glucose and glutamine metabolism both fundamental biological processes in both normal and cancer cells.119 The Myc triad does not dimerize with any member of the network other than Max, while only Max and Mlx can form homodimers. Moreover, only Myc proteins and MondoA carry a TAD to transactivate target genes. In terms of dimerization with Max, Myc and Mxd network members bind Max with different efficiencies, and distinct subcellular localization patterns between Myc-Max and Mxd-Max have been reported.120 Recent studies revealed that Myc and Mnt compete for binding to limited amounts of Max whose availability is subsequently modulated by the turnover of Mxd proteins, typically displaying short half-lives, very much alike Myc, due to ubiquitin-mediated proteasomal degradation.69,100,105,106 In contrast, MondoA and B are stable proteins and tight regulation of transcriptional activity occurs through their nuclear accumulation in response to changes in metabolic flux.69,100,105,106,119 The observed changes in the 29 abundance of individual network players have functional consequences attributed to the competition for available Max and Mlx core members as well as for DNA-binding sites, and may account in part for the extraordinary tight regulation of Myc expression.69,100,105,106  1.2.6.4 Structural Aspects of Protein-Protein and Protein-DNA Interactions within the Myc/Max Extended Network and Beyond Available crystallographic data provides significant insights into structural determinants of molecular interactions that govern the assembly of homo- and heterodimers within the extended Myc network, as well as specific DNA recognition events.121 The first reported X-ray structure of bHLHLZ domain of the Max homodimer bound to the 5’-CACGTG-3’ E-box DNA sequence (PDB ID: 1AN2) revealed the overall topology of the bHLHLZ domain (Figure 1.9) and established the structural bases for DNA recognition.122 Ten years later, the X-ray structures of Myc-Max (PDB ID: 1NKP) and Mad-Max heterodimers (PDB ID: 1NLW) were determined re-enforcing the secondary structure composition of the Max homodimer, featured in Figure 1.9.123  30  Figure 1.9 Structural Aspects of Protein-Protein and Protein-DNA Interactions within the Myc/Max Extended Network and Beyond. The resolved structures demonstrated that Myc/Max/Mad bHLHLZ domains consist of two contiguous α-helixes separated by a random loop. The first α-helical ordered element comprises residues from the basic (b) region and helix 1 (H1). The conserved proline residue at position 30 in the multiple alignment (shown at the top of Figure 1.9) terminates H1 and drives the loop (L) formation connecting the two structured α-helical segments, the latter being composed of the helix 2 (H2) and the leucine zipper (LZ) regions.121-123 The Max homodimer and the Myc- and 31 Mad-Max heterodimers consist of two bHLHLZ monomers that fold into a parallel, four-helix bundle. The two basic regions, protruding from the N-termini of the bundle, insert into a modified B-form DNA conformation, characterized by a narrowed major groove and a widened minor groove, to make sequence specific contacts with the cognate E-box.121-123 The two helical segments at the C-termini form the parallel, coiled-coil or LZ dimerization domain. A well-defined globular core is formed by conserved hydrophobic residues residing in the H1 and H2 helices of the four-helix bundle known to stabilize the Max-Max homodimer, distinguishing the bHLHLZ domain from that of purely coiled-coil LZ proteins.124 Substitutions of Myc HLH and LZ motifs with the corresponding regions of structurally related proteins that do not interact with Max demonstrated that all of the conserved hydrophobic amino acids within H1 and H2 (Figure 1.9) in addition to the LZ motif are required for specific Myc-Max dimer formation.125  Extensive hydrophobic and polar interactions between the HLH and LZ regions stabilize both Max-Max homodimers as well as quasi-symmetric Myc-Max and Mad-Max heterodimer structures that noticeably differ in the corresponding coiled-coil LZ regions. The Max-Max complex contains a packing defect introduced by the charge-neutral Gln91-Asn92/Gln91-Asn92 pairing tetrad occurring at the C-termini of both Max monomers (Max monomer numbering) that promotes Max-Max homodimerization, in contrast to the positively charged pairing Arg423-Arg424/Arg423-Arg424 tetrad that disfavors the formation of the Myc-Myc homodimer. In the quasi-symmetric coiled-coil structures of Myc-Max and Mad-Max, the defect is compensated by charge complementarity at equivalent residues, with H-bonding interactions between Gln91-Asn92 pair with the positively charged Arg423-Arg424 pair (Myc monomer numbering) in Myc-Max heterodimer, and with Glu125-Gln126 pair in the Mad-Max heterodimer, respectively, resulting in the tighter packing of the heterodimers (Figure 1.9 green boxed residues).121-123  32 The Max-Max, Myc-Max and Mad-Max X-ray structures (Figure 1.9) further reveal three main segments responsible for specific DNA recognition at residue-level: residues from the basic and loop regions and the first residue of H2. Three invariant residues within the basic region make base-specific contacts with the DNA 5’-CACGTG-3’ recognition sequence. These are His, Glu and Arg residues at positions 28, 32, and 35 in Max monomer-based numbering, at positions 359, 363, and 367 in Myc monomer-based numbering, and at positions 61, 65, and 69 in Mad. Equivalent residues following the numbering in the Myc-Max dimer structure are His207, Glu211, and Arg215 in Max, and His906, Glu910, and Arg914 in Myc. These critical residues are highlighted in Figure 1.9 with green dots below the multiple sequence alignment. One H-bond between the conserved His and the central guanine of the E-box dictates the specificity for a purine base at that position. The invariant Glu makes two H-bonds with the adenine and the cytosine at positions 2 and 3 in the E-box sequence. Evidence shows that substitutions of the Glu to Gln, Asp, or Leu abolish DNA binding.126 The location of the Glu in the DNA major groove made DNA recognition by the shorter acidic side chain of aspartate incompatible.126 The conserved Arg H-bonds the central guanine and in addition interacts with the phosphate group between the cytosine and the adenine at the first and second positions in the hexanucleotide. As such the Arg dictates the identity of the central 5’-CG-3’ dinucleotide and the specificity for bHLHLZ proteins that bind the canonical class B E-box from those which have a hydrophobic amino acid at that position that bind non-canonical class A site, 5’-CAGCTG-3’.126  It has also been demonstrated that the substitution of this conserved Arg for Met suffices to convert some class A proteins (e.g. AP4, homologous bHLHLZ transcription factor) into a canonical class B E-box binding protein.127 The basic region also makes a large number of contacts with phosphate groups spanning the entire backbone of the E-box. In the Myc-Max 33 structure, residues specific to Myc including Lys902 and Arg903, make additional contacts with the phosphate backbone of the DNA. In addition, non-specific DNA contacts are contributed from Lys918 in the H1 region and Lys936 in loop region. Lastly, specific contacts are contributed from Lys939, the first residue in H2. Substitutions of its equivalent Max residue, Arg239, for Glu or Ala abolished both dimerization and DNA binding. Arg239 makes both side-chain and backbone amide contacts with DNA phosphate groups and as such tethers H2 to DNA and stabilizes the interactions between H1 and H2 by packing against the conserved Phe222 (Phe921 in Myc) located in H1 122,123. Figure 1.10 illustrates the complexity of Myc-Max/DNA recognition interface. Only the Myc-Max heterodimer but not the Mad-Max counterpart was shown to form a bivalent hetero-tetramer capable of upregulating gene expression at distant promoters bearing widely separated E-boxes.123  Figure 1.10 Myc-Max/DNA 5’-CACGTG-3’ E-box recognition interface (PDB ID: 1NKP, 1.9 Å resolution). Specific hydrogen bonding interactions are indicated with yellow lines.  34 Further structural evidence on the specificity of DNA recognition as well as dimerization determinants came from the recently resolved X-ray structure of the Omomyc homodimer.128 The comparison between the protein-DNA interfaces of Omomyc (reported by Jung et al128) to that of the Myc-Max DNA complex (reported by Nair et al123) demonstrated that both complexes bind to the DNA major groove with alike-formed scissor structures at the E-box, and that the basic region of both complexes assume the same phosphate-backbone and base-specific contacts with DNA. These contacts recapitulated the three invariant residues His, Glu and Arg at positions 12, 16, and 20 in Omomyc as critical DNA recognition points. Omomyc is a dominant negative 93 residue bHLHLZ Myc protein fragment specifically designed to introduce 4 single-point mutations in the dimerization domain that correspond to four charged amino acids shown to prevent Myc homodimerization due to major steric and electrostatic clashes.129 The Arg423-Arg424 pair (above-mentioned) that disfavors Myc homodimerization due to charge repulsion is replaced in Omomyc with the residues found in the Max sequence at these positions that favor Max homodimerization, instead: a glutamine and an asparaginine, respectively. In addition, two glutamic acid residues at positions 410 and 417 in Myc are replaced in Omomyc by a threonine and isoleucine, respectively (with the latter also found in Max at the same position). A mutant with all four amino acids substituted with the Max-specific residues would homodimerize only weakly due to an unfavorable shape complementarity of amino acids around position 410.129 These mutations are highlighted in Figure 1.9 with red dots bellow the multiple sequence alignment. Omomyc preferentially forms homodimers stabilized by multiple interactions not present in Myc-Max complex, and introduced by specifically designed mutations.128 Differences in protein-protein interactions between Omomyc and Myc-Max dimers extend beyond the mutated residues 35 to the entire interface, making Omomyc homodimer more stable than Myc-Max by an estimated free energy difference of -9 kcal/mol.128 Omomyc can form dimers with both Myc and Max but, due to repulsive interactions or lack of stabilizing interactions, they are less stable.128 Omomyc homodimers effectively compete with Myc-Max heterodimers for binding to DNA on low-affinity promoters invaded by oncogenic Myc levels, while they do not outcompete binding of Myc-Max to promoters that are highly occupied at physiological levels of Myc.128 In the latter case, binding of Myc-Max to chromatin appears to be stabilized by protein-protein interactions involving Myc transactivation domain.128 Omomyc effectively sequesters Myc away from the DNA and occupies the E-box with transcriptionally inactive dimers (i.e. Omomyc/Omomyc and Omomyc/Max).  Importantly, Omomyc can effectively interfere with the oncogenic function of Myc by inhibiting gene expression characteristic of Myc-dependent tumors.128 It has been demonstrated that Omomyc selectively targets Myc protein interactions in that it binds c- and N-Myc, Max and Miz-1 but does not bind Mad or others HLH proteins.130 Moreover, Omomyc specifically prevents Myc binding to promoter E-boxes and transactivation of target genes while retaining Miz1-dependent binding to promoters and transregression, mechanisms of action accompanied by broad epigenetic changes.130 Thus, in presence of Omomyc, the Myc interactome is directed toward repression and its activity is switched from an oncogenic to a tumor suppressive one.130  1.2.7 Drugging MYC  Myc is widely recognized as one of the most valuable targets in cancer, and the corresponding drug would make a very substantial impact in the clinic; however, for many years Myc has been deemed as “undruggable”.131  36  1.2.7.1 Challenges and Opportunities One challenge associated with drugging the “undruggable” Myc relates to undesirable side effects that Myc inhibition may cause in normal cells, negatively affecting its physiological functions.132 Nonetheless, Omomyc provided early proof-of-concept evidence that inhibiting Myc-Max interactions and its transcriptional output may be an effective therapeutic strategy. Omomyc established the feasibility of intermittent systemic inhibition of Myc-Max protein-protein and protein-DNA interactions for it demonstrated efficacy against tumors with no toxicity to normal tissues, conferring indefinite survival in animal models.132,133 As such, Omomyc alleviated the major concern about the undesired, deleterious side effects that Myc inhibition might have on healthy proliferating tissues. Moreover, Myc is an intrinsically disordered protein (IDP) that transiently acquires minimal secondary structure and exists as a “protein cloud” in a dynamic ensemble of unstable conformations, with no effective pockets on its surface.134 The structural disorder of Myc makes it difficult to characterize its interactions with ligands using experiments alone and importantly, is an inherent challenge in applying conventional structure-based drug design approaches to target its disordered structure.135 Furthermore, Myc function is dependent on dimerization with Max (which is also an IDP), that together form a functional mandatory DNA-binding domain (DBD), through which the Myc-Max complex binds specific DNA recognition sequences 5’-CACGTG-3’, termed E-boxes or enhancer-boxes, at enhancers and promoters of target genes. This event triggers the recruitment of chromatin-remodeling complexes and assembly of the transcriptional machinery thereby switching on more than 15% of the human genome on one hand, and driving oncogenic 37 transformation on the other.136 The Myc-Max complex also binds DNA sites that vary from the palindromic hexanucleotide canonical sequence, not bound with equal affinities. Non-canonical sequences such as 5’-CACGCG-3’ and 5’-CATGGC-3’ represent low-affinity Myc-Max binding sites.137  1.2.7.2 MYC Targeting Strategies A vast array of targeting strategies, both direct and indirect, have been applied to Myc by exploiting its multiple regulatory mechanisms, including Myc transcription, mRNA stability, Myc protein stability and degradation, as well as Myc binding to its interactome. Some of these approaches yielded prototype inhibitors, several of which entered early clinical trials.138 Examples include inhibitors of Myc transcription with direct G-quadruplex stabilizers, antisense oligonucleotides that induce Myc mRNA degradation, aberrant splicing of Myc pre-mRNA or translation block, as well as short-interfering RNAs.138,139  Since c-Myc and N-Myc are highly similar both structurally and functionally, their targeting approaches are generally undistinguished, with perhaps one exception of indirect targeting of N-Myc protein described in subsection 1.2.7.3. Several recent publications138-140 provide in-depth coverage of strategies targeting both c-Myc and N-Myc, their preclinical stage and clinical applicability.  1.2.7.3 Indirect Therapeutic Inhibition Indirect Myc suppression has been achieved via inhibitors of regulators of Myc protein stability and turnover (e.g. GSK3, Ras/Raf/MAPK, PP2A, FBW7, SKP2, hTERT)138,139, inhibitors of pathways involved in Myc translation (e.g. MAPK, mTORC1 and FOXO3a)139, and 38 inhibitors of Myc chromatin remodeling and transcription of BET bromodomain proteins.141 In the latter category, JQ1 was the first reported compound to inhibit Myc-associated chromatin remodeling enzyme Brd4142, followed by novel BET inhibitors, such as ZEN-3694 that entered clinical trials and demonstrated efficacy in a variety of solid tumors and hematological malignancies, alone or in combination with several standards143, and more recent OTX015144 and TEN-010.145 Indirect suppression of MYC transcription and destabilization of the Myc protein in human Burkitt’s lymphoma has been recently achieved by targeting the Myc-HSP90 axis with HSP90 inhibitors (i.e., 17-AAG or 17-DMAG).146 Another potentially effective strategy is indirect targeting of N-Myc stability and turnover by antagonizing Aurora kinase A (AURKA) with small-molecule inhibitors that block AURKA/N-Myc interactions and promote N-Myc degradation. MLN8237 (Alisertib) and CD532 kinase inhibitors induce an allosteric transition in AURKA that results in conformational changes that destabilize N-Myc and triggers its phosphorylation at the N-terminus, leading to ubiquitylation and degradation by FBW7 ubiquitin ligase.147-149   1.2.7.4 Synthetic Lethality Indirect synthetically lethal approaches have also been reported. The synthetically lethal compound dihydroartemisinin (a common metabolite of the highly potent and safe anti-malarial agent, artemisinin) was found to activate the Ser/Thr kinase GSK3β that in turn phosphorylates and destabilizes Myc.150  39 1.2.7.5 Direct Therapeutic Inhibition Although significant efforts have been made in the past 20 years, no approved small molecule drugs have been developed that directly block Myc-Max interactions or binding of the complex to DNA. A number of structurally diverse prototype inhibitors have been reported but they all exhibited lack of in vivo efficacy and/or suboptimal safety.138,151 On one hand, these shortcomings may arise from the lack of rational, structure-guided efforts, as typically such chemicals emerge from high-throughput wet libraries screenings. On the other hand, the disordered nature of Myc protein and the absence of high-quality structures of ligated Myc-Max complexes further restrain therapeutic development efforts.   1.3 Current Advances in the Development of Direct MYC Inhibitors All known Myc-Max small-molecule inhibitors (described in detail in section 1.3.2) fall in two categories:  1. those that act by interfering with protein-protein interactions and block heterodimerization of Myc with Max; and, 2.  those that directly block Myc-Max binding to DNA.  Importantly, both types of inhibitors reduce the abundance of the Myc protein and inhibit the proliferation of several human cancer cell lines where they provoke an energy crisis marked by ATP depletion, neutral lipid accumulation, AMP-activated protein kinase activation, cell-cycle arrest and apoptosis.152 Another promising approach is targeting the morphing DNA topologies within the cis-regulatory sequences upstream of the MYC promoter that regulate MYC gene expression (described in section 1.3.1 below). 40 1.3.1 Direct Inhibitors of MYC Expression The presence of repressive G4 structures in the MYC promoter region provides attractive opportunities for direct MYC inhibition with small molecules that could specifically trigger G4 formation and stabilization, thereby downregulating the transcription of MYC gene. The approach faces similar, if not greater, challenges as those encountered with targeting the Myc-Max protein complex and its transcriptional function: the limited availability of crystallographic data for nucleic acid-ligand complexes, and importantly, the selectivity for MYC G4 relative to other G4-driven oncogenes leading to off-target side effects. It has been shown that, besides MYC, other major cancer driver genes form parallel G-quadruplex structures (G4s), including Bcl2, VEGFA (vascular endothelial growth factor A), c-KIT (KIT proto-oncogene receptor tyrosine kinase), and HIF1α, while other G4 types form in the promoter regions of KRAS, RB1, hTERT, and PDGFA genes, as well as at telomeric ends (e.g., h-Telo) and ribosomal DNA. All G4s differ in their folding patterns, number of G-tetrads, loop length, and composition.75,153  A large number of G4 inducers and stabilizers have been developed in the past 15 years mainly through chemical design and synthesis evaluated by a variety of biochemical/biophysical and biological assays for determination of their direct binding to G4, stabilization efficiency, key structural ligand-G4 interactions, and effects on gene expression.153,154 Three major types of G4 ligands have been described based on the structural organization of the aromatic rings of which they are composed: fused heteroaromatic polycyclic systems; macrocycles; and, non-fused or modular aromatic compounds.154 Three binding modes for G4 ligands have also been described: external stacking attributed to π–π stacking interactions occurring at the external end of G4, the most energetically favored mechanism; intercalation between 2 G-tetrads; and, groove or loop binding.155 The design of modular G4 ligands was motivated by the need for more drug-like and 41 selective compounds targeting the diverse loop and groove regions of G4s and not only the external end G-tetrads of G4s, the preferential binding mode for fused polycyclic and macrocycle ligands, as observed from available X-ray and NMR structural data of G4-ligand complexes.75,153,154 The diversity of G4 ligands include perylene G4-inducer compounds, such as PIPER and its later synthesized derivatives, synthetic dyes, and natural alkaloids. Stronger binders reported to act as both inducers of G4 formation and importantly as G4 stabilizers include cationic porphyrins, quindoline derivatives, and metal complexes.155 The representative cationic porphyrin TMPyP4 (Figure 1.11), an end stacker binding the major parallel MYC G4, provided early proof of principle that stabilization of G4 structures could silence MYC transcription given its effects in Burkitt’s lymphoma Ramos and CA46 cell lines.73 A significant effect was observed in Ramos cells, which preserve the MYC G4 in the NHEIII1 sequence after the well-described chromosomal translocation between chromosomes 8 and 14 (which puts MYC transcription under the control of an immunoglobulin heavy-chain gene enhancer), but not in CA46 cells in which NHEIII1, along with exon 1 (containing MYC’s promoter sequence) are deleted during translocation.73 The observed repression of MYC transcription was further substantiated by a reported decrease in the unfolding of MYC G4 driven by NM32-H2 at increasing concentrations of TMPyP4.156,157 Unfortunately, TMPyP4 has also been shown to stabilize other G-rich sequences, as well as i-motif structures that form on the C-rich strand of the NHEIII1 element of MYC, and to convert the parallel MYC G4 to a mixed parallel/antiparallel type.158 To address the poor selectivity of TMPyP4, Seenisamy et al.159 further designed and synthesized Se2SAP (Figure 1.11), an analog of TMPyP4 with an expanded porphyrin core, which, in comparison to the parental compound, was less toxic and it was able to convert the parallel topology of MYC 42 G4 to a single loop hybrid G4. Se2SAP strongly and selectively stabilized the single external lateral loop of the hybrid relative to other G4-forming sequences (e.g., double-loop hybrid induced by TMPyP4), telomeric G4s, and double- and single stranded DNA, thus demonstrating selectivity for one particular G4 structure.  Figure 1.11 Chemical structures of TMPyP4 and Se2SAP cationic porphyrin G-quadruplex (G4) stabilizers.73,159  The quindoline derivative SYUIO-05 (Figure 1.12), a fused heteroaromatic polycyclic and intercalating compound, was shown to preferentially stabilize MYC G4 over telomeric G-quadruplexes, as determined by FRET-melting, PCR stop assay, and isothermal titration calorimetry (ITC).160 SYUIO-05 significantly arrested cell proliferation of several cancer cell lines and downregulated MYC transcription. Moreover, the compound interfered with the binding of NM23-H2 factor to the MYC G4.160 Several metal complexes have been reported as G4 stabilizers with weak affinity for dsDNA, including platinum II containing compounds161,162, which are nevertheless prone to additive toxicity and treatment-induced resistance when used as chemotherapeutic agents163. Overall, the aforementioned G4 stabilizers have demonstrated efficacy against tumor growth in cancer cells. 43  Figure 1.12 Chemical structure of SYUIO-05 quindoline derivative G4 stabilizer.160  CX-3543 (Figure 1.13), also known as quarfloxin, was the first G4-stabilizer that in 2008 had entered phase II clinical trials for low to intermediate neuroendocrine tumors.164 Its parental compound, the fluoroquinolone QQ58 was originally designed and synthesized by Duan et al.165 as a derivative started from the norfloxin antibiotic and gyrase agent, via a series yielding a closer intermediate quinobenzoxazine A-62176 compound with antibacterial topoisomerase II inhibitory activity73. As QQ58 acted as both a topoisomerase II intercalator and as a G4 interactor, as determined by NMR and molecular simulations, medicinal chemistry efforts that were conducted by Cylene Pharmaceuticals (San Diego, CA, USA) yielded the optimized CX-3543 that demonstrated significant selectivity toward G-quadruplex over duplex DNA with no residual gyrase or topoisomerase II poisonous activity166. While quarfloxin was first considered to be a selective binder of MYC G4, its mechanism of action was later attributed to its preferential disruption of nucleolin binding to ribosomal G4 in the nucleolus, resulting in the inhibition of Pol I transcription and rRNA biogenesis and induced apoptosis in cancer cells.166 CX-3543 had been meanwhile withdrawn from clinical studies of neuroendocrine tumors and its development at Cylene had been discontinued.73 Nonetheless, medicinal chemistry efforts coupled with cell-based and cell-free screens (Pol I transcriptional assay and EMSA among others in the latter category) were further carried out. And in 2011, Cylene Pharmaceuticals reported the development of CX-5461 (Figure 1.13), a potent and selective inhibitor of Pol I-44 mediated rRNA synthesis across a panel of 50 cancer cell lines (average IC50 of 147 nM), with no effect on DNA replication, Pol II-driven mRNA synthesis, or protein translation.167 CX-5461 induced autophagy, but not apoptotic cell death in solid tumor cell lines or in normal cells, as determined by immunocytochemistry-based autophagy and senescence detection assays. Moreover, CX-5461 exhibited potent in vivo antitumor activity in murine xenograft models of pancreatic carcinoma and melanoma upon oral administration at a 50 mg/kg dose once daily or every three days without changes in animal body weight or readily apparent toxicity, positioning CX-5461 for investigational clinical trials.167  Figure 1.13 Chemical structures of CX-3543 (quarfloxin) and optimized CX-5461 G4 stabilizers in early phase clinical trials.164-167  In 2017, Xu et al168 reported a novel mechanism of action for CX-5461 (and the structurally related CX-3543) in that CX-5461 induced DNA damage in DNA-repair deficient cell lines. Moreover, in an in vitro FRET melting assay utilizing three different G4 forming DNA fragments (MYC, c-KIT, and h-Telo), CX-5461 showed strong binding and stabilization of G4 relative to dsDNA. Immunofluorescence experiments with a G4 specific antibody further indicated G4 stabilization in the cellular environment upon treatment with CX-5461 at 45 nanomolar concentrations.168 The CX-5461-induced DNA damage was observed at G4-enriched genomic sequences, the repair of which required breast cancer (BRCA) and non-homologous end joining (NHEJ) pathways.168 Importantly, CX-5461 had a profound anticancer activity in BRCA deficient and chemotherapy-resistant (i.e., to standards of care taxane and cisplatin) triple negative breast cancer patient-derived xenografts (PDX) tumors.168 Currently, CX-5461 is in phase I clinical trials for BRCA1/2 deficient breast tumors.169 Attempts to discover selective MYC G4 stabilizers have been recently reported by Felsenstein et al.170 (2016), who employed small molecule microarrays (SMM) to screen 20,000 drug-like compounds from ChemBridge and ChemDiv repositories. For selectivity purposes, the SMM screening technique was considered advantageous due to its fast throughput, and importantly, incorporation early in the design and discovery process of several unrelated oligonucleotide structures in addition to the targeted G4 Pu27-mer of MYC. This unbiased SMM screen complemented by a PCR stop assay yielded compound 1 (Figure 1.14) containing a novel G4-binding benzofuran scaffold, which inhibited MYC transcription via a G4-dependent mechanism of action. Direct SPR and reversible thermal melt binding assays demonstrated reversible association of compound 1 with the MYC G4 ( KD = 4.5 μM), with no quantifiable or weaker binding to other G4s formed in the promoters of KRAS, Myb, VEGF, Bcl2, and RB1 oncogenes. Compound 1 was able to arrest cell cycle in G1 phase and was selectively cytotoxic to multiple myeloma G4-containing cell lines, including the L363 cell line, but it had no effects on the CA46 Burkitt’s lymphoma cell line, being resistant overall to G4-mediated MYC inhibition.170 Further evaluation of compound 1 selectivity was demonstrated through gene expression analysis and qPCR experiments using L363 cells, which revealed differential G4 MYC-driven profiles between compound 1 and quarfloxin on their effects on levels of Myc 46 downstream expression, with substantial reduction of MYC-regulated genes over RB1, VEGFA, KRAS, and HIF1α ones.170 In a follow-up study, Calabrese et al.171 (2018) aimed at understanding the molecular determinants of binding affinity and selectivity for G4 structures. As such, the authors synthesized a focused library of 25 analogs of compound 1, as reported earlier by the group170. The substitution of the methyl group on the aryl amine moiety of compound 1 to a para-trifluoromethyl yielded the best derivative, DC-34 (Figure 1.14), which showed enhanced affinity and ability to downregulate MYC transcription in multiple myeloma cells in a G4-dependent fashion without affecting the expression of relevant G4-driven oncogenes. DC-34 had a KD of 1.4 μM, as determined by SPR, but a KD of 9.4 μM in a fluorescent intensity assay (FIA). In FIA, DC-34 preferentially bound MYC G4 over other G4 oligos and it did not bind to dsDNA.  Importantly, Calabrese et al171 resolved the NMR structure of MYC G4/DC-34 bound complex (PDB ID: 5W77). They demonstrated that extensive and differing bonding interactions or conformational changes within the tail, loop and G-tetrad elements of the quadruplex govern the recognition and selectivity (correlating with biological activity) of DC-34 for MYC G4 relative to other G4s, such as those of KRAS and Bcl2 genes. DC-34 bound independently and distinctly to 5′ and 3′ ends to form a 2:1 ligand-G4 complex in a similar manner to the previously reported NMR structure of a quindoline derivative ligated to MYC G4 (PDB ID: 2L7V)172.  47  Figure 1.14 Chemical structures of selective Myc G4 stabilizers, Compound 1 and optimized derivative DC-34.170,171  In 2018, Hu et al173 reported the discovery of IZCZ-3 (Figure 1.15), a novel and potent “four-leaf clover-like” compound that specifically stabilized the G4 structure of the MYC promoter. Its discovery was based on the design, synthesis, and optimization of aryl-substituted imidazole/carbazole conjugates, capitalizing on two already available chemical scaffolds, triaryl-substituted imidazole developed by the same group that stabilized parallel G4s but also telomeric multimeric G4s174, and carbazole, derivatives, which have been shown to bind strongly to MYC G4175. As assessed by fluorescence spectroscopy, IZCZ-3 showed significant selectivity for Pu22, the parallel MYC G4, relative to other representative DNA structures, including antiparallel HRAS G4, hybrid telomeric htg22 G4, G-triplex, i-motif, and double- and single-stranded DNA.173 The specifics of IZCZ-3 interactions with MYC Pu22 were further investigated using fluorescence titration (KD of 0.1 μM), CD melting, and molecular modeling studies substantiating the stabilization of the MYC G4 over the parallel promoter G4s of c-Kit, Bcl2, and KRAS.  48  Figure 1.15 Chemical structure of IZCZ-3 G4 stabilizer.173  Molecular docking of the optimized IZCZ-3 electronic structure against three template NMR structures of MYC Pu22, c-Kit, and htg22 showed that IZCZ-3 selectivity for MYC Pu22 was due to the lowest binding energy contributed from optimal end-stacking π–π interactions and placement of the central and positively charged IZCZ-3 imidazole ring in the Pu22 cation channel.173 Evaluation of IZCZ-3 cellular behavior in a panel of assays, including reporter and exon-specific assays, MTT and real-time cellular activity assays, flow cytometry, RT-PCR, and Western blotting, showed that IZCZ-3 induced G0/G1 cell cycle arrest and apoptosis, inhibiting cell growth and MYC transcription due to the selective targeting of MYC G4. Moreover, in mouse xenograft models, IZCZ-3 effectively suppressed tumor growth of human cervical squamous tumors.173 Importantly, also in 2018, Stump et al176 resolved the X-ray structure of the major Pu22 MYC quadruplex at 2.35 Å resolution (PDB ID: 6AU4), providing new avenues for future rational design of specific small molecules targeting MYC G4. The X-ray structure was in good overall agreement with the previously reported NMR structure of Pu22 (PDB ID: 1XAV)76. Major differences were observed in the conformation of the 5′ and 3′ flanking nucleotides and loop regions, but not in the G4 core. In the 6AU4 X-ray structure, the T1-G2-A3 trinucleotide 49 changed its stacked conformation that was observed in NMR to an extended one, protruding away from the G-tetrad formed by guanines G4, G7, G12, and G17. The triplet T20-A21-A22 at the 3′-end also extended away from the G6, G10, G16, and G19 tetrad. Large deviations were observed in the position of loop nucleotides, in particular that of T7.176 Further CD spectroscopy, performed in similar conditions to those employed in the NMR study of the Pu22-quindoline complex (PBD ID: 2L7V) indicated that MYC Pu22 maintained the parallel topology in crystallographic conditions, unlike several cases where different G4 conformations have been observed between X-ray and NMR structures.176 A comparison with the crystal structures of c-Kit (PDB ID: 4WO2) and human telomeric (PDB ID: 4FXM) quadruplexes as well as other NMR structures74 emphasized that the conformations of the 5′ end and loop regions were the main discriminatory features to capitalize on when seeking the selective targeting of particular G-quadruplex structures.176 Co-crystallization studies of MYC Pu22 with novel anthracenyl isoaxole amides (AIMs) G4 ligands (Figure 1.16) developed by the same group are currently ongoing.177,178  Figure 1.16 Chemical structure of anthracenyl isoaxole amides (AIMs) G4 ligands.177,178  On the computational front, targeting MYC G4 with small molecules is currently in its early stages, although advances have been made in the recent years.179 Through virtual screening and NMR spectroscopy, Ma et al.180 (2012) identified a modular natural compound carbamide 1 (out 50 of five analogues) as a G4 stabilizer that differed from previously described G4 binders, both in terms of chemical structure and binding mode, as it demonstrated G4 “groove-binding” specificity (Figure 1.17).  Figure 1.17 Chemical structure of natural compound carbamide 1 identified as G4 stabilizer through computational approaches.180  Using a platform that combined DOCK181 with three-dimensional (3D) pharmacophore screening of 560,000 publically available compounds from ChemDiv and Specs libraries against an NMR derived structure of MYC G4 (PDB ID: 2A5R)182, Kang and Park183 (2015) identified three potential MYC G4 stabilizers. While the compounds were active in Burkitt’s lymphoma Ramos as well as HeLa cell lines, they showed minimal thermal G4 stabilization in FRET melting screening assay. Following an incremental fragment-based docking approach that was implemented in the Surflex-DOCK platform184, Hou et al.185 (2015) identified a novel MYC G4 stabilizer from a screen of 28,530 compounds from the Chembridge repository against the NMR 1XAV Pu22 structure, rescored to eliminate intercalators and groove binders to dsDNA by docking against 1Z3F and 1K2S structures. The pyrollopyrazine-derived top hit, VS10 (Figure 1.18), bearing a novel scaffold, while less effective that the control quindoline SYUIO-05 in luciferase-based assays using Raji and CA46 cells, it showed selectivity for MYC G4 over dsDNA, as determined 51 by SPR and FRET-based competition assays. MD simulations combined with NMR showed that VS10 stabilized MYC G4 by stacking over the 5′ terminal G-quartet.185  Figure 1.18 Chemical structure of VS10 G4 stabilizer identified through computational approaches.185   Rocca et al.186 (2016) used a combination of structure-based pharmacophore screening and docking refinement to identify a novel “dual-specificity” G4 stabilizer binding to both h-Telo telomeric and MYC promoter G4. Six 3D pharmacophore models were constructed using LigandScout software187, with further cheminformatics enrichment and validation, from the 3D coordinates of four human telomeric G4 structures differing in their parallel-folding topology (PDB IDs: 3CE5, 2JWQ, 1NZM, and 3CDM) all complexed with known end stacker active ligands of four different chemical scaffolds. The six 3D pharmacophore models consisted of three or four chemical features, including aromatic ring, H-bonding acceptor/donor, hydrophobic, and positive ionizable. Following their established protocol, ~3 million compounds were screened against the obtained pharmacophore models. Pharmacophore-matching compounds were merged and filtered on the basis of best fitness and satisfaction of Lipinski’s rule of five to yield 1909 compounds subjected to “ensemble docking” simulations against additional G4 telomeric folding topologies: parallel (PDB ID: 1KF1), antiparallel (PDB ID: 143D), and two hybrid types (PDB IDs: 2HY9, and 2JPZ). Further filtering by ADME properties 52 and similarity clustering resulted in a set of 48 purchasable compounds assessed by CD, FRET-melting, and fluorescence intercalator displacement assays for their ability to interact with and stabilize telomeric G4 as well as parallel G4 of MYC (PDB IDs: 2A5P and 2A5R). Compound 56 (Figure 1.19), a fused heteroaromatic naphthyridin-containing derivative had a significant stabilizing effect, not only on parallel telomeric G4 but it was also the best binder to the parallel MYC G4.186 Docking refinements showed that, in the optimized 56-MYC G4 conformation, compound 56 stacked at the 5′ end and associated with MYC G4 (2A5R) via π-π stacking interactions with G4, G8, and A12 nucleobases, cation–π interactions with G4 and G17 bases and H-bonding with G4 nucleotide.186  Figure 1.19 Chemical structure of compound 56 G4 stabilizers identified computationally.186  By employing a rational drug design platform combining in silico, biological and biophysical methods, Bhat et al.188 (2017) identified a novel carbamoylpiperidinium-containing compound, termed TPP (Figure 1.20), which stabilized MYC Pu27 at the known site at its 5′ end guanine stack.   Figure 1.20 Chemical structure of TPP Myc G4 stabilizer identified using a rational drug design platform.188  53 The in silico pipeline included virtual screening of 54,645 chemicals from the Maybridge database, followed by Glide docking in all the three modes (HTVS, SP, and XP) using the Pu27 model built from 2A5P NMR structure, at the 5′ end site, followed by all atom molecular dynamics (MD) simulations in explicit solvent. Analyses of docking binding mode and MD trajectories showed TPP-induced stabilization of the 5′ end capping bases G5, A6, A15 via CH–π interactions, further strengthened by H-bonding interactions with 5′ end G4, G11, and A15, and polar interactions with A6, G11, and A15. Subsequent biophysical analysis using CD, isothermal titration calorimetry, and NMR indicated strong, energetically favorable binding of TPP to the parallel Pu27, also providing atomistic details that are consistent with MD results. Biological assays, including MTT, flow cytometry, RT-PCR, and luciferase-based reporter assays showed that TPP selectively induced apoptosis in T47D breast cancer cells with Myc overexpression, with no effect on normal kidney epithelial NKE cells by a mechanism involving the downregulation of Myc expression (at a 35 μM concentration) by arresting the Pu27 G4, thus interfering with MYC transcription.188 The biology and development of G4 stabilizers as therapeutics or probes has been more in-depth covered by some recent and comprehensive publications.153,154,179 The direct targeting of MYC transcription at the NHEIII1 cis-element described above is gaining tremendous speed as a viable therapeutic strategy to tackle Myc expression in cancer.  1.3.2 Direct Inhibitors of MYC-MAX Function Direct disruption of protein-protein and/or protein-DNA interactions with potent and selective small molecules is the apex of cancer therapy that to date is still elusive. As it has already been outlined before, in the particular case of Myc, this is due to its inherent disorder in 54 the dimerization and DNA binding regions, lack of any obvious targetable pockets, and, importantly, limited evidence to substantiate key interacting residues for small molecule inhibitors even when crystallographic structures serve as guides for drug discovery. The chronological development of small molecule Myc-Max prototype inhibitors and corresponding binding sites follows, with particular emphasis on modern computational drug design methods.  1.3.2.1 Inhibitors of MYC-MAX Heterodimerization  Berg et al (2002)189 identified the first reported inhibitor of Myc-Max protein-protein interactions, IIA6B17, by screening a combinatorial library of approximately 7,000 peptidomimetic compounds that have been assayed via the fluorescence resonance energy transfer (FRET) technique. Out of four initially identified hits, compound IIA6B17 was the most effective in suppressing the growth of Myc-transformed chicken embryo fibroblasts (CEF). Unfortunately, IIA6B17 also inhibited the transformation induced by Jun, a related basic zipper (bZip) transcription factor, suggesting a lack of specificity. Furthermore, Berg et al189 did not indicate whether the compound induced apoptosis, indicative of Myc inhibition. These observations limited the prospective of IIA6B17 as a drug candidate. Further, Shi et al190 introduced structural modification to the members of the same peptidomimetic library and identified two optimized compounds, Mycmycin-1 and Mycmycin-2, which were shown to specifically inhibit Myc-induced oncogenic transformation having no undesired effect on Jun-driven or unrelated Src protein kinase-driven transformation. In 2003, Yin et al191 utilized the yeast two-hybrid system to screen 10,000 drug-like compounds from the Chembridge DIVERSet combinatorial library and identified seven low-molecular weight inhibitors that blocked the Myc-Max interaction at the HLHLZ interface and 55 demonstrated no significant cellular toxicity. Notably, three compounds – 10058-F4, 10074-G5 and 10074-A4 (Figure 1.21) – demonstrated complete specificity toward Myc-Max. In particular, these compounds specifically inhibited Myc transcriptional activity and decreased cell growth of Myc-transformed rat fibroblasts. While the effect of these chemicals on cell viability was similar between Myc-parental lines expressing endogenous levels of Myc (i.e. TGR-1 cells having intact Myc alleles, Myc+/+) and Myc-overexpressing cells, HO15.19 Myc-null cell line (i.e. rat fibroblasts with homozygous deletion of the endogenous Myc gene, Myc-/-) demonstrated a relative lack of response. However, the treatment with Myc-specific compounds led to G0/G1 cell cycle arrest followed by apoptosis.  Thus, compounds 10058-F4, 10074-A4, and 10074-G5 have served as Myc prototypical inhibitors for many subsequent years, and helped accelerating research on therapeutic targeting of Myc. Nonetheless, while effective in cell lines, the use of these chemicals in vivo has been limited by their low potency (e.g. IC50s of 41.1 μM and 22.5 μM for 10058-F4 and 10074-G5, respectively, on growth inhibition of HL-60 Myc-overexpressing leukemia cells).192 Importantly, these chemicals also lack selectivity and display poor pharmacokinetic behavior as they undergo rapid metabolism resulting in low tumoral concentrations insufficient to block Myc-Max interactions in vivo, thus restricting their clinical applicability.193,194 For example, 10058-F4 characterization in vivo in SCID mice bearing DU145 and PC3 human prostate cancer xenografts, demonstrated the lack of efficacy and poor pharmacokinetic behavior of the specific Myc inhibitor in these classical models of moderate to high metastatic potential prostate cancer. Upon single intravenous dose treatment of mice with 20 or 30 mg/kg of 10058-F4, peak tumor concentrations of 10058-F4 were at least 10-fold lower than peak plasma concentrations, eight metabolites were identified in plasma, liver and kidney, and no significant inhibition of tumor 56 growth has been observed.193 Similarly, 20 mg/kg of 10074-G5 administered as single intravenous dose to mice bearing Daudi Burkitt’s lymphoma xenografts showed that the compound reached a peak plasma concentration of 58 μM with a plasma half-life of ~37 min.194 As in the case of 10058-F4, peak tumor concentrations were at least 10-fold lower than peak plasma values and as many as 19 inactive metabolites were observed with a number of these appearing to be glucuronide derivatives of hydroxylated or nitro-reduced 10074-G5, products of primary and secondary metabolic biotransformation in the liver.194,195 Therefore, while well tolerated, the in vivo efficacy of 10058-F4 and 10074-G5 against the tested tumors was hampered by rapid metabolism and insufficient tumoral concentrations to block Myc-Max dimerization in vivo.   Figure 1.21 Chemical structures of 10058-F4, 10074-A4, and 10074-G5 inhibitors of Myc-Max dimerization identified through high-throughput screening of a finite combinatorial library.191  In a subsequent work, Wang et al (2007)196 attempted to develop more potent and selective analogs of 10058-F4 by employing computer-assisted chemical similarity searches among > 500,000 drug-like molecules from the Chembridge database as well as by direct chemical synthesis. The approach taken was to identify similar compounds that maintained > 85% substructure similarity but bore modifications in either the six-member or the five-member 57 rhodanine ring of 10058-F4. 63 compounds were prioritized for IC50 determination in an MTT cell proliferation assay using HL-60 Myc-overexpressing cell line for comparison with that of 10058-F4. Four out of 48 identified six-member-substituted analogs had IC50s comparable to that of 10058-F4, ranging from 23 to 51 μM and four out of the 15 five-member ring analogs had improved potency (IC50 range 4.6-18 μM). These analogs were also able to disrupt Myc-Max heterodimerization in HL-60 cells as well as DNA recruitment as measured by co-immunoprecipitation (co-IP) and electrophoretic mobility shift (EMSA) assays. 17 more analogs were generated by combining the best 5- and 6-member substitutions, but they did not result in further improvements. This inconsistent behavior possibly reflected multiple binding modes of singly or dual substituted analogs to different conformations of the intrinsically disordered Myc monomer.196  Attempted optimizations of 10074-G5 in the form of JY-3-094 derivative (Figure 1.22) and its ester prodrug forms have also resulted in limited success. Thus, structure-activity relationship (SAR) studies of 10074-G5 (the NMR model of its Myc binding site, discussed further below, was used as a guide) led to the generation of an analog, JY-3-094, showing stronger ability to disrupt the association between recombinant Myc and Max proteins192,197, but it did not solve the issue of poor cell penetration. JY-3-094 retained the electron-rich nitro group and the furazan ring of 10074-G5 which were found critical for Myc inhibition197, and included a para-carboxylic acid as a replacement for the ortho-phenyl ring of the ortho-biphenyl of 10074-G5, modification which led to enhanced activity relative to 10074-G5. In EMSA, JY-3-094 disrupted Myc-Max dimerization with an IC50 of 33 μM which is significantly lower than IC50 =146 μM established for 10074-G5. However, due to its ionizable carboxylic moiety, JY-3-094 failed in exhibiting any cytotoxicity to HL-60 and Daudi cells. This lack of activity was remedied by esterification of 58 the para-carboxylic acid of JY-3-094 that resulted in a panel of ester prodrugs that enhanced cellular uptake (IC50 in low micromolar range in both HL-60 and Daudi cells), but unfortunately it impaired the ability to disrupt Myc-Max association in vitro.192 Replacement of the carboxylic acid of JY-3-094 for a phenol ester resulted in compound SF-4-017 that showed comparable potency with the parent in vitro but showed inhibitory activity of Myc-Max dimerization in cells in co-IP, potentially being resistant to esterification.192 Also related to 10074-G5 is the small molecule 3jc48-3 (Figure 1.22), a congener that showed increased potency and stability in cell-based assays.198 3jc48-3 was ~5 times as potent (IC50 =34 μM) at inhibiting Myc-Max dimerization as the parent compound. 3jc48-3 exhibited an approximate two-fold selectivity for Myc-Max heterodimer over Max-Max homodimer, suggesting binding to Myc. 3jc48-3 inhibited the proliferation of Myc-overexpressing HL-60 and Daudi cells with single-digit micromolar IC50 values by causing growth arrest at the G0/G1 phase. Co-IP studies indicated that 3jc48-3 inhibits Myc-Max dimerization in cells, further substantiated by specific Myc-driven gene silencing. Finally, 3jc48-3's intracellular half-life was > 17 hrs.198 3jc48-3 is perhaps one of the most potent, cellular active and stable Myc inhibitors from the classical series developed by the group.192,195,197,198   Figure 1.22 Chemical structures of JY-3-094 and 3jc48-3, improved analogs of compound 10074-G5.192,197,198  59 The most recent efforts to increase the potency and selectivity of classical Myc inhibitors focused on medicinal chemistry and novel delivery technologies. Thus, Wanner et al (2015)199 generated intracellular self-assembly dimeric inhibitors that showed improved potency and activity in cancer cell lines overexpressing Myc (such as Daudi and Raji Burkitt's lymphoma human cell lines). The technique relies on reversible linkage of chemically modified 10058-F4 and 10074-G5 to produce larger molecules, well suited for targeting the Myc-Max dimer surface and capitalize on the drug-like properties of the small molecule components. Moreover, Soodgupta et al (2015)200 reported nanoparticle targeted-delivery of 10058-F4 as an Sn2 lipase-labile pro-drug in the form of MI1-PD, which showed in vivo efficacy as it extended survival in a mouse model of metastatic multiple myeloma.  Further screens have identified small molecule inhibitors of protein-protein interactions of Myc-Max such as Mycro3 and KJ-Pyr-9, that showed improved pharmacokinetics, bioavailability and overall in vivo activity and demonstrated efficacy in mouse models of pancreatic and breast cancers.  Mycro3 (Figure 1.23) was identified by Kiessling et al (2007)201 via high-throughput screening of a library of 1438 pyrazolo [1,5-α] pyrimidines built upon the two predecessor compounds, Mycro1 and Mycro2.202 Mycro3 was the first inhibitor from the constructed library that interfered with the formation of the Myc-Max-DNA complex by inhibiting protein-protein interactions between Myc and Max. Mycro3 potently and preferentially interacted with Myc-Max over the Max-Max dimer or the related bZip AP-1 transcription factors (Jun/Fos proteins), and without affecting the DNA binding of structurally unrelated STAT3 transcription factor to its binding site. Importantly, Mycro3 strongly inhibited Myc-dependent proliferation of U-20S osteosarcoma cells (70% reduction) while not inhibiting the Myc-Max-independent PC-12 cell 60 line which lacks Max. A subsequent study203 demonstrated that the strength of Mycro3 activity was cell-line dependent with TGR-1 cells having intact Myc alleles (Myc+/+) showing higher sensitivity to Mycro3 (IC50 of 0.25 μM) in comparison with other cell lines, such as HO15.19 Myc-null cells (Myc-/-) of common origin to TGR-1 (IC50 of 9 μM) or U-20S cells (IC50 of 10 μM). Moreover, in contrast to 10058-F4 which lacks antitumor activity in vivo, the pharmacokinetic profile of Mycro3 was quite good and characterized by sustained presence in circulation in mice at concentrations adequate for efficacy studies (0.5 μM). Although water-insoluble, in mice Mycro3 was amenable to daily administration by oral gavage as an emulsion for treatment of oncogenic KRAS (KRAS*)-induced pancreatic ductal adenocarcinoma (PDA) which is dependent on Myc activity. Stellas et al (2014)203 showed that genetic ablation of Myc effectively prevented the development of KRAS*-induced pancreatic, mammary and prostatic adenocarcinoma and penile squamous cell carcinoma in mice (consistent with Omomyc data). Treatment with Mycro3 Myc-inhibitor resulted in marked shrinkage of PDA, increased survival in a PDA mouse model, increased apoptosis, and reduced cell proliferation. Mycro3-treated mice carrying orthotopic or heterotopic xenografts of human pancreatic cells also showed significantly attenuated tumor growth.203  61  Figure 1.23 Chemical structures of Mycro3, KJ-Pyr-9, sAJM589, and MYCMI-6 inhibitors of Myc-Max dimerization identified through high-throughput screening of diverse small-sized compound libraries.201,203-206  KJ-Pyr-9, a more recent inhibitor was identified by Hart et al (2014)204 among a 220-membered Kröhnke pyridine combinatorial library207 screened by a fluorescence polarization assay for inhibition of Myc-Max dimerization (Figure 1.23). By employing backscattering interferometry given the low aqueous solubility of the compounds, Hart et al204 determined that KJ-Pyr-9 binds directly to disordered Myc with nanomolar affinity (KD of 6.5 nM) as well as to the Myc–Max heterodimer dissociating it (13.4 nM), but only weakly to the Max homodimer (>1 μM). Despite its low solubility, KJ-Pyr-9 was cell-permeable and interfered with Myc-Max complex formation in cells as determined by a Renilla luciferase-based protein fragment complementation assay (PCA) (with Myc332–439–Max biosensor) supporting the direct intracellular binding of the compound to Myc.  62 It has been further demonstrated that KJ-Pyr-9 preferentially blocks the proliferation of Myc-overexpressing cells (i.e. P493-6 engineered human B-cell line that overexpresses Myc in absence of tetracycline (Tet-off), leading to robust cell proliferation) with no or weak effects on the oncogenic activity of unrelated Src and Jun proteins. The compound also inhibited the proliferation of other cell lines dependent on increased Myc activity: NCI-H460 (large-cell lung cancer), MDA-MB-231 (breast adenocarcinoma), and SUM-159PT (estrogen-independent breast cancer) with IC50 values in the 5 to 10 μM range. The proliferation of Burkitt’s lymphoma cell lines with constitutively high expression of Myc was more sensitive to KJ-Pyr-9 (IC50 between 1 and 2.5 μM). In addition, proliferation of leukemia cell lines K-562, MOLT-4, and HL-60 overexpressing Myc was strongly inhibited, while the colon carcinoma cell line SW-480 was not affected. Importantly, KJ-Pyr-9 was also effective in blocking N-Myc-dependent proliferation. Moreover, KJ-Pyr-9 induced apoptosis by cleavage of caspase 3, and specifically reduced well-established Myc-driven transcriptional signature in the P493-6 Myc-off cell line.208-210  The compound also demonstrated promising pharmacokinetic properties in mouse (3 mice injected at 10 mg/kg intraperitoneal) and in rat (1 mg/kg intravenous) models achieving blood concentrations sufficient to prompt its further investigation in vivo (mice: 3.5 μM in plasma, and 12.4 μM in brain, rat: elimination half-life in plasma ~1.84 h). It was not acutely toxic at doses as high as 10 mg/kg in mice; it crossed the blood-barrier and was present at higher concentration in brain tissue than in the blood after 4 hours. In vivo, KJ-Pyr-9 effectively blocked the growth of xenografts of MDA-MB-231 breast cancer cells in mice treated daily by intraperitoneal injection with 10 mg/kg of the compound. KJ-Pyr-9 represents one promising chemical probe for investigating the modulation of Myc-Max protein-protein interactions as an anticancer strategy. 63 Combinations of KJ-Pyr-9 with other targeted drugs have been suggested for potentially enhanced effect on cancer cells.204 Despite these positive results reported by Hart et al204, the binding mode of KJ-Pyr-9 to Myc remains to be determined to help explain the selectivity of the compound for Myc. Moreover, additional studies in multiple animal models are required to address some limitations of their in vivo observations, including examination of normal rapidly proliferating tissues in a manner similar to Omomyc treatment in healthy tissues.204 sAJM589 is another small molecule Myc inhibitor identified by Choi et al. (2017)205 from a Gaussia-based PCA-based high-throughput screen of ~400,000 drug-like molecules (Figure 1.23). sAJM589 is a potent disruptor of Myc-Max dimerization, a protein–protein interactions (PPI) inhibitor, with an IC50 of 1.8 μM, ~25-fold more potent than 10058-F4 in the PCA biochemical assay. Similar to KJ-Pyr-9, sAJM589 preferentially inhibited cellular proliferation of P493-6 (Tet-Off) cell line in a dose-dependent manner with IC50 = 1.9 μM, while in the presence of tetracycline, its IC50 was >20 μM. sAJM589 also showed a dose-dependent inhibition of various Myc-dependent cancer cell lines: Ramos (Burkitt’s lymphoma), HL-60 and KG1a (acute myeloid leukemia) with IC50s of 0.9, 1.2, and 0.8 μM, respectively, but had no effect on resting macrophages whose proliferation is independent of Myc activity. Co-IP confirmed the disruption of Myc-Max PPI and no effect on the Jun/Fos dimerization. In EMSA, over 2 μM of sAJM589 completely disrupted Myc-Max binding to E-box DNA. Moreover, biolayer interferometry (BLI) showed direct binding of sAJM589 to the LZ region (amino acids 403-437) of a biotin-tagged Myc protein in a dose dependent manner (6.25 to 100 μM sAJM589 concentration range). The authors reported a reduction of Myc protein levels but no reduction of MYC mRNA levels in P493-6 cells upon disruption of Myc-Max PPI with compound. This reduction was attributed for the first time to a novel mechanism that 64 involves destabilization, ubiquitination, and degradation of Myc. Elegantly, Choi et al.205 showed that sAJM589 facilitated Myc turnover reducing Myc protein half-life by ~2-fold, after de novo protein synthesis was blocked by cycloheximide.  Furthermore, when additional treatment with the proteasome inhibitor MG-132, which prevented the proteasome-mediated protein degradation, the increased turnover of Myc protein upon treatment with sAJM589 promoted the accumulation of ubiquitin-conjugated Myc in P493-6 cells. Thus, sAJM589 facilitated Myc ubiquitination by inhibiting Myc-Max binding and freeing the Myc C-terminal region for E3 ubiquitin ligase binding, resulting in its degradation by the proteasome. While Choi et al205 suggested that sAJM589 may provide a basis for the development of potential inhibitors of Myc-dependent cell growth, initial chemical optimization revealed a narrow structure-activity relationship (SAR), with sAJM589 showing superior efficacy in both PCA and cell proliferation assays. In one of the most recent reports, Castell et al. (2018)206 aimed at targeting the Myc-Max PPI and identified a novel small molecule inhibitor, MYCMI-6 (NSC354961), from a library of 1990 compounds from NCI/DTP Open Chemical Repository using a cell-based Bimolecular Fluorescence Complementation (BiFC) assay (Figure 1.23). MYCMI-6 exhibited strong selective inhibition of both c-Myc-Max and N-Myc-Max PPI in vitro at single-digit micromolar concentrations. Thoroughly, the authors demonstrated that MYCMI-6 had no effect on Mxd1 (Mad1)-Max dimerization and on Jun/Fos. Thus, among all Myc inhibitors, only MYCMI-6 and earlier 10058-F5 and 10074-G5 have been shown to be selective for Myc-Max when compared to Mad-Max. Moreover, MYCMI-6 blocked Myc-driven transcription and bound selectively to the Myc bHLHLZ domain with a KD of 1.6 μM, as measured by the surface plasmon resonance 65 (SPR) assay. Furthermore, MYCMI-6 inhibited tumor cell growth in a Myc-dependent manner at an IC50 concentration as low as 0.5 μM, but it was not cytotoxic to normal human cells. Perhaps, the greatest merit of the Castell et al206 study is the demonstration of the effects of MYCMI-6 administration by daily intraperitoneal injection at a dose of 20 mg/kg in vivo. MYCMI-6 induced massive apoptosis and reduced tumor cell proliferation, tumor microvasculature density, and, notably, N-Myc-Max interaction in an N-Myc-dependent mouse xenograft model based on human N-Myc amplified SK-N-DZ neuroblastoma cells, without causing severe side effects. MYCMI-6 treatment was well tolerated in mice with only slight and temporal effects on body weight.  1.3.2.2 Inhibitors of MYC-MAX Binding to DNA  A number of studies aimed to target specifically the binding of Myc-Max transcriptional complex to DNA. The developed Myc-Max/DNA disruptors (Figure 1.24) include:  1. natural-occurring compounds, such as celastrol and celastrol-inspired triterpenoids211; 2. synthetic α-mimetics, such as JKY-2-169 intentionally engineered to recognize the structurally ordered Myc and hence to disrupt Myc-Max/DNA binding212;  3. small molecule inhibitors, MYRA-A and NSC308848 that selectively target the DNA-binding domain (DBD) of Myc-Max213,214 and KSI-3716 that also blocks Myc-Max binding to DNA.215,216  Aiming at identifying a Myc-targeting therapy that could specifically affect cells with deregulated Myc in vivo, Mo et al (2006)213 performed a cellular screen of 1,990 NCI compounds to identify those, which preferentially inhibit proliferation of tumor cells with high 66 Myc expression but do not affect systems with low Myc levels. High Myc expression was obtained by using cells with inducible Myc expression (i.e. mouse fibroblasts with a tetracycline-inducible Myc transgene (Tet-Myc) treated with doxycycline in a dose- and time-dependent manner). This screen identified two compounds known as Myc-pathway response agents, MYRA-A (NSC339585) and MYRA-B (NSC45641), which promote apoptosis in a Myc-dependent manner and inhibit Myc dependent transformation. The Myc-dependent effects of these compounds were further verified in rat fibroblasts (Rat-1 cells). MYRAs effect on cell viability was more significant in rat cells with Myc overexpression (HOmyc3) compared with cells with wild-type Myc (TGR-1), whereas Myc-null cells (HO15.19) were highly resistant to the compounds. Importantly, MYRAs did not disrupt the dimerization of Myc-Max; instead, MYRA-A directly interfered with binding of Myc-Max to DNA. Unfortunately MYRA-A also interfered with the DNA binding of Mnt-Max and Max-Max complexes perhaps unsurprisingly since Mnt and Myc recognize same DNA-binding sites and regulate an overlapping set of target genes in vivo.213 It has been demonstrated, however, that MYRA-A could discriminate Myc network members from other E-box-binding proteins as it did not affect the DNA-binding activity of the E-box binding upstream stimulatory factor, USF. MYRA-A is structurally similar to anthracyclines but, unlike them, does not function as a DNA intercalator (i.e. as a topoisomerase II poison) since it had no effect on USF DNA-binding. Overall, the Mo et al213 report was the first to demonstrate, through MYRA-A identification, the preferential inhibition of Myc overexpression combined with selective effects by induced apoptosis in a Myc-dependent manner rather than proliferative arrest. Moreover, mechanistically, MYRA-A blocked Myc-Max/DNA interactions without interfering with protein-protein interactions perhaps by sufficiently distorting the Myc-Max heterodimer rigid 67 structure to cause loss of DNA binding without altering the Myc-Max interactions enough to promote the dimer dissociation.213   Figure 1.24 Chemical structures of MYRA-A, NSC308848, JKY-2-169, and KSI-3716 inhibitors of Myc-Max binding to DNA identified through high-throughput screening or specifically engineered to disrupt protein-DNA interactions.212-216  In a parallel study, Mo et al (2006)214 showed that MYRA-A also exhibits prominent effects on N-Myc overexpressing neuroblastoma cells. Moreover, they have identified a third compound, NSC308848 that also induced apoptosis in Myc-overexpressing lines (Figure 1.24). In contrast to the MYRAs action, treatment with NSC308848 resulted in decreased Myc protein levels and gave rise to inhibitory effects not only on Myc but also on other transcription factors, including p53. NSC308848 treatment decreased the levels of Myc in p493 B cells, and in HL-60 cells and resulted in a dose-dependent reduction of N-Myc in Tet21N cells while neither MYRA-68 A nor MYRA-B had any significant effects on the protein levels of N-Myc or c-Myc. Thus, the mechanism of action of NSC308848 differs from that of MYRA-A and MYRA-B. Whether NSC308848 affects Myc protein levels by interfering with transcription or by enhancing protein degradation was not determined. Mo et al213,214 findings suggested that the three small molecules can elicit a similar biological response by interfering with the Myc pathway at different levels: all three compounds induced apoptosis in Myc overexpressing cells albeit with different mechanisms of action.213,214  In an attempt to overcome the lack of selectivity of the earlier identified Myc inhibitors, Jung et al (2015)212 further designed synthetic α-mimetics specifically engineered to recognize ordered Myc in its helical form. They synthesized biphenyl-based compounds that preserved the hydrophobic core and electron-rich peripheries of earlier compounds (i.e. 10074-G5 and derivatives) yet further intended to recognize a hydrophobic domain of helical Myc flanked by arginine and other polar residues responsible for the formation of a rigid tertiary structure upon dimerization with Max. The initial in vitro screen consisted of EMSA followed by NMR spectroscopy and SPR biophysical assay to generate further evidence of direct binding and impairment of protein-DNA interactions. The best synthetic compound, JKY-2-169 (Figure 1.24) perturbed Myc-Max binding to the canonical E-box DNA sequence without causing protein-protein dissociation in co-IP. In addition, JKY-2-169 inhibited cell proliferation of Myc overexpressing HL-60 and Daudi cells with IC50s of 20 and 9.5 μM, respectively, promoted G0/G1 cell cycle arrest and accumulation of neutral lipids. Nonetheless, compared to earlier derivatives, its KD value was not significantly improved (~13 μM obtained by NMR, 10 μM in SPR) and the specificity was not enhanced either (similar activity was observed in non-cancer cells lacking Myc overexpression). 69 It has also been demonstrated that JKY-2-169 had additional off-target effects and/or nonspecific toxicities given that the sensitivity of HO15.19 Myc-null cells to JKY-2-169 growth inhibition was comparable to that of TGR1 cells, with IC50s of 20 and 14 μM. Moreover, U266 multiple myeloma (MM) cell line (IC50 of 46 μM), which expresses L-Myc instead of c-Myc and which is the least susceptible among MM cell lines to 10058-F4 growth inhibition (IC50 ~100 μM), was sensitive to JKY-2-169 inhibition. At the time of their reporting, Jung et al212 planned to continue the work with a focus on the determination of the metabolic stability of JKY-2-169 which retains the nitro moiety of 10074-G5 responsible for the compound’s short half-life due to the rapid metabolism to toxic hydroxylamino and other inactive amino derivatives.  In other study, Jeong et al (2010)215 screened with EMSA a library of 6,480 small molecules from Korea Chemical Bank and identified five compounds that blocked Myc-Max binding to DNA with low IC50s of 0.58 μM for KSI-2826, 0.50 μM for FBN-1503, 2.0 μM for KSI-1449, 2.6 μM for KSI-2303 and 0.86 μM for KSI-3716. These chemicals potently suppressed Myc-dependent proliferation and induced apoptosis of HL-60 leukemia cells via G0/G1 cell cycle arrest without altering the expression level of Myc in differentiated HL-60 cells. As with the case of KJ-Pyr-9, this study did not show whether KSI-3716 inhibited Myc-Max interaction in vivo. In a follow-up study, Jeong et al (2014)216 showed that KSI-3716 (Figure 1.24) blocked Myc-Max binding to target gene promoters and decreased Myc-mediated transcriptional activity at concentrations as low as 1 μM as well as the expression of target genes including cyclin D2, CDK4 and hTERT. KSI-3716 exerted cytotoxic effects on bladder cancer cells by inducing cell cycle arrest and apoptosis. Notably, KSI-3716 demonstrated significant growth inhibition of tumors intravesically instilled into the bladder in murine orthotopic xenograft models. With the 70 dose of 5 mg/kg administered twice weekly for 3 weeks the compound demonstrated minimal systemic organ toxicity.216 Thus, Jeong et al216 suggested further development of KSI-3716 as an intravesical chemotherapy agent for bladder cancer given that KSI-3716 showed several physicochemical characteristics suitable for such therapy. These included a drug-like molecular weight and octanol-water partition coefficient (logP), small polar surface area (PSA), a lack of enzymatically cleavable chemical bonds (i.e. amide or ester bonds) and low solubility according to its logS. The above listed properties may confer KSI-3716 a high cell permeability, a relatively low diffusion rate in blood and other tissues, as well as an adequate intracellular concentration once absorbed by bladder and before excretion, and unlikely systemic toxicity due to negligible diffusion in the whole blood. A subsequent report by Seo et al (2014)217 outlined that KSI-3716 had significant activity even against gemcitabine-resistant tumors. Gemcitabine, a synthetic pyrimidine nucleoside analog, is currently one of the most effective first- and second-line chemotherapies as a single agent or combination against metastatic bladder cancer, however multidrug resistance eventually emerges. Notably, KSI-3716 suppressed gemcitabine-resistant xenograft tumors (KU19-19/GEM) formed in presence of as little as 2 mg/kg gemcitabine. Furthermore, sequential addition of gemcitabine and KSI-3716 augmented the therapeutic potency, suggesting that the sequential combination treatment may benefit bladder cancer patients.217  1.3.2.3 Computational Approaches toward MYC-MAX Inhibition  Novel small molecule inhibitors have been recently generated for a broad diversity of biological targets, as significant advances have been made in structure-based computational drug 71 design techniques such as virtual screening (VS) and molecular dynamics (MD) simulations, as well as increased access to large chemical repositories, and availability of crystallographic data. While this is generally true for well-defined, ligated targets such as membrane receptors and a large number of enzymes, accounting for more than 70% of the current drug molecule targets218, targeting intrinsically disordered proteins by computational means represents the high-risk yet high-reward opportunity so long awaited to advance cancer treatment. To maximize the success of structure-based drug discovery (SBDD), cutting-edge MD simulations have recently been employed, for instance, as promising computational techniques to generate heterogeneous structural ensembles of multiple intrinsically disordered protein conformations that could be simultaneously used in VS, more specifically in what is known as “ensemble or multi-conformational docking”.219 In 2016, Yu et al135 provided a first successful example of a general computational approach involving comprehensive MD conformational sampling of an intrinsically disordered region of Myc, which produced an ensemble of representative conformations used for in silico binding site identification and “multi-conformational” molecular docking. Novel compound identification resulted from docking score analysis of the VS ensemble and “multi-conformational-affinity” compound selection. Thus, 201,939 compounds from the SPECS and DCSD libraries as well as a number of selected analogs of 10074-A4 were used for VS using the Glide docking program in standard precision (SP) mode.220,221  Top 5% of docked compounds for different cavities identified on distinct Myc conformations were chosen for manual inspection, and 250 compounds from vendor libraries and 23 analogs of 10074-A4 were selected for experimental testing. Two classes of compounds were selected: 1) “high-conformational-specificity” compounds where the best docking score among 72 the 3 identified cavities was less than -6 and the other two were greater than -4, and 2) “multi-conformational-affinity” compounds where the differences of the 3 scores were less than 2 and at least one of the 3 docking scores was less than -5.   Figure 1.25 Chemical structures of PKUMDL-YC-1201 to -1205 compounds that disrupt Myc-Max dimerization identified through virtual screening (VS) multi-conformational docking against a reference ensemble of Myc disordered conformations generated through molecular dynamics (MD) simulations.135   Out of 273 compounds, 7 actives emerged PKUMDL-YC-1101, -1201-1205 and -1301, all binding the disordered Myc with different affinities as determined by SPR with C-terminal bHLHLZ biotinylated Myc immobilized on the chip. The corresponding KD values of 0.28, 17.2, 32, 0.55, and 18 μM for 1101, 1201, 1203, 1204, and 1205, respectively, were all superior to 36.3 μM KD for the 10074-A4 control. Among these, four compounds blocked Myc function in 73 cell as they inhibited the growth of the Myc-overexpressing HL-60 cells in an MTT assay (EC50s of 6.9, 8.8, and 40 μM for 1203, 1204, and 1205, respectively, relative to 15.1 μM for 10074-A4 control). Moreover, the four compounds affected cell cycle progression in a dose-dependent manner increasing the S-phase while decreasing the G2/M phase, suggesting that the observed effect on viability was not due to cytotoxicity. In addition, the four compounds blocked Myc-mediated transcriptional activity in HL-60 cells as demonstrated in qRT-PCR showing reduction in cyclin D2 and CDK4 downstream Myc regulated genes, confirming that Myc inhibition caused the cell cycle progression arrest. Apart from PKUMDL-YC-1101, all 6 actives from docking as well as 10074-A4 were “multi-conformational-affinity” compounds. Compounds PKUMDL-YC-1201, PKUMDL-YC-1202, PKUMDL-YC-1203, and PKUMDL-YC-1204 shared a common thiourea structure (Figure 1.25). Even though PKUMDL-YC-1203 and PKUMDL-YC-1204 had better experimental profiles they were less soluble in water prompting the authors to further investigate PKUMDL-YC-1205, an analog of 10074-A4, instead.  In an SPR competitive assay with GST-tagged Max protein immobilized on a chip, PKUMDL-YC-1205 abolished Myc binding to Max in the dissociation curves at concentrations in the 100-800 μM range. Chemical cross-linking and anti-Max Western blotting experiments showed disruption of Max-Max/Myc-Max dimerization equilibrium. Treatment with PKUMDL-YC-1205 decreased the Myc-Max heterodimer ratio, consequently leading to an increased Max homodimer ratio. As the parental compound, PKUMDL-YC-1205 is a disruptor of Myc-Max protein-protein interactions. Five independent 100-nanosecond MD simulations in explicit solvent were then performed, using the Amber molecular dynamics package222, for each of the following: Myc/PKUMDL-YC-1205, Myc/10074-A4, and Myc/AJ-292/41944612 (an inactive compound) docked complexes as initial structures. Analysis of the trajectories showed that 74 PKUMDL-YC-1205 and 10074-A4, which were “multi-conformational-affinity” compounds, had longer binding times than the non-active and “high-conformational-specificity“ compound during the simulation course. Yu et al135 provided a useful strategy for SBDD targeting IDPs while also suggesting a tendency for IDPs to bind to “multi-conformational-affinity” compounds – compounds that bind to various groups of conformations with similar affinity, instead of “high-conformational-specificity“ ones – compounds with high affinity to one class of conformation but very low affinity to others.  In 2018, Yao et al223 reported another novel inhibitor 7594-0035 for potential treatment of relapsed/refractory multiple myeloma (MM). In MM, frontline therapy-resistance is strongly associated with Myc.224,225 Compound 7594-0035 has been identified from the ChemDiv database, a commercially available repository containing more than 1 million small molecules, through VS and utilizing the published crystal structure of Myc-Max complexed with DNA123 (PDB ID: 1NKP) (Figure 1.26). Interestingly, although the group considered the ordered dimer structure for virtual screening, they did not attempt to identify pockets on the dimer surface through computational means. Instead, the investigators considered as the docking site the previously described disordered region bound by the 10074-G5 inhibitor (see subsection 1.3.2.4) but in ordered form, due to the fact that the 1NKP X-ray structure of the Myc-Max complex lacks a bound small molecule ligand. For that matter and importantly, no ligated X-ray structure of Myc-Max exists. The Surflex molecular docking226 module of Sybyl-X2.1 suite227 was used for virtual screening. 200 ChemDiv compounds were purchased after two rounds of screening, one intended to accelerate the docking process by reducing the number of conformers and rotatable bonds, and a second one to re-screen the top 1% compounds using default parameters at the selected site. 75 Compound 7594-0035 showed a high docking score (< -6) and following experimental validation showed potent Myc inhibitory activity. This compound inhibited the proliferation of RPMI-8226 and U266 multiple myeloma cells in vitro at concentrations in the 20-40 μM range, induced cell cycle G2 phase arrest and triggered their apoptosis by disturbing the stability of Myc protein while Myc mRNA levels were unaffected.  Treatment with 7594-0035 resulted in rapid degradation of Myc protein after inhibition of protein synthesis with cycloheximide and the proteasome inhibition with MG132. Moreover, administration of 7594-0035 overcame drug-resistance to bortezomib also a proteasome inhibitor considered a breakthrough in the treatment of MM228, and increased the killing effect of MM cells in combination with bortezomib. 7594-0035 also decreased primary tumor growth in vivo (SCID mouse xenograft model subcutaneously injected with RPMI-8226 cells). As it showed significant anti-tumor effect on MM cells, this compound may be a promising therapeutic agent for MM.   Figure 1.26 Chemical structure of 7594-0035 Myc-Max inhibitor identified in silico utilizing the Myc-Max 1NKP X-ray structure, yet targeting a previously reported disordered binding region for 10074-G5 but in ordered form.223  An opposite strategy to reduce Myc-Max activity pursed by computational techniques is minimizing the availability of Max by stabilizing Max-Max repressive homodimers competing with Myc-Max at same E-boxes. It has been rationalized that given the rather unique packing 76 defect of the Max homodimer (described in subsection 1.2.6.4) that makes Max-Max less stable than Myc-Max or other heterodimers of the extended network, specific small molecule stabilizers of Max-Max interactions could potentiate the downregulation of the network.123,229  Thus, in 2009, Jiang et al229 isolated Max-Max stabilizers by applying “blind docking” – a virtual screening technique for sampling large regions of protein complexes, as well as three-dimensional clustering analysis to identify specific binding pockets for small molecule interactors, constituting the first computational study to screen a medium-sized database of compounds over entire structurally ordered dimers. The AutoDock 3.0.5 software230 was used to find the best fit chemicals out of 1668 selected from 140,000 NCI compounds for VS targeting both Myc-Max123 and Max-Max122 X-ray dimer structures (PDB IDs: 1NKP and 1AN2). Clustering of the docked poses consisted of iterative analysis of their predicted binding location, lowest binding scores, and inclusion in a 10 Å distance from center of mass. Three main clusters emerged that contained 85% of docked compounds including those with lowest predicted binding energies. The compounds in each cluster exhibited similar trends in their chemical properties. VS results for the Max homodimer and the Myc-Max heterodimer were similar for cluster 1, which contained 456 compounds total, including those with the lowest predicted binding energies from the entire NCI set, and generally displaying an abundance of negatively charged atoms or a high density of hydrogen bonding atoms. A representative compound for cluster 1 is NSC131615 (Figure 1.27). Cluster 2 contained only 90 compounds, and it was the cluster with the second lowest binding energies, after cluster 1. The top binding compounds from this cluster were similar in charge to those from cluster 1 but were generally more hydrophobic. The representative compound for cluster 2 was NSC292215 (Figure 1.27). The third cluster was the largest of the three, containing 863 compounds, yet compounds in 77 cluster 3 were the weakest predicted binders among these top three clusters. The chemical structures for cluster 3 compounds include one or more rigid, flat surfaces and a lack of a strong negative charge. 68 unique compounds (40 from the Max-Max VS, 40 from the Myc-Max VS, and 12 removed for being common to both sets) representing the top docking solutions from the 3 clusters with predicted binding energies from -11.3 to -8.4 kcal/mol were requested from NCI for further experimental screening in a primary FRET assay.   Figure 1.27 Chemical structures of representative Max-Max stabilizers that alter Myc-Max function by blocking binding to equivalent DNA binding sites.229  FRET analyses correlating with the molecular docking results, showed that 85% of the compounds of cluster 3 were specific for the Max homodimer, compared to 67% for cluster 2 and only 13% for cluster 1. Representative of cluster 3, compound NSC13728 is the most effective compound selected for further characterization in this study (Figure 1.27). In vitro binding assays showed that NSC13728 enhanced Max-Max homodimerization while also interfered with Myc-Max heterodimerization (co-IP, ELISA, SPR and ultracentrifugation analysis). EMSA analysis showed that NSC13728 which targets the intersection of the LZ and 78 HLH regions did not significantly affect DNA binding of the Myc-Max heterodimer or that of the Max homodimer in contrast to the compounds of cluster 1 and 2, predicted to interact with the basic DNA binding region of Myc or Max, which decreased DNA binding significantly. Nonetheless, NSC13728 interfered with Myc-mediated oncogenic transformation of CEF with an IC50 of 3 μM, but not with that induced by the Src oncoprotein. NSC13728 inhibited the growth of the breast cancer MCF7-35IM cell line that carries an inducible Myc transgene, and when used in combination with the estrogen antagonist ICI 182,780 that inhibits endogenous Myc expression resulted in an additive reduction. NSC13728 also attenuated Myc-dependent transcription at concentrations in the 2.5-10 μM range, because of the stabilization of the Max homodimer. Although limited in the chemical space employed for virtual screening, this study provided nonetheless proof-of-concept that ordered dimer structures are more promising in silico docking targets for small-molecule disruptors over the disordered monomeric forms.  1.3.2.4 Binding sites for MYC-MAX Small Molecule Inhibitors  Despite the growing number of novel Myc-Max inhibitors, there is a paucity of information with respect to their precise binding mode. In fact, binding sites have been determined solely for a handful of inhibitors.  Identification of early binding sites for compounds 10058-F4 and 10074-G5, Myc402-412 and Myc363-381 (Figure 1.28), respectively, resulted from point mutations and truncations of synthetic peptides of the Myc DNA-binding and dimerization domains studies, followed by circular dichroism (CD) and NMR spectroscopy.231 A fluorescent polarization assay, capitalizing on the intrinsic fluorescence exhibited by 10058-F4 and 10074-G5, monitored the direct binding to purified recombinant Myc bHLHLZ domain, Myc353-437. Binding of 10058-F4 was impaired by 79 mutations of residues at the interface between H2 and LZ regions: Leu404Pro, Gln407Lys, and Val406Ala-Glu409Val and by deletion of the LZ region Myc406-437 (Figure 1.28). By contrast, binding of 10074-G5 was enhanced by amino acid substitutions between the basic region and H1: Arg367Gly and Glu369Lys-Leu370Pro and eliminated by truncations Myc370-409 and Myc400-439. Other ~20 mutations did not have a substantial effect on binding affinity.   Figure 1.28 Binding sites for structurally diverse Myc-Max inhibitors mapped along Myc and Max protein sequences. Residue numbering above a protein sequence follows the convention for individual monomers, while that below follows the numbering in the Myc-Max X-ray structure. Circle-enclosed digits indicate binding sites. The coloring scheme suggests three major druggable sites on Myc in its disordered monomeric form.  Flexible molecular docking of 10058-F4 into Myc402-412 peptide segment showed that the inhibitor was located at the center of a C-shaped cavity that formed upon binding of compound to the peptide segment. 10058-F4 docked in an orientation that allowed for intramolecular hydrophobic interactions to take place between its aromatic ring and the ethyl tail and intermolecular hydrophobic interactions with the N-terminal hydrophobic aromatic and aliphatic side chains of the peptide: primarily, Tyr402, Leu404 and Val406 but also Ile403 and Ala408. Additionally, the carbonyl oxygen of 10058-F4 was within H-bonding distance with Ser405 and 80 Gln407 side chains. This model matched independent NOESY results that indicated the formation of a similar hydrophobic cluster.  The model of 10074-G5 binding to Myc363-381 revealed that the inhibitor docked to a pocket that was dynamically formed by a sharp twist in an N-terminal helical segment extending from Leu370 to Arg378, in agreement with independently generated NOESY data. In its binding pose, the ortho-biphenyl moiety of 10074-G5 was in close proximity to the aromatic ring of Phe375, while the furazan and nitro electron-rich moieties interacted with the positively charged Arg366-367. Binding of 10058-F4 and 10074-G5 to the disordered Myc induced global conformational changes that abrogated Myc-Max interactions. The small molecule-induced bound structures nevertheless remained disordered and differed from the ordered structure induced by Max in the bHLHLZ of Myc. Free and bound structures of the above-mentioned Myc peptides were modeled using the PREDITOR web server232 based on dihedral constraints that were obtained from observed NMR chemical shifts, and further energy minimized using CHARMM27 parameters233. Docking simulations were then performed between the optimized bound structures and respective inhibitors using the AutoDock Lamarckian genetic algorithm (LGA) to obtain the final models representing the most reasonable conformations out of the dynamic ensemble of each complex.230,231 The potential determinants of binding specificity of these small molecules to the nonconventional, flexible binding sites identified in Myc, were further assessed by comparison between the Myc bHLHLZ sequence and those of related Max and Mad bHLHLZ proteins. This further assessment revealed several conservation anomalies where sequence conservation was observed to some extent between Max and Mad but not Myc231. Yin et al191  have previously shown that 10058-F4 and 10074-G5 selectively inhibit Myc-Max but not Mad-81 Max. Out of 22 anomalies scattered throughout the Myc353-437 sequence, four occurred within the 10074-G5 binding site (Asn368, Glu369, Phe375, and Ala376). 5 more occurred within 10058-F4 binding site (Leu404, Ser405, Val406, Gln407, and Ala408; for Glu409 there is no consensus at all) and 3 (the already mentioned Phe375 and Ala376, and also Glu383) occurred within the binding site of yet another inhibitor identified earlier by Yin et al191, 10074-A4, a non-fluorescent compound (discussed further below). The conservation anomalies are indicated with a star (*) below the multiple sequence alignment in Figure 1.9.  Moreover, the early-identified binding regions Myc402-412 and Myc363-381 (Figure 1.28) for compounds 10058-F4 and 10074-G5 contained two of the three clusters of hydrophobic residues found in Myc bHLHLZ: 374-Phe-Phe-Ala-Leu-377 and 401-Ala-Tyr-Ile-Leu-404.231 These were more hydrophobic than the corresponding sequences of other bHLHLZ proteins: Leu-Glu-Lys-Leu and Leu/Met-His-Ile-Lys/Gln in Mad; Leu-Glu-Arg-Leu and Ala-His-Ile-Lys in Mxi; and Phe-His-Ser-Leu and Glu-Tyr-Ile-Gln in Max, respectively. As IDPs generally lack bulky hydrophobic amino acids234 to form well-structured hydrophobic cores, Follis et al231 findings highlighted the importance of minimum hydrophobic content as most relevant for targeting protein-protein interactions of IDPs with small molecules, setting a first practical example.  In 2014, 10074-G5 and 10058-F4 binding sites were confirmed and, importantly, were established on both c- and N-Myc in their monomeric disordered form in the work of Muller et al.235 Subsequent CD and NMR spectroscopy experiments conducted by Hammoudeh et al (2009)236 revealed that compounds 10058-F4, 10074-A4 and 10074-G5 bind specifically the Myc monomer at 3 simultaneous and independent binding sites (Figure 1.28). These sites consisted of contiguous stretched of amino acids: Myc366–375 within the Myc363-381 peptide for 10074-G5 (site 1 in Figure 1.28), Myc375–385 within the Myc370-409 peptide for 10074-A4 (site 2 in 82 Figure 1.28), and Myc402–409 within the Myc402-412 peptide for 10058-F4 (site 3 in Figure 1.28), that matched the highest chemical-shift peaks also found on full-length Myc353-437. The reported Myc-bound models by compounds were generated based on the average of multiple dynamic structures of the intrinsically disordered domain of Myc as previously reported by the group.231 The affinity of 10058-F4 for its cognate binding site was KD = 5.3 μM but reduced more than two-fold when the minimal peptide binding site (residues 402-412) was extended to include the entire bHLHLZ domain (KD = 13 μM). A similar decrease in affinity also occurred in the case of 10074-G5 – from KD = 2.8 μM to 4.4 μM. This suggested that the specificity of site recognition by these small molecules as well as the fine-tuning of binding affinities arose from residues that extended to a certain degree beyond the minimal binding sites defined previously by synthetic peptides.231 Competition titrations using fluorescent polarization performed by Hammoudeh et al.236 further demonstrated that out of the seven inhibitors that were originally identified by Yin et al.191, compounds 10031-B8, 10075-G5, and 10009-G9 also bound at the 10058-F4 site in the H2-LZ region (site 3 Figure 1.28) but with three- to eight-fold lower affinity than 10058-F4. Moreover, compound 10050-C10 (the largest one) bound at the 10074-G5 site in the basic-H1 region (site 1 Figure 1.28), this time with a three-fold higher affinity (KD = 0.9 μM) than 10074-G5, indicating the plasticity of intrinsically disordered binding sites. Compound 10074-A4, unlike 10058-F4 and 10074-G5, is a non-fluorescent chemical that exists as a racemic mixture of two (R- and S-) enantiomers.236 The structural features of the binding interaction between 10074-A4 and its binding site, Myc370-409 (site 2 in Figure 1.28) deduced by CD (KD = 21 μM), were further characterized by NMR, which indicated that the location of 10074-G4 binding was immediate C-terminal to that of 10074-G5 (site 1 in Figure 1.28).236 Significant localized 83 interactions within the H1 and loop regions, as determined by NMR backbone chemical shifts, were observed to include the predominant hydrophobic interactions between 10074-G4 and hydrophobic residues Leu377, Ile381, and Leu384 of the peptide, as well as interactions with Arg378 and Asp379.  A comparison of free and bound peptide models further showed that the compound was enclosed in a cavity shaped by N-terminal residues Phe374, Phe375, Ala376, and Leu377 of the loop region, and Leu377 of the H1 helix. Docking of 10074-A4 to the bound conformation of the peptide showed that the compound was stabilized via extensive hydrophobic interactions. The docking of both enantiomers of 10074-A4 displayed a similar mode of binding and similar docking scores. Poses for both enantiomers were generally consistent with the independently generated NMR NOE data.  The docking simulation provided a general understanding of the binding interaction, but it could not generate precise binding information or identification of a favored binding enantiomer.236 As already mentioned, the group further designed the JKY-2-169 α-mimetic compound, the best of a series of biphenyl based compounds that were intended to specifically recognize the ordered region Myc366-378 that included the binding region of 10074-G5 as well as hydrophobic N-terminal binding region of 10074-A4. Arg372, Phe374, Leu377, and Arg378 delineated the binding site of JKY-2-169 (site 4 in Figure 1.28).212 Nowadays, MD simulations are state-of-the-art computational techniques massively employed to study and understand the characteristics of highly dynamic conformations of IDPs. MD is the computational method that best complements X-ray and NMR-based techniques that may lack sufficient resolution to fully investigate IDPs, for they cannot reveal the complete dynamic behavior of IDPs in atomistic details, nonetheless experimental structural data, such as 84 NMR chemical shifts, are necessary to evaluate MD ensemble models for improved accuracy.237 MD is markedly useful to study weak interactions between a ligand molecule and an IDP, since these interactions are often associated with rapid conformational changes.238 In a 2012 report, Michel et al.239, using explicit solvent enhanced sampling meta-dynamics simulations on 10058-F4, reported that the molecule interacts with a Myc402-412 conformational ensemble at multiple different dynamically formed pockets within the peptide segment and that the ligand binding was driven by weak and non-specific interactions. Nonetheless, 10058-F4 made preferential contacts with Tyr402, Ile403, and Leu404, which lie within the most hydrophobic segment of the entire bHLHLZ domain. In Max, such segment does not exist. In 2013, Jin et al134 studied the binding characteristics of 10074-A4, the only Myc compound known insofar to bind to the loop region, by using extensive MD simulations, with both implicit and explicit solvent models. Jin et al134 found that 10074-A4 associated with Myc370-409 and behaved like a “ligand cloud” around a “protein cloud”, with distinct features from that of a non-binding negative control peptide segment, Myc410-437. 10074-A4 bound Myc370-409 at different sites along Myc disordered conformations. The obtained MD results were consistent with the previously reported NMR data, thus providing evidence of the reliability of computational approaches. Importantly, Jin et al134 obtained a conformational ensemble of apo (free) and holo (bound) forms that are suitable for use as reference structures for drug design targeting Myc through SBDD approaches. The above-reported simulations consisted of 34.5 μs total time in implicit solvent with 30 REMD replicas each of 1.15 μs for four groups. Namely, their own-built extended structure of the Myc370-409 peptide (150 ns), the apo NMR defined structure (270 ns), the structure obtained after 80-ns simulation time from the extended structure (210 ns), and the most occupied 85 representative conformations that were previously generated by REMD (replica-exchange molecular dynamics) from the built-in extended structure (520 ns). The explicit solvent simulations were conducted for a total of 21 μs for three groups: 1 μs of seven trajectories, for each 1 apo and 2 holo for chiral forms of A4, using as starting structure the NMR refined structure, and six of the representative conformations generated by the 150-ns extended peptide REMD simulations. The negative peptide control was simulated for 80 ns in implicit solvent with the extended Myc410-437 peptide used as initial structure. The final structure was applied on all-atom explicit simulations, 1 μs each per 10074-A4 chiral forms. Naturally, VS aided by MD re-emerged in 2016 as a powerful synergistic combination of computational methods for the discovery of novel inhibitors targeting representative conformations of IDPs at multiple sites. Indeed, Yu et al.135 utilized the reference ensemble of the apo and holo conformations of the Myc370-409 intrinsically disordered region generated earlier134 for in silico binding site identification using the CAVITY program240. Three sites were identified: two sites in the apo conformation, termed Apo1 in Myc379-409 and Apo2 in Myc370-386, and one site in the holo conformation, Holo1 in Myc374-388. Virtual screening was then conducted targeting the aforementioned sites that led to the discovery of the PKUMDL-YC prefixed inhibitors that are described in subsection 1.3.2.3. Structurally, compounds PKUMDL-YC-1201, PKUMDL-YC-1202, PKUMDL-YC-1203, and PKUMDL-YC-1204 are somewhat related as they share two common chemical features, namely thiourea and acylamino groups, which as docked engaged through their amino hydrogen atoms in H-bonding interactions with the backbone oxygen atoms of Glu383 and Asp379, respectively. These observations prompted Yu et al241 to conduct a structure-activity relationships (SAR) analysis to gain insights into the determinants of their binding affinity. In 86 their docking mode, the thiourea and acylamino groups of PKUMDL-YC-1203 and PKUMDL-YC-1204 formed additional hydrogen bonds with the backbone carbonyl of Ile381. On top of these, PKUMDL-YC-1203 and PKUMDL-YC-1204 further engaged Arg378 in H-bonding interactions. PKUMDL-YC-1203 was less active than PKUMDL-YC-1204 due to electrostatic repulsion between the oxygen from the benzyloxy group of PKUMDL-YC-1203 and the backbone oxygen atom from Asp379. Because PKUMDL-YC-1203 and PKUMDL-YC-1204 were less soluble in water, a detailed binding analysis by Saturation-Transfer Difference (STD) NMR was conducted for PKUMDL-YC-1205, the analog of 10074-A4, instead. Five independent 100-nanosecond MD simulations using the molecular docking complexes as initial complex structures further investigated the interactions between PKUMDL-YC-1205 or 10074-A4 (the favored S enantiomer form) and Myc370–409. Yu et al241 showed that, despite the fact that the compounds were residing in their binding sites in the MD initial structures, they hovered along the Myc370–409 structure during the simulation course. To characterize the binding site, the interaction distances within 5 Å between the compounds and residues in Myc370–409 were calculated. PKUMDL-YC-1205 and 10074-A4 showed a strong tendency to bind to the N-terminal residues 370–387 and 375–385 in Myc370–409, respectively (site 5 in Figure 1.28). Particular contacts for PKUMDL-YC-1205 within the Myc370–387 region were modeled as Lys371, Arg372, Ser373 (confirmed by NMR TOCSY spectrum), Phe375, Gln380, Ile381, and Pro382. Yu et al241 study provided a useful strategy for structure-based drug discovery targeting IDPs and proof-of-concept that “protein clouds” are druggable. Also mentioned earlier is the Myc-Max inhibitor 7594-0035 discovered by Yao et al223 through docking-based virtual screening using the X-ray structure of Myc-Max complexed with DNA [75] (PDB ID: 1NKP). Interestingly, although the group considered the dimer structure for 87 docking, no attempt has been made to identify new pockets on the dimer surface in silico. Instead, the investigators considered the previously described disordered region, Arg363-Ile381 of Myc bound by the 10074-G5 inhibitor, but in ordered form as the site for inhibitor binding, given that there is no small-molecule ligand bound to the Myc-Max complex in the 1NKP X-ray structure. Therefore, during the preparation of the protein for docking, only the region Arg363-Ile381 of the ordered dimer was set as the docking site (site 6 in Figure 1.28), and the loop382-392 region was removed. The rational for excluding the loop was that in the stable ordered state of Myc, the loop is in close proximity to the pocket, with the Lys392 side chain inserting into the active site.223 In its docking pose, compound 7594-0035 formed a strong H-bonding network with the side chains of residues Arg364, Arg367, and Asn368, and the main chain of residue Arg364. Apart from the polar interactions, the indole ring of compound 7594-0035 fitted in a hydrophobic cavity formed by Leu379, Lys371, Phe374, and Phe375. Likely both hydrogen-bonding and hydrophobic interactions played key roles in the binding of 7594-0035 to Myc.223  Through “blind docking” and clustering analysis, Jiang et al229 identified three main sites, coined “site 1, site 2, and site 3” utilizing both the ordered Max-Max122 and Myc-Max123 X-ray structures to discover in silico specific Max-Max stabilizers that inhibited Myc-Max function (as described in the subsection 1.3.2.3). “Site 1” represents a deep concave protein surface between the two basic helices of the Max-Max dimer at the DNA interface. VS results for the Max homodimer and the Myc-Max heterodimer were similar for site 1, which is consistent with the high degree of structural similarity between the dimers in the basic region. Its high binding strength arose from both strong electropositive potential of several basic residues: Arg35, Arg36, and Arg60 from both Max monomers, and His38 and Lys40 from each Max monomer (see 88 Figure 1.28 for equivalent residues from Myc). NSC131615, as the representative compound for site 1, counteracted the charges while further engaging in H-bonding with residues in the site.  “Site 2” was found as a more hydrophobic pocket-neighboring site 1. Compounds targeting this site (e.g. NSC292215) interacted with several positively charged residues: His44, Arg47, and Arg60 from the basic region of one chain of Max, as well as Arg35 from the other chain, and with neutral HLH region residues (e.g. Ile63) nearby. Minor inconsistencies between the Max-Max and Myc-Max dimers were observed in the HLH region of this binding site, such as the substitution of Phe922 in Myc for His44 in Max and a shift of the loop backbone in Myc that is caused by an insertion at residue 933 (see Figure 1.28 for numbering).  “Site3” was a shallow C-shaped cavity at the intersection of HLH and LZ regions. The prominent structural feature for site 3 binding compounds, representative of which is compound NSC13728, was the presence of rigid, flat moieties that fitted within the relatively narrow pocket between LZ residues Tyr70 and Arg75 and HLH loop residue Pro51, with Met74 and Asn78 neighboring the site. Arg75 was the only positively charged residue near the binding site. In the docking predictions, site 3-binding compounds rarely engaged in hydrogen-bonding with side chains neighboring the cavity. A major disparity was observed between site 3 identified on the Myc-Max dimer relative to that of Max-Max. In the latter complex, the pocket used by site 3-binding compounds was blocked in the Myc-Max structure by Max residues Arg254 and Gln261, disfavoring compound binding to this site in Myc-Max. Max-Max stabilizers at site 3 were predicted to dock instead to various minor sites in the Myc-Max HLH region. The latter aforementioned computational studies provide evidence that ordered dimer structures are more promising in silico targets for therapeutic inhibition over disordered monomeric forms.  89 1.3.2.5 The Need for MYC-MAX Small Molecule Inhibitors As described above, several research and development groups took on the challenging task of targeting Myc with small molecules. Despite significant efforts and promising preliminary results, no effective anti-Myc drugs have been developed yet. Targeting repressive DNA topologies in the Myc promoter region with G-quadruplex stabilizers is one direct strategy to block Myc gene expression in cancer. Except for the CX series of G4 stabilizers in early clinical trials, compound drug-likeliness and off-target side effects are major limitations of current G4 stabilizers that remain main concerns for future development efforts.  Targeting the Myc-Max protein-protein and/or protein-DNA interactions with small molecules is the other direct strategy aimed at altering Myc-Max transcriptional output, which, to date, did not yield an inhibitor to enter clinical trials. The majority of literature reported small molecule inhibitors of Myc-Max function, described in section 1.3.2 above, while affecting growth in cancer-specific cell lines with few showing effectiveness in animal models, they generally demonstrate suboptimal in vivo safety and efficacy profiles restricting their clinical utility. One major pitfall is their general lack of potency. Their chemical structures and physicochemical properties reflect their limited translational potential. Many compounds in the latter category bear toxic and/or promiscuous moieties, as well as reactive centers and/or metabolic liabilities. Some are too large and/or are insoluble in water or are too lipophilic, limiting their permeability and their bioavailability. It is unknown whether further preclinical development of the most promising inhibitors (e.g. Mycro3, KJ-Pyr-9) is ongoing. To prove their target selectivity, the binding mode of these compounds to Myc-Max remains to be determined. Overall, the reported Myc-Max inhibitors require further optimization 90 to improve their in vivo activity, pharmacokinetics and bioavailability to become orally available, daily-administered clinically viable drugs.  As an effective Myc drug would make a huge impact in the clinic, development of novel Myc-Max inhibitors represents an urgent and unmet need. Computational approaches have the intrinsic power to accelerate the discovery of novel Myc-Max inhibitors by taking advantage of very large chemical databases emerging in recent years, available crystallographic structures, more accurate methods of binding affinity estimation, improved force fields, and state-of-the-art molecular dynamics simulations.  1.4 Drug Discovery and Development One of the current challenges faced by the pharmaceutical industry is to increase the efficiency of drug development given that the approval rate of drugs entering clinical trials is less than 12% and yearly very few drugs reach the market.242 The development of an effective and safe new prescription drug that gains market approval is a time-consuming and increasingly expensive undertaking with a total average cost of US$2.6 billion.242 Moreover, an ongoing struggle for the pharmaceutical industry is to find new promising drug targets as the current drug therapy utilizes less than 10% of the potential targetable space. Among ~1600 FDA-approved drugs for ~500 targets, there is a significant bias toward compounds that target well-characterized proteins mostly belonging to the families of G protein-coupled receptors (GPCRs), kinases, ion channels, and ligand-inducible nuclear receptors, with well-defined and deeply buried pockets that readily interact with small molecules.243 A large number of cancer-associated proteins with intrinsically disordered regions, in particular over 1600 human transcription factors (TF) accounting for ~20% of all oncogenes, have been 91 pessimistically considered ‘undruggable’ and, as such, have been for a long time largely neglected in drug discovery campaigns.244,245 Targeting transcription factors and their disorder with small molecules that could effectively stop TF oncogenic functions is nowadays gearing up in the realm of novel drug discovery to turn the undruggables into druggables. Such strategies include modulation of their expression or degradation, blocking protein-protein interactions and/or preventing DNA binding at DNA-interacting sites driven by enhanced knowledge on their structure-function relationships, and their dynamics of binding to DNA harnessed by technological advances. However, discovery and optimization of selective TF inhibitors that could further translate into novel TF-directed drugs is still a considerable challenge, compared, for instance, to the kinase protein family inhibitors, owing to the lack of structural similarity, folding patterns and known binding sites within TFs, as well as the large complexes and various signaling pathways they participate in.246  The discovery of targeted small molecules and their development into potent, selective and orally available clinical drugs follows a lengthy and costly process. The initial drug discovery stage involves target identification and validation, hit identification and lead optimization. The development stage encompasses preclinical studies and clinical trials.  Figure 1.29 Drug discovery and development pipeline.  A key starting point in the drug discovery pipeline is proper selection and validation of a macromolecular target (protein, DNA, RNA or complexes) upon which a potential therapeutic 92 drug can act to elicit a biological response that can translate into a beneficial therapeutic effect against an intended disease.247 In the last several decades, modern genetics, cell and molecular biology have unlocked the genetic code of humans248 and of animals as well as of pathogenic organisms making it possible to identify and characterize the macromolecules and biochemical pathways involved in disease and to make these target biomolecules available to aid in drug design. Advances in understanding the etiology, phenotypic characteristics and mechanism of the targeted disease as well as the advent of omics technologies provided a large amount of normal versus disease typing data that, when mined and analyzed by informatics means, greatly accelerated the identification and prioritization of novel potential drug targets.249 Validation represents the process in which modulation of the selected target leads to desired changes in in vitro and in vivo disease-specific models, which may or may not exist thus need to be developed. Some techniques used to validate drug targets are genetic knockouts (i.e. gene knockout usually in mice), conditional knockouts (i.e. tissue specific and/or induced knockouts), and RNA interference (i.e. siRNA, shRNA, miRNA) that block mRNA and consequently reduce protein synthesis.250 After identification and validation of a biological target of interest and establishment of appropriate cell-free biochemical and cell-based functional assays as well as development of animal models to translate in vitro to in vivo activity, hit identification is the next sequential step to identify and confirm active molecules (hits). Traditionally, hit identification efforts relied on high-throughput screens (HTS) of combinatorial libraries. While demonstrating a high success rate in hit generation, HTS is a very 93 expensive endeavor further limited by a lack of prior understanding of the molecular mechanism behind the activity of identified hits.251 Thus, emerging hits require confirmation of the designed structure, screening activity and feasibility of re-synthesis. In contrast, virtual screening is a very efficient approach to conduct in silico searches over millions of chemicals, increasing yields of potential hits (Figure 1.30).252  Figure 1.30 General workflow for virtual screening large chemical libraries.  The hit-to-lead optimization process aims to transform a confirmed hit into a prospective preclinical candidate through iterative cycles of computer-assisted drug design, medicinal chemistry synthesis and experimental validation (Figure 1.31).   94  Figure 1.31 Schematic of hit-to-lead optimization process. The challenge in this stage is the simultaneous optimization of potency and drug-like profile while maintaining the ligand properties required for safety and biological activity of the intended target and disease. Analogs of confirmed hits are prioritized based on in silico, in vitro and in vivo experiments to select candidates with optimal potential for development into a safe and effective drug.  Structure-activity relationships (SAR) are developed in this stage to determine pharmacokinetics (PK) and pharmacodynamics (PD) properties that can be applied to derivatives in view of synthesis and evaluation.253 The loose definition of PK is what the body does to a drug in terms of dynamics of absorption, distribution, metabolism and elimination (ADME). PD on the other hand is what the drug does to the body describing the molecular mechanisms of drug action in terms of dose-response relationships, in other words the relationships between the amount (concentration) of drug administered via a particular route (e.g. oral, intravenous, intraperitoneal) and the measured physiologic effect of the drug, advantageous (efficacy) and untoward (toxicity) reactions. Linking PK/PD represents the time course of a drug mechanism of 95 action, as PK relates drug concentration to time and PD relates drug effect to concentration to obtain the drug effect in time.254 Preclinical stage involves drug synthesis and formulation research of typically 10,000 molecules of which less than 100 chemicals will enter preclinical development, which involves in vivo animal efficacy (anti-tumor effect) and safety evaluation in doses equivalent to approximated human exposures, and bioavailability, pharmacokinetics/pharmacodynamics (PK/PD) and preliminary toxicology evaluation.255-258 To advance a promising drug to further studies in human subjects, results from preclinical investigations require FDA-approval of an IND (Investigational New Drug Application). Clinical development typically involves three phases that aim to establish the optimal dosage, frequencies and administration routes and to demonstrate the safety and efficacy of an investigational drug in humans, prior to applying for FDA-approval in the form of an NDA (New Drug Application) to market the drug. In phase I trials, dose-escalation studies starting from a single dose based on findings from preclinical trials are conducted in a small number (less than one hundred) of healthy volunteers and also patients with corresponding disease to establish tolerability (i.e. dosage levels and side effects), pharmacokinetic properties and safety.259  Phase II assesses the effectiveness and safety of the drug, confirms PK/PD, evaluates drug-drug interactions and determines the maximum tolerated dose (i.e. minimum dose for maximum efficacy) in large numbers (few hundreds) of actual patients who have the disease of interest. Because of their limited size, Phase II trials lack the power to establish the drug efficacy.260 Subsequent Phase III trials are considered therapeutic confirmatory as they are conducted in large numbers of patients (up to few thousands) in multi-clinical centers in well-designed and 96 well-controlled independent trials to prove sufficient levels of efficacy and safety before new drugs are registered and approved for marketing by FDA. The aim of Phase III clinical trials is to demonstrate the treatment effects of the experimental therapy compared to the standard of care therapy for the disease of interest.261 Post-marketing surveillance is conducted after an NDA receives FDA approval in Phase IV in the general population of patients given drug for therapy to assess further adverse reactions and patterns of drug utilization.254  1.4.1 Computer-Aided Drug Discovery In recent years, computer-aided drug discovery (CADD) has become an integral part of an early drug discovery campaign. While greatly accelerating the hit/lead generation stage, CADD techniques help reduce the costs associated with high-rates of failure of drug candidates in late-stage clinical trials. A major advantage of CADD relative to the traditional HTS technique is that it uses a targeted knowledge-driven approach that can provide valuable insights about the interaction patterns and binding affinity between the target and ligand, thus explaining the molecular basis of therapeutic activity.252 Advancements in computational methods and machine learning techniques applied to CADD, as well as exponential growth of purchasable and/or readily synthesizable chemical space and increased availability of structural data makes possible to expand the repertoire of drug targets and to generate novel small molecule inhibitors for a broad diversity of biological targets including previously neglected ones. The impact of computing on drug discovery has intensified in recent years due to rapid development of graphics processing unit (GPU) accelerated hardware architectures, massively parallel general-purpose or specialized 97 supercomputing, and advanced algorithms for high-level computations on long timescales previously unaffordable.  To guide and expedite discovery of promising drug candidates, computer-aided drug design takes advantage of structural knowledge of the drug target (structure-based drug design, SBDD) on one hand, or known ligands with demonstrated bioactivity (ligand-based drug design, LBDD) on the other. The SBDD methods rely on the structural information of the therapeutic target, which typically is derived from X-ray crystallography, NMR spectroscopy, homology modeling and more recently from cryo-EM microscopy, to identify key features of protein-drug interactions at high-resolution. The SBDD methods include molecular docking and molecular dynamics (MD) simulations. The LBDD approach, on the other hand, relies on pharmacophore modeling, chemical similarity expansion and QSAR. The synergistic use of diverse SBDD and LBDD methods, as well as their integration with modern experimental techniques can effectively speed up rational drug discovery research both in academia and in industry.253 Molecular docking. Docking is a central and rapid SBDD technology employed in large-scale virtual screening and lead optimization efforts to predict protein-ligand interactions and binding free energies.262 Docking algorithms aim to find an optimal placement of a flexible ligand in the binding site of the receptor by generally applying a lock-and-key fitting model, and to generate a score for the resulting docking conformation, referred to as a binding pose, by using a scoring function that approximates the ligand binding free energy. Docking scores serve as bases for ranking and prioritization of virtually screened compounds as they provide a means to distinguish high-affinity binders from those with weak or no binding ability. Several commercial docking programs are currently available, Glide220,221, ICM263, and OpenEye suite264. For example, the Glide docking score is an empirical scoring function comprised of terms that 98 account for the physics of the protein-ligand binding process, including a lipophilic-lipophilic term, hydrogen bond terms, a rotatable bond penalty, and contributions from the Coulomb and vdW interaction energies between the ligand and the protein.143,144 For extra precision, Glide score incorporates terms to account for hydrophobic enclosure for protein-ligand complexes.265 As docking programs greatly differ in their underlying conformational sampling algorithms (i.e. systematic, stochastic, or simulations-based) and scoring functions (i.e. force field-based, empirical or knowledge/statistical-based), their parallel usage is a practical way to find consensus and boost confidence in a docking pose.266  A major drawback of routinely applied static SBDD docking calculations that follow a rigid lock-and-key binding model is the unaccountability for structural flexibility of the target (i.e. induced-fit), improper consideration of solvation effects, and lack of conformational sampling of significant rearrangements generally observed during binding of the ligand to the target critical for identification and optimization of a drug candidate. Furthermore, molecular docking scoring functions are parameterized and benchmarked on a limited, well-characterized set of biological targets, and thus are insufficiently designed or optimized for general purpose, as they do not incorporate the large majority of the current drug and target space.267,268 By using approximated scoring functions, molecular docking generally fails to provide reliable estimates of binding energies that correlate with experimentally determined values, thus post-processing methods are necessary to validate and/or refine docking solutions and obtain accurate free energies of binding. MM/GBSA. Molecular mechanics with generalized Born and surface area continuum solvent model (MM/GBSA) is a fast force field-based, implicit solvent method for computing the free energy of binding of protein-ligand complexes, with improved accuracy over docking as it 99 accounts for target flexibility and solvation effects.269 MM/GBSA approximates the total energy of a system as a summation of two energy terms: the gas-phase potential energy computed by a molecular mechanics (MM) force field of choice (e.g. OPLS3 or AMBER), and the solvation free energy calculated using the generalized Born surface area (GBSA) model (eq 1).  Etot = Evac + ∆Gsolv (1) ∆Gsolv = ∆Gel + ∆Gnonpolar (2) In a typical GBSA implicit solvation model, the solvation free energy (eq 2) comprises polar contributions calculated from the generalized Born (GB) implicit solvent model, an efficient approximation of the theoretically rigorous but computationally expensive Poisson-Boltzmann (PB) continuum electrostatic model, and nonpolar contributions captured by the (solvent accessible) surface area (SA)/hydrophobic effect term. The polar term, ∆Gel, represents the free energy of first removing all charges in the vacuum, and then added them back in the presence of a continuum solvent environment while the nonpolar term, ∆Gnonpolar, represents the free energy of solvating a molecule from which all charges have been removed (i.e. the partial charge on every atom is set to zero).270 Despite its limitations of implicit solvation, lack of conformational sampling and entropy treatment, MM/GBSA is a computationally efficient method as it can reduce the size and complexity associated with explicit treatment of bulk solvent-solvent and protein-solvent interactions.270 It is particularly useful, nonetheless, in rescoring high-confidence docking poses based on their predicted relative binding free energies (affinities) as implemented in Schrӧdinger’s Prime MM/GBSA.271 Molecular dynamics simulations. Molecular dynamics (MD) simulations is an advanced CADD technique employed to study and understand the characteristics of highly dynamic conformations of macromolecular complexes in explicit solvent, a more realistic, physiological 100 relevant environment.272 The main advantage of MD springs from the explicit treatment of structural flexibility and solvent-associated entropic effects. MD simulations provide a detailed all-atom description of molecular interactions within a system as it evolves in time in the form of atomic trajectories computed by numerical integration of Newton’s equations of motion (eq. 3) for each atom in a solvated and neutralized system. fi(t) = miai(t) = −𝜕𝜕𝜕𝜕�𝑋𝑋(𝑡𝑡)�𝜕𝜕𝑥𝑥𝑖𝑖(𝑡𝑡)    (3) fi(t) is the force acting in the ith atom of the system at a given point in time t, ai(t) is the corresponding acceleration and mi is the mass. x(t) is the vector of atomic coordinates (positions in Cartesian space of interacting atoms in the configuration of the system at time t). U is an empirical force field (FF) that approximates the net atomic force in biochemical systems.  A FF is comprised of the combination of a potential energy function – a mathematical model used to calculate the energy of the system as a function of its structure (i.e. atomic coordinates), and optimized force constants – parameters used in the potential energy function that are essential to accurately and specifically treat physical interactions within the system.273 For instance, the functional form of the OPLS force field274 is a summation of bonded and nonbonded energy terms given by equation 4. U = ∑bonds Kr(r – req)2 + ∑angles Kθ(θ – θeq)2               + ∑dihedrals [𝑉𝑉12(1 + cos φ) +  𝑉𝑉22(1 - cos 2φ) +  𝑉𝑉32(1 + cos 3φ) +  𝑉𝑉42(1 - cos 4φ)]      + ∑i<j [qiqje2/rij + 4εij(σij12/rij12 - σij6/rij6)]fij  (4) Bonded terms include bond stretching and angle bending energies represented by harmonic terms where K denotes force constants and eq the equilibrium (or reference) values of bond length r and angle θ, as well as a torsion term that sums over all dihedral angles. Nonbonded 101 interactions are represented by sums over Coulomb and Lennard-Jones energies that model electrostatic and van der Waals (vdW) interactions between all pairs of interacting atoms i and j. In eq 4, qi,j are partial charges, rij is the separation between interacting sites, and σij are εij adjustable parameters that represent the collision diameter (the separation where the Lennard-Jones 12-6 potential is zero) and the energy well depth, respectively. For biological macromolecule simulations, numerical methods are necessary to split the integration of Newton’s equation of motion in discrete time-steps.275 Several integrators exists, one of which is the simple and widely used Verlet integrator.276 To propagate an MD trajectory, the Verlet integrator calculates next positions in time from current positions, velocities and accelerations, the latter in turn calculated by the forces acting on each atom by taking the negative of the first derivative of the potential energy with respect to positions (eq 3). MD simulations are then fully carried out under the specific force field, periodic boundary conditions (for better description of bulk properties) and a particular thermodynamic ensemble, such as the isothermal-isobaric NPT ensemble in which the number of atoms (N), the pressure (P) and the temperature of the system (T) are conserved (kept constant) (for better mimicking macroscopic behavior).277 Theoretically, MD simulations in explicit solvent with accurate potentials and run over sufficient time could model biological relevant concerted motions, such as large conformational changes, secondary and tertiary structure formation and protein folding. In practice, MD simulations are computationally expensive, time-consuming and dependent on development of improved force fields. Nonetheless, modern MD simulations codes optimized for GPU (graphics processing units) computing architectures can, in a short time span, run MD simulations of target-ligand complexes that take into account the flexibility of both the target and the ligand and 102 provide detailed information on atomic motions at long nanoseconds timescales, which are rather difficult to obtain experimentally.  MD is the computational method that best complements X-ray and NMR-based techniques that may lack sufficient resolution to fully investigate challenging complexes, for they cannot reveal their complete dynamic behavior in atomistic details, nonetheless experimental structural data, such as NMR chemical shifts, are necessary to evaluate MD generated ensemble models for improved accuracy.237 MD is markedly useful to study weak interactions between a ligand molecule and intrinsically disordered protein (or protein region), since these interactions are often associated with rapid conformational changes.238 Besides providing an efficient way to generate conformational ensembles for a target and/or for a ligand and to study target-ligand interactions at an atomic level of detail, more advanced MD methods provide enhanced conformational sampling as well as accurate estimates of relative free energies of binding.  Free Energy Perturbation MD simulations. Accurate prediction of protein-ligand free energies of binding is a principal goal of CADD. Free energy perturbation (FEP) is a physics-based rigorous method that employs atomistic molecular dynamics simulations in explicit solvent to determine the free energy difference between (two) related ligands as they are morphed via alchemical pathways typically used in medicinal chemistry.278 As such, FEP is particularly useful in drug discovery lead optimization efforts as it can accurately rank series of congeneric derivatives and drive chemical synthesis decisions. It has been demonstrated that compounds predicted to be potent based on FEP free energy differences (∆G ≈ RTln(IC50)) achieved a significant reduction in the number of false positives relative to compounds synthesized as a result of other computational approaches.278 Desmond FEP279,280, part of Schrӧdinger’s suite of programs, is a robust tool for performing and simulating ligand mutations 103 and derive relative binding free energy differences. It takes advantage of the improved OPLS3 force field, enhanced sampling relative to standard MD simulations by incorporation of the replica exchange with solute tempering (REST) methodology281-283 and optimized algorithm for execution on GPU-enabled clusters.284 Chemical similarity. Chemical similarity search is a fundamental technique for ligand-based drug discovery that aims to identify compounds with structures and bioactivities similar to a query molecule. It relies on the assumption that compounds with similar structures most likely have similar activities as stated by the ‘chemical similarity principle’.285 A major drawback of the technique is that the chemical similarity principle does not always hold true as exemplified by ‘activity cliffs’ where small chemical modifications of functional groups can cause large variations in activity.286 The Tanimoto coefficient (Tc) similarity score is the most common measure to evaluate the structural similarity between two molecules both in two or three dimensions.287 The 2D Tc is based on chemical fingerprints, such as Molecular ACCsess System (MACCS) substructure 2D binary fingerprints and represents the fraction of bits (i.e. 0 and 1, indicative of absence or presence of specific substructure keys) shared by two binary-encoded feature vectors of compared compounds. A limitation of 2D Tc is that, while numerically can indicate that two compounds are overall similar, does not provide information on the specific chemical groups they share. The 3D Tc is a chemical similarity metric based on 3D structural features of compounds that calculates the fraction of shared molecular volumes between two compared chemicals.287,288 A representative implementation of 3D Tc is the rapid overlap of chemical structure (ROCS)289 program, a most popular shape and color similarity in drug discovery. In ROCS, molecular volumes represent shape: volumes overlay indicate similarity and any volume mismatch is accounted as a measure of dissimilarity. To refine shape-based 104 overlays, ROCS includes a color atom-based force field to measure chemical similarity between chemicals in terms of donor, acceptor, anion, cation, hydrophobes and rings. Pharmacophore modeling. Pharmacophore modeling is a powerful computational technique to generate and use 3D information to search for novel active compounds.290 It can be both ligand-based and structure-based. Ligand-based pharmacophore modeling is particularly useful when no target geometry is available and utilizes multiple 3D structures of known ligands or active compounds to extract their common chemical features. The structure-based pharmacophore approach works directly with the 3D structure of a macromolecular target or target-ligand complexes and is useful in identifying ligand-binding sites and key interaction points between ligands and the target.  A pharmacophore is an abstract concept defined as “the ensemble of steric and electronic features that is necessary to ensure optimal supramolecular interactions with a specific biological target structure and to trigger (or to block) its biological response”.291 In other words, the essential features responsible for molecular recognition and binding of a ligand to the target and their arrangement in 3D space. The set of features of a pharmacophore include hydrogen-bond donors or acceptors, positively or negatively charged groups, hydrophobic regions and aromatic rings, each assigned a tolerance radius. Once generated, a pharmacophore model is readily applicable to virtual screening and/or filtering of chemical databases thus greatly complementing docking-based virtual screening while reducing its limitations, as well as de novo design and lead optimization.290  Quantitative Structure-Activity Relationships (QSAR). QSAR292 is a broadly used cheminformatics approach that relates structural properties of molecules to their biological activity through a vast array of chemical descriptors and application of machine learning 105 algorithms. Chemical descriptors are numerical features extracted from chemical structures that represent the core of QSAR modeling as they reflect various levels of chemical structure representation (i.e. dimensionality), from 0D through 6D.293  QSAR descriptors 0D descriptors are the simplest and most commonly used constitutional or count descriptors represented by a scalar that reflect the chemical composition without any information about the molecular geometry or atom connectivity. Molecular weight, logP, and counts of atoms, rotatable bonds, rings, H-bonding acceptors and donor are examples of 0D descriptors.  1D descriptors include lists of structural fragments, fingerprints (uniquely encoded numerical strings that contains information on the structure) and SMILES strings (simplified molecular-input line-entry system, a specification in form of line notation using ASCII characters to describe the structure of chemical species).  2D chemical descriptors describe properties computed from two-dimensional representations of molecules, to include topological indices and autocorrelation vectors. 3D descriptors are geometrical descriptors that extract chemical features from 3D coordinate representations and are the most sensitive to structural conformations and variations among ligands.  3D descriptors include substituent constants, surface and volume parameters and quantum-chemical descriptors. A limitation of 3D-QSAR chemical descriptors is the computational complexity of conformer generation and structural alignments that may not correspond to relevant bioactive conformations.294  4D descriptors were thus developed to account for ligand conformations and orientations in the protein binding site. In addition, 5D- and 6D-QSAR models incorporate flexibility, induced-106 fit, and solvation effects to address their critical role for accurate prediction of protein-ligand interactions.253 Inductive descriptors. Pioneered by Dr. Cherkasov, ‘inductive’ descriptors are 3D-QSAR parameters that utilize atomic-scale molecular information to describe whole molecules, based on calculated effects of the atomic constituents of the molecule with demonstrated predictive power.295,296 Novel 4D inductive descriptors have been recently developed in house to quantify at atomic level key protein-ligand interactions that enabled the prediction of adverse drug responses by clinically relevant mutants of human androgen receptor.31  QSAR models Two types of QSAR models are broadly used. Regression models establish quantitative relationships between molecular descriptors and the target property capable of predicting activities (e.g. IC50, KD) of novel compounds. Regression models are continuous models as the response is a continuous numerical property. Classification models are categorical as they classify instances in distinct classes to generate a discrete qualitative response (e.g. active or inactive). A limitation of QSAR models is that it can provide relatively accurate predictions only for confined series of compounds thus restricting their applicability domain. Application of artificial intelligence (i.e. deep learning) is becoming routine in computational drug discovery as is able to surpass major limitations of QSAR modeling based on traditional machine learning algorithms.  ADMET Prediction. As efficacy and safety are principal goals of any drug design campaign, compounds are evaluated early in the drug discovery process to profile their ADMET properties in an effort to minimize the attrition rate due to poor pharmacokinetics or bioavailability (so called “fail early fail cheap” strategy).297  107 ADMET Predictor of Simulations Plus298 is a particular useful tool for prediction of ADMET properties and overall risk of a set of molecules based on their chemical structure. It can quickly predict over 140 properties utilizing a large number of molecular descriptors and artificial neural network ensemble-based models. Predicted physicochemical properties include logP, pKa, transporters inhibition, solubility and permeability. Pharmocokinetic models predict human plasma protein binding, volume of distribution and fraction unbound in human liver microsomes. Metabolism models predict cytochrome P450 (CYP) sites of metabolism and metabolites for nine CYP isoforms (the majority of Phase I metabolic enzymes responsible for biotransformation of most drugs), CYP kinetic parameters, CYP inhibition for five major CYPs (1A2, 2C9, 2C19, 2D6, and 3A4), human liver microsomal unbound intrinsic clearance, and UGT substrate for nine UGT isoforms (Phase II metabolism). Toxicity prediction covers a large range of toxicities including cardiac (hERG binding), hepatotoxicity, endocrine (estrogen and androgen binding filter), carcinogenicity, and sensitivity (skin, respiratory). ADMET Predictor can rapidly build new QSAR/QSPR models using custom data. Structure-based virtual screening often selects biologically promiscuous compounds, so called ‘frequent hitters’, attributed in part to their hydrophobic character that tend to favor their detection in docking (or to aggregation effects in screening assays).267 Such promiscuous compounds demonstrate unspecific inhibition of multiple targets and are discarded in drug discovery campaigns due to their lack of selectivity. FAF-Drugs299 is a free ADME-Tox tool particularly useful in filtering out such frequent hitters as well as toxic and unstable molecules.  108 1.5 Objective and Rationale of the Study Significant progress has been made in recent years in the treatment of prostate cancer by adoption of novel drugs targeting the AR-signaling pathway both in hormone-sensitive and castrate-resistant settings as most recently demonstrated by the efficacy of second generation AR-targeting therapeutics, enzalutamide and abiraterone. However, despite the prolonged survival benefit for treatment-responsive CRPC patients, resistance develops in majority of cases. Moreover, administration of increasingly potent therapy eventually leads to neuroendocrine transdifferentiation resulting in the most aggressive AR-independent neuroendocrine phenotype of prostate cancer (NEPC). As CRPC and NEPC are currently incurable, novel strategies are required to meet the urgent clinical need for management of these rapidly lethal forms of the disease. Overall, the current therapeutic landscape of prostate cancer solely directed at targeting AR is insufficient. Encoded by major oncogenes, Myc transcription factors have been for decades the most sought after targets in many cancer types. In CRPC and NEPC, the exacerbated expression of c-Myc and N-Myc, respectively, plays critical roles in disease progression and enzalutamide-resistance. c-Myc has been demonstrated to promote AR gene transcription and enhance the stability of both AR and AR-V7 proteins in CRPC cells and PDX models. Importantly, c-Myc knockdown by shRNA as well as inhibition with 10058-F4 c-Myc small molecule prototype alleviated enzalutamide-resistance. N-Myc, on the other hand, is a primary inducer of NEPC.  Among a spectrum of target genes upregulated by oncogenic levels of Myc is the hnRNP A1 splicing factor with similar transforming activity observed in many cancers. In CRPC, overexpressed hnRNP A1 has been shown to selectively stimulate AR-V7 alternative splicing while its knockdown with siRNA suppressed AR-V7 levels and growth of CRPC cells. 109 Moreover, its inhibition with a natural occurring flavonoid, quercetin – a very frequent hitter, resensitized enzalutamide-resistant cell lines and mouse xenografts to enzalutamide in vivo. The selective disruption of protein-protein or, importantly, protein-nucleic acid (i.e. DNA, RNA) interactions is a pinnacle of cancer therapy yet is a challenging endeavor especially when targeting intrinsically disordered proteins or their complexes. As best illustrated by Myc, its inherent disorder in the dimerization and DNA binding regions and lack of any obvious targetable pockets on its unstructured monomeric form, hampered the application of conventional structure-based drug design approaches.  Targeting the functionally mandatory Myc-Max complex proved similarly challenging as the Myc-Max/DNA interface is very large and surface exposed. Moreover, the lack of high-quality structures of the Myc-Max complex ligated with small molecules to evidence key protein interacting residues and serve as guides for rational drug design further restrained therapeutic development efforts. The majority of literature reported small molecule prototype inhibitors that bind and alter the disordered Myc monomer and subsequently block Myc-Max dimerization, as well as the few Myc-Max/DNA disruptors emerged mainly from high-throughput screens (HTS) of limited chemical libraries unlikely to contain clinically optimized structures with only three Myc inhibitors discovered through computational approaches. While these small molecule inhibitors affected growth in cancer-specific cell lines with few showing effectiveness in animal models, they generally demonstrated suboptimal in vivo safety and efficacy profiles. Successful targeting the oncogenic transformation of the Myc-Max complex with small molecules, although considered high-risk, represents the high-reward opportunity long awaited to advance prostate cancer treatment. Application of CADD methods involving novel in silico binding sites identification, large-scale virtual screening to maximize the throughput of structure-110 based drug design, hit identification and lead optimization techniques, both structure- and ligand-based, and ADMET profiling, represents the novel strategy employed in this work to target Myc-Max transforming activity in advanced prostate cancer. The strategy led to the discovery of novel small molecule inhibitors of both c-Myc-Max and N-Myc-Max function, as well as inhibitors of the downstream effector hnRNP A1. Their development involved specific disruption of protein-nucleic acid interactions by targeting the DNA- or RNA-binding domains of their respective targets. The current lead inhibitors are the foundation for ongoing in silico medicinal chemistry, iterative SAR and upcoming pharmacokinetics and anti-tumor growth profiling to further their development into promising drug candidates for in vivo efficacy and safety studies and for potential clinical translation. As individual agents or, importantly, used in combination with current standards of care, the developed Myc-Max inhibitors (both c- and N-) and/or hnRNP A1 inhibitors could serve as a new generation of prospective therapeutics for treatment of CRPC and NEPC. They could not only block the AR signaling pathway in AR-dependent settings but they also could reduce AR-V7 levels to overcome enzalutamide-resistance in CRPC, and in AR-independent setting prevent NEPC differentiation.   1.6 Thesis Layout Chapter 1 (above-presented) provides background information on PCa, drug resistance mechanisms, and the role of MYC major oncogene in PCa progression and therapy-resistance to AR-targeted therapies. It includes a comprehensive overview of the knowledge accumulated to date on MYC regulation, structure and function, its critical role in cancer, and various therapeutic strategies employed to tackle MYC-driven malignant transformation. Moreover, it covers the current advances in the development of small molecule inhibitors of Myc function and 111 the techniques employed in these campaigns. It also includes an overview of drug discovery and development process with emphasis on modern computer-aided drug design methodologies. As indicated in the preface, this knowledge, research and development is included in two publications. Chapter 2 presents materials and methods applied in the current work.  Chapter 3, 4 and 5 summarize the most relevant results of the research aimed at developing novel small molecule inhibitors of MYC potent transforming activity in CRPC and NEPC by the application of state-of-the-art computer-aided drug design methods. As indicated in the preface, two recently published research articles include part of the results presented in chapters 3 and 5.  Chapter 6 provides concluding remarks and future directions.   112 Chapter 2: Materials and Methods All methods applied in chapters 3, 4 and 5 are summarized in this chapter.  2.1  In silico modeling 2.1.1 Protein Modeling Protein structures utilized in this thesis are crystallographically resolved 3D structures deposited in the RCSB Protein Data Bank (PDB) or homology models built using X-ray structural data as template. The X-ray structures include the c-Myc-Max heterodimer bound to its DNA recognition sequence, PDB ID: 1NKP (1.9 Å resolution)123, and the structure of hnRNP A1 splicing factor bound to its RNA recognition sequence, PDB ID: 4YOE (1.92 Å resolution)300. As there is no publically available X-ray structure of N-Myc-Max heterodimer a homology model was built in house at high 0.7 Å resolution using MODELLER301,302 with the c-Myc-Max 1NKP X-ray structure used a template.  Briefly, the sequence of the bHLHLZ domain of the N-Myc protein was obtained from UniProt database (P04198 entry, amino acids 377-474). A PIR (Protein Information Resource) formatted alignment was then prepared between the c-Myc bHLHLZ X-ray structural template and the N-Myc target sequence for input to MODELLER. The tool generates multiple models for the intended target by optimal satisfaction of spatial restraints (i.e. distances, angles, dihedrals, non-bonding interactions obtained from the alignment, force field or statistical preferences), refines the models to minimize violations of these restraints, and generates a DOPE energy score used to assess the quality and select the best structure from the collection of models. The lowest energy N-Myc structure was selected and merged with the Max structure from 1NKP X-ray to build the full N-Myc-Max dimer model using the Molecular Operating Environment (MOE).303 113  2.1.2 Binding Sites Identification The Site Finder algorithm, implemented in the fully integrated drug discovery platform Molecular Operating Environment (MOE), was the method of choice for binding site identification and analysis of the above-mentioned protein structures. Site Finder is a geometric method, which uses alpha spheres (virtual atoms) to probe a protein surface for suitable small-molecule binding pockets. Site Finder first identifies regions of tight atomic packing, filters out highly solvent exposed sites, calculates alpha spheres on sites and classifies them as either hydrophobic or hydrophilic depending on whether the virtual atom is in a good hydrogen bonding spot in the receptor. It then produces a collection of sites based on pruning (alpha spheres corresponding to inaccessible regions or exposed to solvent are eliminated) and clustering (by number and chemical type) of alpha spheres. The sites are then ranked according to their Propensity for Ligand Binding (PLB) score based on the amino acid composition of the pocket.304  2.1.3 Protein Structure Preparation The standard Maestro 9.3 suite, Schrӧdinger LLC305 protein preparation protocol 306,307 was applied to all the PDB or homology structures. The process recommends removal of all crystallographic waters and complexed nucleic acids, followed by addition of hydrogen atoms and missing side chains. Prepared protein structures were then energy-minimized using the OPLS-2005 or the more accurately parameterized OLPS3 force field.   114 2.1.4 Ligand Preparation The chemical substances presented in this thesis were either part of the drug-like sets of ZINC12 or the most recent and largest ZINC15 databases308,309 or were directly downloaded from specialized small molecule screening libraries from established vendors, such as Enamine-REAL. In preparation for docking, each chemical imported into a large MOE molecular database was washed and energy-minimized under the MMFF94x force field and Born solvation as per ligand preparation protocol implemented in MOE.303  Briefly, the wash procedure involves salt disconnection, removal of smaller molecular fragments, deprotonation of strong acids and protonation of strong bases at physiological pH, followed by addition of explicit hydrogen atoms and minimization. The optimized structures exported in SD file format are then ready for use in various docking programs.   2.1.5 Virtual Screening The Glide220,221 program, part of Maestro 9.3 suite, Schrӧdinger LLC, was utilized as the starting point to perform rigid structure-based molecular docking of ~5 million drug-like chemicals. For sampling and scoring of screening compounds, Glide approximates a complete systematic search of the conformational, orientational and positional space of the docked ligand in the receptor site. The search begins with the generation of an equally spaced 2 Å grid that represents the shape and properties of the binding site region of the receptor by a series of hierarchical fields that progressively provide more accurate scoring of a ligand pose.  A relatively small number of low energy conformers generated for the core of the ligand are then systematically positioned on the grid and greedily scored (using pre-computed scores from a 1 Å spacing grid) to narrow the search space. A small number of best scoring poses are 115 further refined and optimized using a stochastic technique (Monte Carlo simulated annealing) with the OPLS force field energy function. For the selection of the best pose, Glide uses a model energy function that combines empirical and force field-based terms. As per the Glide screening protocol, small grids (20 to 30 Å in size) centered on the residues of the predicted binding sites were first generated using the prepared protein structures of the intended targets. Docking was conducted using Glide standard precision (SP) mode with all other settings set to default. The Glide score, an interaction energy score that includes a lipophilic-lipophilic term, H-bonding terms, a rotatable bond penalty, contributions from the Coulomb and vdW interaction energies between the ligand and the protein, and solvation terms, was used to rank the generated docking poses. Potentially weak binders, determined as those having a Glide score worse than a threshold selected on a per-target basis (e.g. Glide score > -5.5 kcal/mol for Myc-Max or Glide score > -6 kcal/mol for hnRNP A1), were discarded. “Blind docking” was additionally performed using Glide in SP mode on large grids defined to encompass the entire structure of the protein target on selected subsets of compounds for use in consensus voting in hit-to-lead optimization. The best Glide scoring compounds were re-docked using various docking programs that differ in their underlying sampling algorithms and scoring functions: eHiTS310, ICM263 and OpenEye Hybrid264,311, in view of consensus voting that can boost confidence in the predicted poses. To illustrate the differences, unlike Glide, eHiTS utilizes a fragment-based docking approach. The ligand is divided into rigid fragments and connecting flexible chains. eHiTS docks all rigid fragments to all possible places in the receptor cavity independently of each other, thus it uses a non-incremental sampling approach that does not rely on conformers generation. It also utilizes a much finer 0.5 Å grid to position exhaustively the fragments and evaluated with a 116 scoring function that is discretized with 0.2 Å binning (relative to 1 Å binning in Glide). Furthermore, in the final minimization phase, eHiTS uses a deterministic gradient optimization to find the minimum energy pose, avoiding the use of stochastic methods.310  OpenEye Hybrid performs a systematic, exhaustive, non-stochastic examination of all possible protein-ligand poses within the binding site, filters for shape complementarity and chemical feature alignment before selecting and optimizing poses using the Chemgauss scoring function.312 It uses a large ensemble of conformers generated with OpenEye Omega2 program.313 Hybrid further reduces the search space based on shape and chemical complementarity to known bound ligands or active compounds.  ICM docking utilizes a stochastic global optimization algorithm (Monte Carlo simulations) that attempts to find the global minimum of an energy function that includes five grid potentials describing the interactions of a flexible ligand with the receptor and the internal conformational energy of the ligand. Thus, it globally optimizes a set of ligand internal coordinates in the space of grid potential maps (with 0.5 Å spacing) calculated for the receptor pocket. ICM utilizes its own tailored and sophisticated scoring function to rank and select the best pose among a number of low energy conformers, generated independent of the protein pocket.263   2.1.6 Consensus Scoring and Voting The overall consensus was built based on top-ranking Glide docking scores (more negative scores correspond to higher binding affinity) and calculation, using MOE scripts314,315, of two main indicators. More specifically, the root mean square deviation (RMSD), an atom-based metric reflecting the deviation in atomic coordinates between poses obtained from the various docking programs calculated using the mol_rmsd.svl script, and predicted pKi, a good indicator 117 of potency, calculated using scoring.svl analysis tool for non-bonded intermolecular interactions. The filtering thresholds used were set to RMSD ≤ 3 Å (an RMSD of 0 indicates perfect superposition; the higher the RMSD the greater the deviation), and pKi ≥ 5 (the larger more potent theoretically). RMSD was also calculated for poses obtained using the Glide “blind docking” virtual screening technique.  Small number of compounds (up to 150) were purchased in a few rounds from established vendors based on favorable Glide scores, RMSD or pKi indicators as well as satisfaction of important interactions with the pocket residues upon visual inspection. Consensus was further enriched by application of ligand-based and more accurate structure-based techniques in the development process.  2.1.7 Pharmacophore Modeling A pharmacophore model of two essential hydrophobic features (1.5 Å diameter each) of the Myc-Max binding site (formed primarily by Leu917, Ile218, Phe921 and Phe222) was built and used to search for and filter matching hits from an initial database of top ranked Glide poses and as a ranking criteria for prioritizing consensus compounds in subsequent rounds of hit-to-lead optimization.   2.1.8 Chemical Similarity Searches To find structural analogs, 3D similarity searches were conducted utilizing the ROCS program from OpenEye289,316 against a large ensemble of conformers consisting of at most 200 conformers for each of the approximately 9 million entries of the drug-like purchasable chemical 118 space of the ZINC15 database317. Omega2 of OpenEye313 was used to generate conformers. Current hits represent the query molecules for ROCS similarity searches.  ROCS is a fast shape-based superposition method, which uses a combination of global three-dimensional shape overlay and color-based chemical complementarity in terms of hydrogen-bond donors, hydrogen-bond acceptors, hydrophobes, anions, cations and rings, to compare the query to a large collection of molecules and rank the matching hitlist according to the TanimotoCombo score, a rigorous measure of shape and color overlap. Compounds with high TanimotoCombo scores that most closely resemble the query structure receive priority for subsequent virtual screening. 2D substructure searches were also conducted using online functionalities of the ZINC and Enamine-REAL chemical repositories to find additional analogs.  2.1.9 MM/GBSA Simulations To obtain estimates of the relative binding free energies (affinities) of compounds to the target site, the docking poses were subjected to MM/GBSA (molecular mechanics with generalized Born and surface area continuum solvent model)318 simulations with implicit solvent as implemented in the Prime program of the Schrödinger software suite271. Each protein-ligand complex was MM minimized utilizing the OPLS3 force field274 and the variable dielectric VSGB 2.0 solvent model.319 The Prime MM/GBSA binding free energy is calculated as the difference between the energy of the bound complex and the energies of the unbound target and inhibitor compounds as shown below: ∆Gbind = Ecomplex(minimized) – Eligand(minimized) – Ereceptor(minimized) Prime energy components include OPLS3 MM contributions: bond, angle, dihedral, Coulomb and van der Waals energies, as well as VSGB2.0 electrostatic and nonpolar solvation energies. 119 VSGB2.0 energy model accounts for solvation effects with generalized Born approximation as well as augmented variable dielectric internal polarization effects of target side chains, and physics-based empirical corrections for modeling directionality of H-bonding interactions, hydrophobic packing and π-π stacking interactions, among others.319 During the minimization of the complexes, binding site residues within 5.0 Å of the ligand were allowed to undergo fluctuations to take into account the protein flexibility, while the rest of the protein structure was kept fixed. MM/GBSA was used as a ranking method not reflective of correct free energies.   2.1.10 Molecular Dynamics Simulations In order to obtain an all-atom description of the molecular interactions between lead inhibitors and the intended target within the identified binding sites, classical molecular dynamics (MD) simulations with explicit solvent were setup and analyzed utilizing the Desmond MD simulation system279 integrated in the Maestro molecular modeling environment.  Target-ligand complexes were prepared for use in MD as initial structures. Hydrogen atoms were added following removal of original hydrogens. The structures were then refined to assign appropriate protonation and tautomeric states for titratable residues and ligand at physiological pH. The systems were neutralized by addition of chloride counterions to balance the protein charges and solvated utilizing the SPC or TIP3P water models. The solvated systems were subjected to the OPLS3 force field under periodic boundary conditions in an orthorhombic box with edges lying 10 Å from the outermost atoms of the system.  The system was allowed to relax prior to the 100 nanoseconds MD production run. MD simulations were conducted under the isothermal-isobaric ensemble with the temperature kept constant at 300 K using the Nosé-Hoover thermostat320 and the pressure kept constant at 1 atm 120 using the Martyna-Tobias-Klein barostat321 with a relaxation time of 2 ps and isotropic coupling. The integrator used to evolve the system was RESPA (reference system propagator algorithm)322 with a timestep scheduling of 2 fs for bonded and non-bonded van derWaals and short-range electrostatic interactions, and 6 fs for non-bonded long-range electrostatic interactions. The cutoff for the Coulombic short-range interactions was set to 9 Å. A harmonic restraint with a 1 kcal/mol force constant was imposed on the alpha carbons of the protein backbone. MD simulations were executed on the GPU-enabled Helios cluster of Compute Canada high-performance computing platform.  2.1.11 Free Energy Perturbations Molecular Dynamics Simulations Desmond FEP+280, part of Schrӧdinger’s suite of programs, was used to derive relative binding free energy differences between morphed derivatives of a congeneric series as the most robust and accurate ranking method predictive of their relative potency.  Desmond FEP+ implementation employs standard Desmond MD simulations using OPLS3 force field and settings as described in 2.1.10, with enhanced sampling by incorporating the replica exchange with solute tempering (REST) methodology.281 FEP/REST enables simulations of a selected subsystem treated with replicas (or lambda windows, by default 12) in a higher effective temperature regime relative to the rest of the system to enable convergence of free energy calculations.  The elevated effective local temperatures focus in the region of the ligand and binding pocket residues where an alchemical transformation is simulated. Protein residues close to the binding pocket may also be included in the enhanced sampling region as part of Desmond/FEP+ automated REST region selection. In FEP/REST, the effective temperature of the enhanced 121 sampling region, as a function of lambda (i.e. coupling parameter of the potential energy function that describes the transformation), gradually increases from room temperature with lambda = 0 for the initial physical state to a much higher temperature with an intermediate lambda value equal to 0.5. Then, the temperature is gradually lowered to room temperature while lambda is increased to 1 corresponding to the final physical state. In this way, the potential energies for the two end-points reach the correct physical states, and enhanced sampling is achieved through the increased effective temperatures of the intermediate lambda windows.323  The relative free binding energy is calculated practically from a thermodynamic pathway that involves two distinct alchemical transformations. A first transformation is used to determine the free energy of morphing an initial ligand into a final ligand in solvent. The second alchemical transformation is used to determine the free energy of morphing the initial ligand into the final ligand in the receptor. The differences between the free energies obtained from these two transformations can be accurately related to the difference in the binding free energy between the two ligands.323 The setup of the FEP+ calculations involved the generation of a multi-cyclic perturbation map of the intended congeneric series transformations utilizing the prepared protein structure and the docking poses of the chemical series imported in Maestro’s environment. The cycle is broken down in individual perturbations each simulated on Compute Canada GPU-deployed Desmond simulation engine and then merged back to provide a “cycle closure” free energy estimates accounting for minimal convergence errors.323  122 2.1.12 ADMET Prediction For control of the ADMET (absorption, distribution, metabolism, excretion and toxicity) profile of prototype inhibitors as a requirement for safety and biological efficacy, SimulationsPlus ADMET Predictor software298 was used as a ranking tool to eliminate compounds containing highly toxic moieties and as a guide to predict metabolic liabilities in the hit-to-lead optimization process. FAF-Drugs299 ADME-Tox tool was used to filter out frequent hitters as well as toxic and unstable molecules.  2.2 Experimental Validation 2.2.1 Cell Culture and Reagents LNCaP, 22Rv1 and LASCPC-01 human PCa cells and IMR32 neuroblastoma cells were purchased from the American Type Culture Collection (ATCC) and grown in RPMI 1640 supplemented with 5 to 10% fetal bovine serum (FBS). HO15.19 Myc-null rat fibroblast cells were a generous gift from John Sidivy at Brown University and were cultured in Dulbecco’s modified Eagle’s medium DMEM (ATCC 30-2002) supplemented with 10% FBS. 10058-F4 and 10074-G5 chemicals were ordered from Sigma.  The UBE2C reporter plasmid was purchased from GeneCopoeia (product ID #HPRM16429). The Biolux Gaussia luciferase assay kit was purchased from New England Biolab (#E3300L). PrestoBlue cell viability reagent was purchased from Invitrogen #A-13262.  2.2.2 Transfection and Reporter Assays Cell transfection was performed using TransIT-2020 transfection reagents according to the manufacturer’s instructions (Mirus). LNCaP cells were plated at 10000 cells per well and treated 123 for 1 day with compounds. Myc reporter activity was measured using the Cignal Myc Reporter Assay Kit from Qiagen (#336841) according to the manufacturer’s instructions. For the UBE2C reporter assay, 22Rv1 cells were plated at 10000 cells per well in 96-well plates in RPMI media supplemented with 5% charcoal-stripped serum (CSS) and treated for 1 day with 1 μM, 10 μM and 25 μM of compound. The UBE2C reporter screening assay was developed in house to monitor the levels of the AR-V7 isoform in 22Rv1 cells by using a plasmid containing the UBE2C promoter linked to a luciferase reporter.  2.2.3 Cell Viability Assay LNCaP Myc-positive were plated at 5000 cells per well in RPMI 1640 containing 5% CSS in a 96-well plate, treated with test compounds (0-25 μM) for 96 hours. Cell density was measured using the PrestoBlue assay according to the manufacturer’s protocol. The percentage of cell survival was normalized to the cell density of control wells treated by vehicle. Viability of Myc-negative HO15.19 cells was done similarly but in DMEM supplemented with 5% CSS. IncuCyte® technology was utilized to measure in real-time dose-dependent treatment effects on viability of LASCPC-01 NEPC cells, IMR32 neuroblastoma cells and HO15.19 Myc-null cells. PrestoBlue assay was also used to assess the effect on viability of 22Rv1 cells treated with hnRNP A1 compounds at increasing concentrations (up to 25 μM) for 3 days. High-sensitivity fluorescence was used as detection method according to the manufacturer’s protocol.  2.2.4 Protein Purification Histidine tagged Max (residues 23-102) and GST tagged Myc (residues 368-454) were overexpressed in E. coli BL21-DE3 cells. Cells were co-lyzed in lysis buffer (20 mM Tris pH 8, 124 500 mM NaCl, 5% glycerol, 10 mM imidazole, 8 mM BME, 2.1 mM PMSF). After sonication and centrifugation, the complex was first purified by using a Ni-NTA affinity resin. After overnight dialysis to remove the imidazole, the protein sample was applied to a size exclusion chromatography equilibrated with (20mM Tris pH 8, 150mM NaCl, 5% glycerol, 0.2mM TCEP). Fractions containing equal amount of Myc and Max on SDS PAGE were collected and used for the binding assay. The presence of both proteins was validated by Western blot using a specific antibody of each protein (Max (h2) Sc-8011 and c-Myc (9E10) Sc-40, Santa Cruz Biotechnology). hnRNP A1 was biotinylated in situ using an AviTagTM sequence (GLNDIFEAQKIEWHE) (Avidity, LLC, Aurora, CO, USA) incorporated at the N-terminus of the hnRNP A1. E. coli BL21 containing both biotin ligase and hnRNP A1 vectors were induced with 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) and the protein was expressed for 4 h at 25 °C in the presence of 125 μM biotin. The bacteria were then lysed by sonication, and the resulting lysate was purified by immobilized metal ion affinity chromatography (IMAC) with nickel–nitrilotriacetic acid (Ni–NTA) resin followed by size exclusion chromatography (superdex s75, GE Healthcare).  2.2.5 Bio-layer Interferometry (BLI) Assay The direct interaction between biotinylated E-box oligo (TGAAGCAGACCACGTGGTCGTCTTCA) immobilized on a streptavidin biosensor and a purified Myc-Max complex (0.05 mg/ml) was quantified by BLI using OctetRED (ForteBio). The DNA was first bound to the super-streptavidin sensors over 1000 sec at 25°C. The sensors were next moved into wells containing the reaction buffer (20mM Tris pH 8, 150mM NaCl, 5% glycerol, 0.2mM TCEP, 5 % dimethylsulfoxide) for measuring the baseline and next into the 125 Myc-Max complex alone or in presence of the tested Myc-Max inhibitors to study the association of the complex to the DNA. The direct interaction between purified UP1 RNA-binding domain of the hnRNP A1 protein (N-terminal residues 1-196) and compounds was also quantified with BLI technique using OctetRED (ForteBio). The purified hnRNP A1 at 0.1 mg/mL was immobilized on the streptavidin sensors (SA) overnight at 4 °C. The sensors were then blocked, washed and moved into wells containing various concentrations of the tested hnRNP A1 compounds in reaction buffer (20 mM Tris pH 8.5; 150mM NaCl; 0.5 mM TCEP; 5% DMSO and 10% glycerol).  2.2.6 Chromatin Fractionation Assay Sub-cellular fractionation of LNCaP cells was performed as previously described.324,325 Compounds were administered for 24 hours at various concentrations prior to biochemical fractionation of cytosol, soluble nuclear, and chromatin bound fractions. Western blotting for c-Myc detection was performed with c-Myc (9E10) Sc-40 specific antibody from Santa Cruz Biotechnology. Equal loading of samples and confirmation of successful sub-cellular fractionation was performed with a mAb against cytosol α-tubulin (Cell Signaling Technology, DM1A), and a mAb directed against nuclear histone H3 protein (Abcam, ab32356).  2.2.7 Western Blotting To assess the effect of Myc-Max inhibitors on apoptosis, after 48 hours of treatment with compounds, LNCaP cells were lysed, and protein sample preparation followed by Western blotting were performed. Blots were incubated with primary antibodies against c-Myc, PARP (Sigma 084M4766V), PARP cleaved-Asp214 (Sigma SAB4500487) and β-Actin (Sigma A2066) 126 overnight at 4°C, followed by appropriate peroxidase-conjugated secondary antibodies. β-actin served as an internal control.  To analyze the reduction of AR-V7 splice variant with hnRNP A1 inhibitors, Western blotting was performed using 22Rv1 cells starved in RMPI with 5% CSS media and treated with compounds for 24 h. The blot was incubated with rabbit anti-actin antibody (1:500 dilution) and mouse anti-AR-NTD 441 monoclonal antibody (1:500 dilution). Visualization of the immuno-complexes was done by an enhanced chemiluminescence system (Millipore, Burlington, MA, USA) followed by exposure to X-ray films.  2.2.8 qRT-PCR Quantitative RT-PCR was employed to quantify the mRNA levels of the AR-V7 splice variant in 22Rv1 cells upon treatment with hnRNP A1 compounds at 10 and 25 μM concentrations. 22Rv1 cells were starved in RPMI media supplemented with CSS for 48 h, before 24 h treatment with DMSO (0.1%) or compounds. RNA was extracted with TRIzol (Invitrogen, Carlsbad, CA, USA) followed by cDNA synthesis (SuperscriptII, Invitrogen, Carlsbad, CA, USA). RT-PCR (125 ng cDNA, 5 μM primers, Sybr green master mix) was performed on a ViiA 7 thermal cycler. Actin RNA was used as normalized control. qRT-PCR was also employed to assess the effect of Myc-Max inhibitors after 24 hours treatment on downregulation of target genes.  127 2.2.9 Metabolic Stability The metabolic stability (i.e. half-life) of selected compounds was assessed in vitro using M1000 pooled CD-1 male mouse liver microsomes (0.5 mL at 20 mg protein/mL) (XenoTech) and LC-MS (liquid chromatography mass spectrometry) analysis.   128 Chapter 3: Development of Novel c-Myc-Max Inhibitors using Methods of CADD 3.1 Background Drugging Myc is a difficult task due to the highly disordered nature of the protein, lack of well-defined pockets on its surface, and obligate dimerization with the Max protein for binding to DNA and transcriptional activation of target genes. Small molecule prototype inhibitors of Myc-Max developed to date that alter the transcriptional output of the complex demonstrate inadequate in vivo safety and efficacy thus restricting their clinical applicability. Therefore, there is an urgent need to discover novel and effective small molecule inhibitors of the Myc-Max complex.  Presented here is the application of CADD methods toward discovery and development of novel Myc-Max inhibitors with demonstrated ability to disrupt the functional mandatory Myc-Max binding to DNA and subsequently to block gene expression characteristic of Myc-Max-dependent tumors. The CADD techniques employed encompass in silico novel binding sites identification at the ordered Myc-Max heterodimer interface, and both structure-based methods such as large-scale docking-based virtual screening and molecular dynamics simulations as well as ligand-based techniques including pharmacophore modeling, chemical similarity searches and ADMET profiling, complemented by experimental validation.   3.2 Results 3.2.1 Discovery of c-Myc-Max Inhibitors  3.2.1.1 Binding Sites Identification  The published 1.9 Å X-ray structure of the c-Myc-Max heterodimer bound to its E-box DNA recognition sequence (PDB ID: 1NKP)123 was selected for identification of plausible 129 binding pockets at the structurally ordered interface of the dimer. Using the Site Finder module of the Molecular Operating Environment (MOE)303, three main sites (Figure 3.1) emerged as the most probable druggable sites according to their Propensity for Ligand Binding (PLB) scores.  The top PLB-ranked site was located at the fork formed between the DNA binding bHLH regions of c-Myc (hereafter Myc) and Max, with which the dimer inserts itself into the DNA major groove (site 1 in Figure 3.1). This site identified on the ordered dimer interface contains highly conserved residues from both Myc, Max monomers, and as such differs from all previously reported binding sites on Myc monomer only, both in its disordered and ordered forms. Hydrophobic residues Leu917 and Phe921 from the Myc monomer and the equivalent residues Ile218 and Phe222 from the homologous Max monomer, and charged residues Lys939 from Myc, and Arg212, Arg215, Asp216, Lys219 and Arg239 from Max shape this pocket. Mutagenesis studies as well as evidence from Max-Max and Omomyc homodimers crystallographic data on the specificity of DNA recognition through invariant amino acid-base interactions and contacts with the phosphate groups of the DNA backbone (described in section 1.2.6.4) highlight the functional relevance of residues in this site. Importantly, in contrast to the Myc-Max site, the corresponding Mad-Max site at the DNA interface – the only major predicted by Site Finder on the 1NLW X-ray structure, differs in amino acid composition, charge and size. The Mad-Max pocket is composed of Arg14, Leu17, Arg18, Leu21, and Thr39 from Mad (equivalent to Arg914, Leu917, Lys918, Phe921 and Lys939 on Myc), and Arg214, Arg215, Ile218, Lys219, Phe222 and Arg239 on Max. More residues contribute to this site on Mad compared to Myc, with two additional positively charged arginine residues in the site as well as two important substitutions, one from an aromatic to an aliphatic residue at position 921 in Myc 130 (Phe921Leu), and the other from a positively charged to a neutral amino acid at position 939 in Myc (Lys939Thr).  Therefore, designing specific inhibitors or stabilizers of the two complexes is feasible. Overall, the positively charged residues lining up the top-ranked predicted site on Myc-Max may provide hydrogen bonding and/or salt bridges propensity for potential binders to disrupt the above-mentioned DNA base and phosphate contacts while the hydrophobic core of the site may substantially anchor the binders via hydrophobic interactions. Our newly identified small molecule inhibitors (described hereafter) targeting this site block Myc-Max binding to DNA as hypothesized.  The second PLB-ranked site (site 2 in Figure 3.1) contains residues in the HLH region from the ordered Myc monomer only: Lys918, Phe921, Phe922, Arg925, Glu935, Lys936, Ala937, Pro938, Lys939 and Ile942. It neighbors the highly-positively charged first site but is more hydrophobic. Myc specific contacts in this site are Phe922, Lys939 and Pro938. In Mad, these residues are non-conservatively substituted for a glutamic acid, and two threonine residues, respectively.  The third site located at the intersection of H2 and LZ regions (site 3 in Figure 3.1) is the smallest and the most surface-exposed pocket. Residues Tyr949, Ser952, Val953, Glu956 and Leu960 from Myc shape it, along with Met253, Arg254, Asn257, His258 and Gln261 from Max. This is the site in which the most conservation anomalies exist with most Myc residues diverging from the consensus among bHLHLZ proteins, including Ser952 (consensus Lys), Val953 (consensus Ala), Glu956 (no consensus) and Leu960 (consensus Ala). While most specific, this is the least druggable site. 131  Figure 3.1 Three independent binding sites identified in silico at the Myc-Max/DNA and Myc-Max dimerization interfaces. (Top) The amino acid composition of the sites, numbered and circle-enclosed, mapped below the Myc and Max protein sequences. (Bottom) 3D representation of the sites (colored solid surfaces) at the Myc-Max interfaces. The sites were identified by probing the surface with virtual atoms (colored alpha spheres within the three found pockets).  3.2.1.2 In silico Identification of Hit Compounds  The highest PLB-ranked pocket at the DNA interface of Myc-Max was prioritized for targeting with small molecules (Figure 3.2).  132  Figure 3.2 In silico model of the basic/helix-loop-helix/leucine zipper (bHLHZ) domain of Myc-Max bound to the 5’-CACGTG-3’ DNA recognition sequence. The model was constructed based on the 1.9 Å crystal structure of Myc-Max heterodimer in its bound form to the canonical E-box sequence (PDB ID: 1NKP). Myc (blue), Max (red) and the DNA backbone (green) are represented as ribbons. The DNA bases forming the E-box recognition sequence are represented as sticks. The Myc-Max predicted binding site at the DNA interface (side view) is represented as a grey solid surface and emphasized with a black circle; (B) Magnified view of the Myc-Max predicted binding site (bottom view of the structure of the complex in panel A which is rotated 90° on the horizontal axis). The pharmacophore utilized for subsequent virtual screening and filtering of potential binders is shown as cyan meshed spheres. Inhibitors targeting the predicted site are expected to block Myc-Max binding to DNA.   The drug-like subset of the ZINC12 molecular database308,309, containing more than 6 million purchasable chemicals, further reduced to 4.7 million compounds by filtering on physicochemical properties such as charge, number of rings and rotatable bonds, was virtually screened against the desired pocket on the Myc-Max dimer DBD. Glide (Maestro 9.3 suite, 133 Schrӧdinger LLC) software220,221,305 was employed as the primary structure-based docking technique (with the standard precision mode). The generated docking poses were then filtered by the Glide docking score (binding energy score used to rank docking poses and distinguish strong binders in their optimal placement in the respective pocket from compounds that bind weakly) using a -5.5 kcal/mol cutoff.  The top ranked 12503 remaining compounds were further filtered by structure-based pharmacophore screening using MOE’s tools. A pharmacophore model of two essential hydrophobic features (1.5 Å diameter each) of the binding site (formed primarily by Leu917, Ile218, Phe921 and Phe222) was built and used to search for matching hits in the database of top ranked Glide poses (Figure 3.2). 1019 pharmacophore-matching hits were then selected for visual inspection. 116 compounds having a good balance of Glide docking score and ligand efficiency (the ratio of binding affinity over the number of heavy atoms) made additional side-chain or backbone hydrogen bonds with the charged residues in the site. 69 chemicals were selected for purchase, in particular those predicted to form hydrogen bonds with the backbone carbonyl oxygen of Arg215.  The purchased compounds were further subjected to experimental evaluation using a primary screening transcriptional assay (see below). From the primary cell-based screening, 10 hits were identified (Table 3.1) showing better than 50% inhibition of Myc-Max transcriptional activity at 25 μM. Hits with more than 70% inhibition were further investigated for effect on the AR-V7/UBE2C downstream pathway using an UBE2C reporter assay.    134  Table 3.1 Chemical structures and activities of initial hit compounds that bind the ordered Myc-Max DBD. VPC ID Structure Myc-Max transcriptional activity %inhibition  (25 μM) Myc-Max/AR-V7 pathway %inhibition (25 μM) IC50 (μM) 70005  65 n/a >25 70021  95 73 10 70027  53 n/a >25 70033  81 51 10 135 70053  73 50 11 70063  106 94 9 70064  78 64 >25 70066  65 n/a >25 70067  98 71 22 70068  73 58 >25 10058-F4*  91 70 29 136 10074-G5*  88 n/a 16 * Literature prototype Myc inhibitors used as controls.  Structurally the initial hit compounds fell in two main categories: rhodanine-containing compounds, such as VPC-70067 highly similar chemically to the known Myc prototype inhibitor 10058-F4, and compounds having a common thiourea or urea bioisoster linker with substituted aromatic rings at both ends, such as VPC-70063. The remaining hits, for instance VPC-70033, had dissimilar scaffolds.   3.2.1.3 In silico Binding Mode of VPC-70063  Compound VPC-70063 was the best performer in cell-based and cell-free assays designed for this study (see section 3.2.1.4). The predicted binding pose of VPC-70063 (1-benzyl-3-(3,5-bis(trifluoromethyl)phenyl)thiourea), obtained using computational modeling methods, is shown in Figure 3.3. The chemical structure of VPC-70063 is composed of a benzyl ring at one end, a thiourea linker and a highly hydrophobic 3,5-bis(trifluoromethyl)phenyl moiety at the other end. 137  Figure 3.3 (A) Predicted binding pose of VPC-70063 in space-filling representation within the Myc-Max DBD pocket (cyan = carbon, blue = nitrogen, green = fluorine, yellow = sulfur). Myc, Max, the DNA backbone and the pocket are colored using the same scheme as in figure 3.2. VPC-70063 chemical name: 1-benzyl-3-(3,5-bis(trifluoromethyl)phenyl)thiourea. The 3,5-bis(trifluoromethyl)phenyl moiety of VPC-70063 is deeply buried in the hydrophobic core of the pocket. The benzyl ring of VPC-70063 significantly overlaps with the DNA backbone. VPC-70063 is expected to compete with DNA for binding to the Myc-Max DBD pocket; (B) Pocket residues predicted to interact with VPC-70063 at the Myc-Max DNA interface. Hydrogen bonds formed by the thiourea moiety of VPC-70063 are indicated with magenta dashed lines. Hydrophobic interactions formed by the benzyl ring as well as the phenyl ring and the trifluoromethyl groups are represented with green dashed lines. Protein residue numbering follows that from the crystal structure of Myc-Max dimer (PDB ID: 1NKP; chains A, B).  138 Within the binding pocket of the Myc-Max DBD domain, VPC-70063 is predicted to form 2 hydrogen bonds between the 2 thiourea amine hydrogens and the backbone carbonyl of Arg215 (Figure 3.3B, magenta dashed lines). In addition, VPC-70063 forms a large number of strong hydrophobic interactions (Figure 3.3B, green dashed lines). On one hand, the 3,5-bis(trifluoromethyl)phenyl moiety strongly interacts with the hydrophobic core of the pocket, including the aliphatic and aromatic side-chains of Leu917, Phe921, and Lys939 of Myc, and Ile218, Phe222, and Arg215 of Max. On the other hand, the benzyl ring interacts with the aliphatic side chains of Arg215 and Arg212 of Max. Noteworthy, as shown in Figure 3.3A, the 3,5-bis(trifluoromethyl)phenyl group of VPC-70063 is matching the hydrophobic features of the constructed pharmacophore with an RMSD of 0.92 Å, it is deeply buried in the hydrophobic core of the Myc-Max DBD pocket being stabilized via hydrophobic interactions. Furthermore, in the binding pose, the benzyl ring of VPC-70063 overlaps significantly with the DNA backbone.  The docked structure was further minimized in implicit solvent and the binding free energy was calculated with the MM/GBSA method (molecular mechanics with generalized Born and surface area continuum solvent model). The estimated binding free energy of VPC-80051 obtained from MM/GBSA simulations was −40.2 kcal/mol. In these simulations in which the binding site residues were allowed to relax (i.e., protein flexibility within 5.0 Å from the ligand) during the minimization procedure of the complex, VPC-70063 maintained the protein-ligand interactions predicted by rigid docking (Figure 3.4). Therefore, it was expected that VPC-70063, in as much as other hits having similar interactions, would overcome the binding of DNA to the Myc-Max DBD site. It is not surprising then, that VPC-70063 blocks the binding of Myc-Max to DNA as determined by BLI and chromatin fractionation assays (see section 3.2.1.4). 139  Figure 3.4 Myc-Max/VPC-70063 interactions obtained from MM/GBSA simulations with implicit solvent in two-dimensional representation. Purple arrow pointed lines represent backbone hydrogen-bonding interactions, green line indicates π interactions, and the curved contour represents the shape of the pocket around the ligand colored by the character of the interacting protein residues: (green = hydrophobic, blue = positively charged, red = negatively charged).  3.2.1.4 In vitro Characterization of Hit Compounds  Effects of hit compounds on Myc-Max transcriptional activity  The selected chemicals were subjected to experimental evaluation using a commercially available transcriptional assay Cignal (c-Myc luciferase reporter assay in LNCaP cells). Compounds 10058-F4 and 10074-G5, known Myc inhibitors from the literature, were used as positive controls. A transiently transfected Myc-driven luciferase reporter allowed the monitoring of Myc-regulated signal in LNCaP upon treatment with the in silico identified compounds. From a larger number of hits, 10 compounds caused more than 50% reduction of the Myc-driven luciferase levels at 25 μM (see Table 3.1). A thorough dose response analysis was 140 performed using LNCaP cells to evaluate the potency of hit compounds. The compounds inhibit Myc-Max transcriptional activity with low to mid-micromolar potency, with the following IC50 values (half-maximal inhibitory concentration with 95% Confidence Intervals) established as: 22.7 μM [16.6 to 31.2 μM] for VPC-70067 comparable to that of the control compound 10058-F4 (28.9 μM; [19.7 to 42.5 μM]), and, importantly, 8.9 μM [6.6 to 11.8 μM] for VPC-70063 (Figure 3.5A). Effects of hit compounds on Myc-Max downstream-regulated pathways  Myc inhibition was recently reported to reduce levels of the constitutively active androgen receptor splice variant AR-V7 in 22Rv1 cells.67 AR-V7 has been shown to specifically regulate the expression level of the Ubiquitin Conjugating Enzyme E2C (UBE2C) in androgen-deprived 22Rv1, through the UBE2C promoter.68 Hence, a complementary transcriptional screening assay was developed in house to monitor the expression levels of the AR-V7 isoform in 22Rv1 cells by using a plasmid containing a UBE2C promoter linked to a luciferase reporter. The dose-dependent reduction of luciferase levels by the identified hits indicates a Myc-related reduction of AR-V7 level in the cells (see Table 3.1). Compound VPC-70063 showed the highest reduction of UBE2C promoter activity suggesting the reduction of AR-V7 levels in 22Rv1. The AR-V7 reduction with the hit was confirmed by Western blot (Figure 3.5B). Effects of hit compounds on cell viability  The effect of hit compounds on Myc-driven cell proliferation was evaluated by measuring the cell viability of LNCaP cells after treatment with increasing concentrations of compounds. VPC-70063 showed the best inhibition of LNCaP cell proliferation (IC50 = 2.5 μM; [95%CI: 2.1-2.8 μM].  To rule out that this inhibition was due to non-specific cytotoxicity of VPC-70063, the c-Myc knockout HO15.19 cell line (Figure 3.5C) was treated with the compound. The 141 proliferation of the HO15.19 cell line was slightly affected by VPC-70063, up to a maximum of 36% inhibition at 25 μM. However, at a VPC-70063 concentration of 3 μM where 70% of LNCaP cells are inhibited there was no significant effect on the c-Myc knockout cells. VPC-70067, 10058-F4 and 10074-G5 have IC50 values of 11.1 μM [95%CI: 10.6-11.4 μM], 18.31 μM [95%CI: 17.7-18.8 μM] and 8.7 μM [95%CI: 8.3-9.1 μM], respectively (Figure 3.5C).  Figure 3.5 Effects of VPC-70063 on Myc-Max transcriptional activity, AR-V7 levels and viability of selected cell lines. (A) Dose response effect in LNCaP PCa cells on the transcriptional activity of c-Myc by using a c-Myc mediated luciferase reporter. Data are presented as mean ± SEM of triplicates and expressed as a percentage of luciferase activity relative to DMSO control. (B) Inhibition of Myc-Max reduces the levels of AR-V7 in 22Rv1 cells. (C) The effect of VPC-70063 in comparison with 10074-G5 on cell viability of Myc positive (LNCaP) and Myc negative (HO15.19) cell lines. The percent of cell viability is plotted in dose dependent manner. Data points represent the mean ± 95%CI (confidence interval) of triplicates and expressed as percent of cell viability relative to DMSO control.  142 Mechanism of action of the identified Myc-Max inhibitors  Apoptosis Myc inhibition induces cell death by cell cycle arrest and apoptosis.326 Cleavage of PARP-1 by caspases is considered a hallmark of apoptosis. The ability of compounds to induce PARP cleavage after treatment was thus measured. As predicted, VPC-70063 induced PARP cleavage suggesting that the observed effect was through apoptotic pathways (Figure 3.6).  Figure 3.6 Inhibition of Myc-Max with VPC-70063 induces apoptosis of LNCaP cells as indicated by cleavage of PARP in Western blot.  Direct binding and disruption of protein-DNA interactions To study the direct effect of hit compounds on the interaction between Myc-Max heterodimer and DNA, Bio-Layer Interferometry (BLI, ForteBio) technique was employed. BLI is a label-free technology allowing the measurement of direct interactions between two partners, one immobilized on a sensor and the other one present in a solution. BLI was thus applied to quantify the disruption of the interaction between a biotinylated E-box oligo immobilized on a streptavidin biosensor and a purified Myc-Max complex in presence of compounds. Therefore, histidine-tagged Max (residues 23-102) and GST tagged Myc (residues 368-454) were overexpressed and co-purified. The fraction containing equal amounts of Myc and Max was collected and used for the binding assay (Figure 3.7A). The presence of both proteins was validated by Western blot using a specific antibody for each protein. As determined by BLI, the 143 Myc-Max heterodimer was prevented from interacting with the immobilized 5’-CACGTG-3’ DNA in presence of VPC-70063 (and VPC-70067) similarly to the control 10074-G5 (Figure 3.7B). Additionally, the ability of the best compound VPC-70063 to disrupt the interaction of Myc-Max with DNA was assessed in a dose-dependent manner. At concentrations below 100 μM, there was no significant effect on the complex dissociation. At higher concentrations, though, a dose response decrease of Myc-Max binding to DNA was observed, showing that VPC-70063 was able to disrupt the complex formation (Figure 3.7C).  Figure 3.7 BLI quantification of disruption of Myc-Max interactions with DNA upon treatment with VPC-70063. (A) Purification of GST-Myc and His-Max using size exclusion chromatography. The fraction highlighted with a black rectangle on the Western blot corresponds to the fraction used for the binding assay. (B) Inhibition of Myc-Max interaction with the biotinylated E-box quantified by BLI in presence of 500 μM of studied compounds. (C) Dose response inhibition of Myc-Max binding to DNA in presence of VPC-70063.  Since Myc is a nuclear transcription factor, the effect of VPC-70063 on the disruption of Myc-Max binding to DNA was further investigated using a chromatin fractionation assay (see section 2.2.6 for experimental details). As expected, Myc was not present in the cytosol but in both the soluble and chromatin bound nuclear fractions (Figure 3.8). Importantly, treatment with compound VPC-70063 resulted in strong dissociation of Myc-bound fraction from DNA 144 (supporting the BLI data) and a significant reduction of Myc protein levels in a dose-dependent manner at low concentrations in the 2 to 10 μM range.  Figure 3.8 VPC-70063 strongly dissociates Myc from chromatin and significantly reduces Myc protein levels in a dose-dependent manner at single-digit μM range. Tubulin and histone H3 are cytosolic and chromatin-binding protein biomarkers, respectively, used as controls.  Microsomal stability Metabolic stability is an important property of drug candidates as poor metabolism is one of the major causes of drug failure during development.297 An important measurable value to quantify metabolic stability of compounds is their half-life time (t1/2) which can be determined in liver microsomes that contain a large number of the phase I (i.e. CYP) and phase II (i.e. UGT) drug-metabolizing enzymes. As such, the metabolic stability of VPC-70063 as well as control 10074-G5 was assessed in vitro using pooled CD-1 male mouse liver microsomes. The half-live of VPC-70063 was 69 min relative to only 3 min for 10074-G5 literature control known to undergo rapid metabolism (Figure 3.9). Thus, VPC-70063 demonstrated moderate yet encouraging metabolic stability. 145  Figure 3.9 Fraction of original concentration of compounds 10074-G5 (left) and VPC-70063 (right) versus time in mouse liver microsomes. Imipramine (Imip) was used as control.  To summarize this section, a small number of compounds were identified that inhibited Myc-Max activity with low to mid-micromolar potency and with minimal generic cytotoxicity. VPC-70067, a close analog of the previously identified Myc inhibitor 10058-F4, highly similar in structure, potency and mechanism of action, served as proof-of-concept that the in silico drug discovery strategy performed as expected.  Compound VPC-70063, of a chemically different scaffold, was the best performer in a panel of in vitro assays, and the forerunner for hit-to-lead optimization efforts.327   3.2.2 Optimization of c-Myc-Max Inhibitors  The findings reported in section 3.2.1 prompted us to leverage the full power of our ever-growing CADD platform (Figure A.1) that proved successful in targeting unconventional sites on protein surfaces and in yielding promising preclinical drug candidates for previously 146 uncharted targets19,328-335 to enhance the target affinity and to eliminate the cytotoxicity observed at higher concentrations upon treatment with initial hits.  Consequently, ligand-based similarity searches followed by molecular docking and consensus scoring computations were performed to identify analogs of the initial hit compounds. Three-dimensional similarity searches were conducted utilizing the ROCS program from OpenEye289,316 against a large ensemble of conformers consisting of at most 200 conformers for each of the approximately 9 million entries of the drug-like purchasable chemical space of the ZINC15 database.317 Conformers were generated using Omega2 of OpenEye.313  Current hits served as query molecules. In addition, substructure-based searches against ZINC15 as well as ENAMINE-REAL, the largest catalog of readily synthesizable chemicals were performed. Molecular docking, using various docking programs differing in their underlying scoring functions and sampling algorithms: Glide220,221, eHiTS310, ICM263 and Hybrid264,311, was employed to position analogs into the Myc-Max DBD. Consensus voting and filtering using various thresholds as described in section 2.1.6 followed. To eliminate toxic moieties and metabolically labile centers, the drug-like profile of derivatives was assessed using Simulations Plus ADMET Predictor software.  For more accurate scoring, relative free energies of binding were estimated using the MM/GBSA minimization of ligated structures in continuum solvent (Figure 3.10). Computationally expensive molecular dynamics simulations were also carried out for selected compounds to gain further insights into induced fit motions and solvent effects, and thus, the dynamic behavior of protein-ligand complexes in explicit solvent. All high-confidence analogs were subsequently subjected to full experimental profiling. 147  Figure 3.10 In silico drug discovery pipeline.  3.2.2.1 VPC-70063 Derivatives  Application of combined structure- and ligand-based CADD techniques led to identification and consequent ordering of 142 analogs of VPC-70063 out of which 54 were active toward inhibiting Myc-Max transcriptional activity in LNCaP cells. With a resulting success rate of 38%, the in silico hit-to-lead optimization protocol yielded more potent derivatives of the parental compound that maintained the hydrophobic core and H-bonding interactions and displayed enhanced inhibitory effects on Myc-Max transcriptional activity and viability of LNCaP Myc-dependent cell line, and reduced toxicity in Myc-independent HO15.19 cell model (Table 3.2). For instance, compounds VPC-70495, 70532 and 70514 with improved IC50 values as low as 1 μM, reduced toxicities to as low as 10% against the viability of HO15.19 148 cell line at 12 μM, and enhanced viability inhibition of LNCaP cells as high as 73% at 12 μM, received more favorable or similar in silico scores relative to the parental compound. The corresponding Prime MM/GBSA free energy estimates were -42.5, -42.3 and -40.2 kcal/mol, respectively, relative to -40.2 kcal/mol of VPC-70063, and matched the pharmacophore with RMSD values of 0.88, 0.89, and 0.93 Å, relative to 0.92 Å of the parental compound.  Table 3.2 Chemical structures and activities of best analogs of VPC-70063. VPC ID Structure Myc-Max Transcriptional activity LNCaP cells %inhibition (10 μM) IC50 (μM) Viability HO15.19 cells %inhibition (12 μM) Viability LNCaP cells %inhibition (12 μM) 70495  80 5 25 50 70498  106 10 0 29 70530  94 6 10 56 149 70532  101 6 10 73 70535  95 5 10 55 70511  112 2 27 75 70514  110 1 27 70   3.2.2.2 Development of Lead Compound VPC-70551 Of a distinct chemical class from that of VPC-70063, VPC-70033 (N-(4-[[5-chloro-2-nitro-4-(trifluoromethyl)phenyl]amino]phenyl)acetamide) was second best among the initial hits with an IC50 of 10 μM in inhibiting Myc-Max transcriptional activity (Table 3.1). Interestingly, VPC-70033 bore, among others, the electron-rich nitro moiety found on literature Myc inhibitor 10074-G5 as well as its improved congeners, which was critical for binding to the Myc monomer at the DNA interface. Through its nitro group, VPC-70033 engaged the side chain of Arg913 of Myc via H-bonding, in addition to forming an H-bond with the side chain of Arg215 of Max via the carbonyl oxygen of its acetamide moiety. Moreover, it matched the pharmacophore of the hydrophobic core of the Myc-Max binding site with an RMSD of 1.04 Å.  The Glide docking score of VPC-70033 was -5.66 kcal/mol. Poses obtained from ICM and Hybrid docking programs showed little deviation from the Glide pose, with calculated RMSD 150 values of 0.71 Å and 0.77 Å, respectively. The relative free energy of binding obtained from flexible MM/GBSA in implicit solvent was -41.7 kcal/mol in the same range as that of hit VPC-70063 and its best derivatives. In MM/GBSA simulation, VPC-70033 maintained the interactions with the pocket residues obtained from rigid docking, while forming additional salt bridges with both Arg239 of Max and Arg913 of Myc via its nitro moiety (Figure 3.11).  The predicted ADMET profile of VPC-70033 suggested risks associated with absorption and distribution (i.e. aqueous solubility, lipophilicity, and fraction unbound) and hepatic toxicity, with an overall score (i.e. ADMET Risk) of 4.2, below the cut-off value of 6 (out of 20).   Figure 3.11 Myc-Max/VPC-70033 interactions obtained from MM/GBSA simulations with implicit solvent. Purple arrow pointed dashed lines represent side chain hydrogen-bonding interactions, red to blue colored lines represent salt bridges and the curved contour represents the shape of the pocket around the ligand colored by the character of the interacting protein residues: green = hydrophobic, blue = positively charged, red = negatively charged.  A first round of in silico ligand-based ROCS chemical similarity searches with VPC-70033 as query molecule yielded compound VPC-70127 (N'-[2-nitro-4-151 (trifluoromethyl)phenyl]pyrazine-2-carbohydrazide), a distant analog of VPC-70033 (Figure 3.12). Among 5000 analogs resulting from the ROCS search, VPC-70127 rank was 1861 in the matching hitlist with a TanimotoCombo score of 1.22 on a scale from 0 to 2 reflecting the lower similarity with the parental compound.  As indicated by the ShapeTanimoto and ColorTanimoto individual ROCS scores with values 0.86 and 0.36, respectively, on a scale from 0 to 1, the two compounds were similar in three-dimensional shape but chemically more dissimilar as they differed in the arrangement and number of hydrogen-bond donors, hydrogen-bond acceptors, and rings as well as chemical cores. While the scaffold of VPC-70033 was diphenylamine, that of VPC-70127 consisted of terminal phenyl and pyrazine rings connected via a carbohydrazide linker (i.e. N'-phenylpyrazine-2-carbohydrazide). Both compounds shared a phenylamine (aniline) moiety with nitro and trifluoromethyl substituents in ortho- and para- positions.   Figure 3.12 VPC-70127 a novel scaffold obtained through ROCS chemical similarity searches using VPC-70033 as a template.  The selection of VPC-70127 derivative for purchase and subsequent experimental evaluation was driven by very good agreement in the results obtained from various in silico techniques of the CADD platform. Thus, in the consensus-based rigid docking, the Glide pose of VPC-70127 had a score of -5.56 kcal/mol and the poses obtained from eHiTS and OpenEye 152 Hybrid programs deviated slightly from the Glide reference with calculated RMSD of 2.83 and 1.45 Å, respectively, all below the set filtering thresholds. In its predicted docking pose within the Myc-Max DBD binding site (Figure 3.13), VPC-70127 strongly interacts with the hydrophobic core of the site via its trifluoromethyl-phenyl ring, its pyrazine ring significantly overlaps with the DNA backbone and in addition forms H-bonds with Arg215 of Max. The nitro group forms H-bonds with Arg913 of Myc. Importantly, the hydrazide linker forms a very strong H-bond via its carbonyl oxygen with Arg239 of Max, mutation of which as previously described abolishes binding to DNA.  Figure 3.13 (A) Docking pose of VPC-70127 in space-filling representation within the Myc-Max DBD pocket. (B) Myc-Max/VPC-70127 interactions within the DBD site. Hydrophobic interactions are represented with green dashed lines. The strong hydrogen bond formed by the carbonyl oxygen of hydrazide linker of VPC-70127 novel scaffold with the side chain of Arg239 is indicated with magenta dashed line.  153 Optimization of the docked structure in implicit solvent perfomed using the MM/GBSA method revealed that during the minimization procedure of the complex in which the binding site residues were allowed to relax (i.e., protein flexibility within 5.0 Å from the ligand), VPC-70127 maintained the protein-ligand interactions predicted by rigid docking (Figure 3.14).   Figure 3.14 Myc-Max/VPC-70127 interactions obtained from MM/GBSA simulations with continuum solvent. Purple arrow pointed dashed lines represent side chain hydrogen-bonding interactions, red to blue colored lines represent salt bridges, green lines indicate π interactions, and the curved contour represents the shape of the pocket around the ligand colored by the character of the interacting protein residues.  Thus, Myc-Max residues of the hydrophobic core, Leu917, Phe921 of Myc, and Ile218, Phe222 of Max interacted favorably with the trifluoromethyl-phenyl moiety of 70127 via hydrophobic interactions. The side chain of Arg239 formed H-bonds and salt bridges with the carbonyl oxygen of the hydrazide linker and the nitro group, respectively, as well as hydrophobic interactions with the phenyl ring. The side chain of Arg215 formed hydrophobic contacts as well as an H-bond with one nitrogen of the pyrazine ring. The side chain of Arg913 formed an 154 additional H-bond as well as salt bridges with the nitro group of VPC-70127. The estimated binding free energy of VPC-70127 obtained from MM/GBSA simulations was -36.8 kcal/mol comparable to that of the distant parental compound, VPC-70033. Computationally intensive molecular dynamics (MD) simulations were subsequently carried out to gain further insights into induced fit motions and solvent effects, and thus, the dynamic behavior of the Myc-Max/VPC-70127 complex when fully subjected to force field in explicit solvent as opposed to the MM/GBSA implicit solvent approach. Analysis of the trajectory obtained from MD simulations showed that VPC-70127 favorably interacted with the DBD pocket for more than 30% of the 120 ns simulation time. Ligand binding was driven mainly by hydrogen-bonding interactions between the carbonyl oxygen of the hydrazide linker with Arg239 side chain and cation-π interactions between the phenyl ring with Arg239 side chain that occurred for more than 90% of the simulation time (Figure 3.15). In addition, further contacts occurred for 47% of the simulation time through salt bridges between Arg239 side chain with both the nitro group and the carbonyl oxygen of the linker. Seconding Arg239, Phe222 formed hydrophobic interactions with the trifluoromethyl-phenyl ring. Additionally, throughout the simulations, Ile218 was the second major contributor to hydrophobic interactions, while Arg215 and Asp216 provided additional hydrogen bonding interactions. The ADMET profiling of VPC-70127 utilizing the ADMET Predictor software showed a low overall ADMET risk of 2.9 relative to 4.2 of the parental compound VPC-70033. Unlike the distant parental compound, VPC-70127 had no predicted absorption and distribution problems and, importantly a low metabolic risk of 0.93 with only 3 predicted metabolites resulting from oxidative processes around the pyrazine ring due to the action of CYP3A4 (79%) and CYP1A2 (21%) liver metabolizing enzymes (Figure 3.16). 155  Figure 3.15 Dynamics of Myc-Max/VPC-70127 complex. VPC-70127 interacts favorably with Myc-Max DBD residues in 120-ns explicit solvent MD simulations without restraints.   156  Figure 3.16 Predicted metabolism of VPC-70127.  Further experimental validation of VPC-70127 demonstrated best inhibition of Myc-Max transcriptional activity in LNCaP cells with an IC50 of 1 μM (Figure 3.17A) and >100% inhibition of Myc/AR-V7-driven UBE2C activity. The compound had a very strong effect on apoptosis via PARP pathway programmed cell death as indicated by PARP cleavage in Western blot (Figure 3.17B). Moreover, it effectively inhibited the growth of LNCaP cells, with only 19% cytotoxicity at 12 μM in the HO15.19 Myc-/- knockout cell line (Figure A.2). 157  Figure 3.17 Strong effect of VPC-70127 on Myc-Max transcriptional activity, IC50 = 1 μM (A) and apoptosis (B) in LNCaP cells.  The consistent behavior of VPC-70127 obtained from in silico techniques, good ADMET profile and, importantly, its potent activity and specificity in Myc-driven cell models prompted a second round of in silico medicinal chemistry. As such, chemical similarity searches both ROCS-based (against the drug-like chemical space of ZINC15) and substructure-based, with and without the reactive nitro group of VPC-70127 as templates, against Enamine-REAL repository were conducted to identify potentially improved VPC-70127 derivatives. Following the docking protocol and further filtering (of best 2000 scoring compounds) based on satisfaction of the hydrophobic requirements of the pharmacophore constructed for the Myc-Max site, 24 analogs of VPC-70127 were selected for purchase (along with the 142 analogs of VPC-70063). Out of 24, 11 compounds showed >50% inhibition of Myc-Max transcriptional activity in LNCaP cells at single lower concentration of 12.5 μM (and 3 more at 25 μM), an excellent 58% success rate in the hit-to-lead optimization process (Table 3.3).     158  Table 3.3 Activities of best on-demand synthesizable ENAMINE-REAL derivatives of VPC-70127. VPC ID Myc-Max Transcriptional activity %inhibition (12.5 μM) RMSD Pharmacophore (Å) Glide docking score* (kcal/mol) MM/GBSA ∆Gbind (kcal/mol) 70551 98 0.60 -4.8 -33.1 70549 75 0.60 -4.9 -33.9 70554 71 0.51 -4.9 -32.6 70545 66 0.98 -5.0 -35.3 70547 66 0.71 -4.5 -31.5 70546 58 1.23 -4.8 -34.6 70548 55 0.42 -5.5 -35.0 70550 51 0.97 -4.8 -38.6 * OPLS3 FF  Three of the active compounds (VPC-70551, VPC-70545 and VPC-70546) resulting from substructure searches without the reactive nitro group of the parental compound VPC-70127, retained the hydrazide linker of 70127, connecting substituted phenyl rings at both ends. Importantly, these three compounds were selected from a database of ~600 hydrazide linker-containing Enamine-REAL on-demand synthesizable analogs based on satisfaction of the DBD site pharmacophore. The activities of the three derivatives, 98%, 68% and 58% of Myc-Max driven transcription, respectively, positively correlated with the pharmacophore RMSD values of 0.6, 0.98 and 1.23 Å, respectively. The lower the RMSD, the higher the activity (Table 3.3 and Figure 3.18). 159  Figure 3.18 Chemical structure of three active congeners of VPC-70127 that retain the hydrazide linker attaching substituted phenyl rings.   The most potent among the three was VPC-70551 (N'-[4-cyano-2 (trifluoromethyl)phenyl]-4-(trifluoromethyl)benzohydrazide). The chemical structure of VPC-70551 consisted of a cyano-trifluoromethyl-phenyl ring at one end, the hydrazide linker and a trifluorophenyl ring at the other end. A comparison between the chemical structures of the three derivatives suggests that the cyano group of VPC-70551 is the substituent responsible for the observed higher potency of VPC-70551 in inhibiting Myc-Max transcription. A dose response analysis on the effect of VPC-70551 on Myc-Max transcriptional activity in LNCaP cells indicated an IC50 = 4 μM comparable with that of the parental compound. The specificity of the compound was further assessed by real-time monitoring for its effects on the viability of the LNCaP Myc positive and HO15.19 Myc null cell line. While strongly affecting the growth of LNCaP cells, it had minimal effect on the HO15.19 cell line, with only 10% 160 cytotoxicity at a 10 μM concentration likely due to the removal of the reactive nitro group of the parental compound (Figure 3.19).   Figure 3.19 (A) Effect of VPC-70551 on Myc-Max transcriptional activity in LNCaP cells, IC50 = 4 μM. (B). Minimal effect on viability of HO15.19 Myc null cell line.  Moreover, the replacement of the pyrazine ring of parental VPC-70127 having three predicted metabolites with the trifluoromethyl-phenyl ring of 70551 yielded no predicted metabolites (Figure 3.20).   Figure 3.20 VPC-70551 improved chemical scaffold. Lack of the nitro group (encircled in red) made VPC-70551 less toxic relative to VPC-70127, while replacement of pyrazine for trifluoromethyl-phenyl (encircled in green) improved the metabolic stability of VPC-70551.  Notably, the in vitro evaluation of VPC-70551 microsomal stability demonstrated a very good half-life t1/2 = 140 min (Figure 3.21). 161  Figure 3.21 Fraction of original concentration of lead compound VPC-70551 versus time in mouse liver microsomes (t1/2 = 140 min). Imipramine (Imip) was used as control.  VPC-70551 is the current lead candidate presently under optimization efforts that involve additional iterative rounds of medicinal chemistry, in silico modeling and biological validation based on observed structure-activity relationships (SAR). To help guide further development, a thorough analysis of the structural determinants of binding affinity and thus activity of VPC-70551 was conducted.  The docking pose obtained from Glide SP agreed with the conformations predicted by ICM and OpenEye Hybrid programs with calculated RMSD of 1.5 and 3.3 Å, respectively. In the docked conformation, the cyano-trifluoromethyl-phenyl group of VPC-70551 was the main contributor to the binding affinity.  On one hand, the cyano functional group engaged the side chains of Lys939 of Myc and Arg214 of Max via H-bonding, mimicking the H-bonding interactions that these residues form with phosphates of the DNA backbone. On the other hand, the ortho-trifluoromethyl-phenyl ring buried the ligand into the core of the Myc-Max binding site via hydrophobic interactions with the two conserved phenylalanine residues, Phe921 and Phe222, as well as Leu917 and Ile218 of the 162 two monomers. The second para-trifluoromethyl-phenyl ring made additional hydrophobic interactions with the aliphatic side chain of Arg215, and in the predicted orientation, it overlapped with the DNA backbone (Figure 3.22). The estimated binding free energy of the Myc-Max/VPC-70551 complex obtained from the flexible MM/GBSA model in implicit solvent was -33.1 kcal/mol. In this minimized conformation, similar protein-ligand interactions were observed, in particular the H-bond between the cyano moiety with Arg214 side chain. In addition, the carbonyl oxygen of the hydrazide linker formed an H-bond with the side chain of Arg215, thus tightening binding of the ligand (Figure 3.23).  Figure 3.22 (A) Docking pose of VPC-70551 in space-filling representation within the Myc-Max DBD pocket. (B) Rigid Myc-Max/VPC-70551 interactions within the DBD site. Hydrophobic interactions are represented with green dashed lines. Hydrogen bonds formed by the cyano group with Lys939 and Arg214 are indicated with yellow dashed lines.  163  Figure 3.23 Myc-Max/VPC-70551 interactions obtained from MM/GBSA simulations in implicit solvent. Purple arrow pointed dashed lines represent side chain hydrogen-bonding interactions, green lines indicate π interactions, and the curved contour represents the shape of the pocket around the ligand colored by the character of the interacting protein residues.   Compared to the implicit solvent MM/GBSA approach, MD simulations more accurately estimate the thermodynamics and kinetics of ligand-target recognition and binding by explicitly treating structural flexibility, enthalpic (i.e. strength of protein-ligand interactions) and entropic effects (i.e. that of water as physiological solvent). Analysis of the trajectory obtained from MD simulations of the optimized Myc-Max/VPC-70551 complex with restraints imposed on alpha carbons of the dimer backbone provided further insights into the binding specificity and critical protein-ligand interactions (Figure 3.24).  164  Figure 3.24 Dynamics of Myc-Max/VPC-70551 complex. VPC-70551 interacts favorably with Myc-Max DBD residues in 120-ns explicit solvent MD simulations.  Throughout 120 ns simulation time, VPC-70551 made preferential contacts with pocket residues, primarily salt bridges between the cyano group and Arg214 side chain as well as salt 165 and water bridges between the hydrazide linker and Arg239 for more than 30% of simulation time, and hydrophobic interactions with Ile218. Additional cation-π interactions were formed between the side chains of Arg215 of Max and equivalent nearby Arg914 of Myc with the two phenyl rings of the ligand.  Superimposition of conformations of the simulated complex taken at reference (0 ns) and selected end time step (116 ns) revealed that the side chains of Arg215, Arg212 and Arg914 underwent larger fluctuations relative to other side chains in the pocket (Figure 3.24). The side chain of Arg215 moves closer to the ligand enabling cation-π interactions. Notably, little changes in the positions of the hydrophobic residues of the pocket and the two arginine residues interacting with the ligand, Arg239 and Arg214, were observed. The binding of the ligand and the reduced flexibility of the main interacting residues can be explained by the extended H-bonding network between the ligand, Arg214 side chain further interacting with Glu211, as well as tethering of Arg239 side chain against the aromatic ring of Phe222. As previously described (section 1.2.6.4), mutations of Arg239, Arg215 and Glu211 abolished DNA binding highlighting the importance and specificity of these residues for ligand binding. Overall, the data suggests that VPC-70551 is able to compete with DNA by disrupting critical Myc-Max/DNA recognition points. As illustrated in Figure 3.25, VPC-70551 is strongly anchored in the hydrophobic core of the site matching both designed pharmacophore features with its ortho-trifluoromethyl-phenyl group. Tight binding is further achieved via the cyano substituent due to H-bonding interactions with Arg214 (and Lys939 in docking conformations) that replace the H-bonding contacts of these side chains with the phosphate group of adenine DA309 at second position of the E-box 5’-CACGTG-3’ strand bound by Myc. Moreover, the carbonyl oxygen of the hydrazide linker sterically clashes (2.5 Å distance) with oxygen OP2 of 166 the phosphate group of adenine DA109 at second position of the E-box strand bound by Max that forms H-bonds with the side chain of critical Arg239.   Figure 3.25 VPC-70551 competes with DNA by disrupting critical Myc-Max/DNA interactions. H-bonds are indicated with yellow dashed lines. Pharmacophoric features are shown as cyan meshed spheres. Clashed with the DNA backbone are indicated by orange disks.  167 Lastly, in the model, the second para-trifluoromethyl-phenyl group significantly clashes with the DNA backbone of adenine DA109 and cytosine DC110 bases in second and third position of Max-bound E-box. Specifically, the benzene ring clashes with sugar-phosphate backbone of adenine DA109 at as small distances as 0.9 and 1.2 Å between the ring carbons and O5’ and OP2 oxygens of DA109, respectively, thus interfering with the H-bonding interactions of DA109 with Arg239. Moreover, the trifluoromethyl substituent strains the DA109 sugar (e.g. C3’ carbon at a distance of 1.9 Å) and collides with the OP2 oxygen of DC110 phosphate at a 1.5 Å distance, thus disrupting the phosphate H-bonding interactions of DC110 with critical Arg215.  The in silico protein-ligand contacts analysis of VPC-70551 designated a novel active scaffold: N'-[4-cyano-2-(trifluoromethyl)phenyl]benzohydrazide for further SAR-based development.   3.3 Discussion Transcription factors regulate gene expression by interacting with specific nucleic acid recognition sites. Sequence-specific protein-DNA binding is, as such, a central mechanism underlying transcriptional regulation of the cell.  Myc is a major oncogenic transcription factor often deregulated and implicated in most if not all human cancers. The potent transforming activity of Myc represents a highly valued target for therapeutic intervention yet the selective disruption of its protein-protein or protein-DNA interactions is difficult to address with small molecule drugs.  In the absence of clinically approved anti-Myc drugs, targeting the Myc-Max complex represents a critical step towards creating new therapeutics for lethal CRPC. The majority of 168 literature reported small molecule prototype inhibitors that bind and alter the disordered Myc monomer and subsequently block Myc-Max dimerization, as well as the few Myc-Max/DNA disruptors emerged mainly from high-throughput screens (HTS) of limited chemical libraries unlikely to contain clinically optimized structures with only three Myc inhibitors discovered through computational approaches. While these small molecule inhibitors affected growth in cancer-specific cell lines with few showing effectiveness in animal models, they generally demonstrated suboptimal in vivo safety and efficacy profiles. Thus, successful targeting the oncogenic transformation of the Myc-Max complex with small molecules, although considered high-risk, represents the high-reward opportunity long awaited to advance prostate cancer treatment. Application of computer-aided drug discovery methods, including binding sites identification, large-scale virtual screening, hit identification and lead optimization techniques (both structure- and ligand-based), and ADMET profiling, represents the novel accelerated strategy employed in this work to target Myc-Max oncogenic activity in CRPC. The strategy led to the discovery of novel small molecule inhibitors of Myc-Max transcriptional activity, able to reduce Myc-driven expression of AR-V7 splice variant that promotes CRPC, reduce growth of Myc-driven prostate cancer cell lines while minimally affecting a Myc-knockout cell line, induce apoptosis, and specifically disrupt Myc-Max/DNA interactions upon direct binding to the complex. Taking advantage of the publically available X-ray structure of Myc-Max bound to DNA recognition sequence and usage of in silico techniques led to the identification of a novel druggable site on the DNA-binding domain of the structurally ordered Myc-Max complex. The site at the DNA interface contains highly conserved residues from both Myc and Max 169 monomers, and as such is unique as not only differs from the few experimentally determined binding sites on Myc monomer only, both in its disordered and ordered forms, reported in the literature, but also the sites found on other Max-binding partners (i.e. Mad-Max). The use of the CADD approach resulted in identification of a number of novel Myc-Max inhibitors of different chemotypes. Resulting from an initial round of virtual screening and pharmacophore modeling, VPC-70067, a compound highly similar in structure, potency and mechanism of action to the literature reported Myc prototype inhibitor 10058-F4 served as proof-of-concept of the adopted in silico strategy. In addition, a novel compound VPC-70063 with a chemically different scaffold was identified as the best performer in a panel of in vitro assays as it inhibited Myc-Max transcriptional activity (IC50 = 8.9 μM), Myc-Max downstream functions, levels of the AR-V7 splice variant in CRPC cells, and cell growth in various PCa cell lines. In addition, VPC-70063 induced apoptosis as expected for a Myc inhibitor. Its specificity was further illustrated by the observed inhibitory effect on the Myc-Max association with DNA and by the cell viability experiment with Myc-negative cell line HO15.19. It should be noted, however, that in the latter system some cytotoxicity has been detected at higher concentrations of VPC-70063. Even though the extent of inhibition of Myc-negative cells was much less profound than that for Myc-positive ones (7% versus 50% at 2.5 μM administration), the observed general cytotoxicity was a concern and prompted its optimization. It should also be mentioned that the positive control used, compound 10074-G5 has also demonstrated 37% suppression of HO15.19 at 25 μM dosage.  The most recently reported small molecule Myc-Max inhibitor MYCMI-6 has also exhibited about 15% inhibition of HO15.19 at 10 μM. It is possible then these inhibitors could affect other Max containing complexes in HO15.19 cell line and/or that the corresponding 170 prototypical small molecule Myc-Max inhibitors all require further optimization and reduction of Myc-independent cytotoxicity. In fact, we calculated the quantitative estimate of drug-likeness (QED) score336, using the QED webserver337, for known Myc inhibitors to predict their desirable “chemical beauty”. Ten out of the twenty-two quantified compounds had a QED score greater or equal to 0.5, like 75% of orally available drugs, while only seven were ranked as having a desirable drug-likeliness profile with scores above 0.61. Although a low QED score might not rule out the potential usefulness of a small molecule as a drug candidate336, the estimated drug-likeliness profiles of the Myc-Max inhibitors suggest that the large majority, if not all, requires further optimization to become orally available clinically viable drugs. Of note, VPC-70063 was more potent and metabolically stable in mice liver microsomes (IC50 = 9 μM; t1/2= 69 min) than control 10074-G5 (IC50 = 16 μM; t1/2= 3 min). The full power of the in silico platform combining structure-based and ligand-based techniques harnessed towards optimization of VPC-70063 led to a large number of active derivatives with enhanced potency and reduced cytotoxicity that maintained the hydrophobic core (i.e. interactions with Phe921, Phe222, Leu917, and Ile218) and H-bonding interactions (i.e. with Arg215 backbone) of the parental compound. Development of the current lead inhibitor, VPC-70551, of a different chemical class, involved two rounds of in silico medicinal chemistry starting from initial hit VPC-70033. The first round of ROCS-based chemical similarity searches with VPC-70033 as a chemical template yielded a distant analog VPC-70127. Its selection was driven by excellent agreement in the results obtained from various in silico techniques, including consensus docking, MD simulations in explicit solvent, and ADMET prediction suggesting low metabolic risk (only three predicted metabolites). VPC-70127 demonstrated best inhibition of Myc-Max transcriptional activity with 171 an IC50 =1 μM, strong reduction of AR-V7 levels, a very strong effect on apoptosis and, importantly, it effectively inhibited the growth of LNCaP cells, with only 19% cytotoxicity at 12 μM in the HO15.19 Myc-/- knockout cell line.  The second round of similarity searches against the ENAMINE-REAL database of readily synthesizable chemicals using VPC-70127 structure without its reactive nitro group as template led to VPC-70551. Albeit demonstrating a small reduction in potency, IC50 = 4 μM, it demonstrated the minimum observed effect relative to other optimized hits and literature compounds on the viability of the Myc-knockout cell line, only 10% at 10 μM, likely due to the removal of the nitro moiety. In addition, replacement of a pyrazine ring of VPC-70127, predicted to metabolize, for a trifluoromethyl-phenyl ring of VPC-70551 eliminated the metabolic risk (i.e. no predicted metabolites). Indeed, assessment of VPC-70551 microsomal stability demonstrated an excellent half-life, t1/2 = 140 min. VPC-70551 is the current lead candidate for ongoing SAR-based development. A thorough analysis of the structural determinants of binding affinity (from pharmacophore matching, docking and MD simulations) and thus activity of VPC-70551 showed that the cyano-trifluoromethyl-phenyl ring of the compound is required for activity. Not only anchors 70551 in the hydrophobic core of the pocket, but also engages via the cyano moiety basic residues Arg214 (and Lys939 in docking) thus replacing the H-bonds that these residues otherwise form with DNA phosphates. The hydrazide linker is also required for activity as is involved in H-bonding interactions with Arg239 (Arg215 in docking), mutation of which abrogates DNA binding. Finally, the second benzene ring significantly overlaps with the DNA backbone and contributes to the binding of the ligand. Thus, the in silico protein-ligand contacts analysis of VPC-70551 designated a novel active scaffold: N'-[4-cyano-2-(trifluoromethyl)phenyl]benzohydrazide. In 172 SAR, substitutions with R-groups around the second benzene ring may result in enhanced activity and stability.  To summarize this section, VPC-70551 represents a potent and stable c-Myc-Max inhibitor that upon further optimization may soon lead to the development of the first Myc-Max drug candidate for treatment of CRPC. 173 Chapter 4: Targeting N-Myc-Max with c-Myc-Max Inhibitors  4.1 Background An emerging mechanism of resistance to targeted therapies in several cancer types including PCa is lineage plasticity, a process by which differentiated cells lose their identity, acquire alternative lineage programs and enhance the adaptability of tumor cells to changing environments associated with metastasis and treatment resistance.338 Emerging data from metastatic biopsies from patients progressing on AR-directed therapies (abiraterone or enzalutamide) suggest that a subset of late stage CRPC tumors can progress to NEPC by divergent clonal differentiation with an ancestry traceable back to a single mixed adenocarcinoma/castration resistant prostate cancer (CRPC-Adeno) precursor.63 During the course of disease progression and treatment resistance, these mixed tumors can evolve toward a predominantly poorly differentiated neuroendocrine (NE) phenotype with loss of AR-signaling dependence, AR expression, luminal/epithelial cell identity and acquisition of an alternative neuronal-like lineage phenotype.38 Several candidate drivers of plasticity toward NEPC have been identified including loss of TP53 and RB1 or upregulation of N-Myc, AURKA, EZH2, BRN2, and SOX2.338 Expression of N-Myc is not present in the epithelial cell lineage of the developing prostate but is amplified and/or overexpressed in the majority of NEPC cases and a subset (20%) of CRPC-Adeno tumors that display NEPC features.10 In AR independent setting, N-Myc overexpression thus leads to the development of highly aggressive and poorly differentiated N-Myc induced NEPC tumors. While current efforts are mainly focused on targeting AURKA/N-Myc protein-protein interactions occurring at the N-terminus of N-Myc and triggering its degradation, Phase II 174 clinical trials with the kinase inhibitor MLN8237 (Alisertib), for instance, demonstrated that only 7% of the evaluated patients (i.e. 4 out of 56) benefited from the treatment.339 MLN8237 (Alisertib) and CD532 kinase inhibitors induce an allosteric transition in AURKA that results in conformational changes that destabilize N-Myc and triggers its phosphorylation at the N-terminus, its ubiquitylation, and ultimately its proteasomal degradation. On the other hand, targeting N-Myc-Max interactions at the C-terminus of N-Myc and disrupting the critical binding of the complex to DNA to inhibit N-Myc-Max function is absent. Considering that c-Myc and N-Myc are highly similar both structurally and functionally as they share > 70% similarity in their bHLH DNA binding domain (DBD), upregulate transcription of a common set of genes involved in transformation, proliferation, apoptosis, metabolism and stem-like-state140 only upon the obligate dimerization with Max, targeting N-Myc-Max with c-Myc-Max DBD inhibitors is expected to have similar blocking effects. As previously outlined prototypical c-Myc inhibitors 10074-G5 and 10058-F4 inhibited both c-Myc and N-Myc proteins by binding at their independent and equivalent sites experimentally established on both disordered monomers. Moreover, the most recently reported disruptors of c-Myc-Max protein-protein interactions, KJ-Pyr-9 and MYCMI-6, demonstrated effectiveness against N-Myc-Max activity, albeit their specific sites and mode of binding remain unknown (see section 1.3.2.1). Blocking N-Myc-Max DNA-binding and its transforming activity with c-Myc-Max DBD inhibitors developed by CADD approaches is unprecedented and represents a promising novel targeted strategy for treatment of lethal NEPC.  175 4.2 Results As there is no publically available X-ray structure of N-Myc-Max and given the high structural similarity of the bHLHLZ domains of c-Myc and N-Myc, a homology model of the N-Myc-Max complex was built at high 0.7 Å resolution using MODELLER with the 1NKP X-ray structure of the c-Myc-Max heterodimer bound to E-box used as a template (see section 2.1.1). The same computational methods for binding site identification used in the case of c-Myc-Max revealed, unsurprisingly, a binding site of identical composition and chemical character at the DNA interface of the N-Myc-Max homology model (Figure 4.1).  Figure 4.1 High-resolution N-Myc-Max homology model using c-Myc-Max X-ray structure as a template. Protein residues forming identical binding sites for small molecules are indicated below the multiple alignment with stars: c-Myc (dark blue), N-Myc (cyan), Max (red).  176 The CADD protocol subsequently applied toward N-Myc-Max inhibition starting with virtual re-screening and re-scoring of the developed c-Myc-Max inhibitors followed by experimental evaluation of their inhibitory effects against relevant cell models of neuroblastoma and NEPC showed that optimized c-Myc-Max inhibitors had similar in silico predicted binding affinities for N-Myc-Max and similar experimental activities. Standing as the best example, the current lead compound VPC-70551, docked in a highly similar fashion in both c-Myc-Max and N-Myc-Max models with an RMSD between the two poses of 0.76 Å and almost identical Glide scores (-4.8 and -4.6 kcal/mol, respectively) (Figure 4.2). Further flexible MM/GBSA minimization of the docked structures in implicit solvent revealed that in both complexes VPC-70551 formed H-bonds between the cyano moiety with the side chain of Arg214 of Max and between the carbonyl oxygen of the hydrazide linker with the side chain of Arg215 of Max. The compound formed additional hydrophobic interactions between its trifluoromethylphenyl rings at both ends of the linker with Arg215 and Ile218 of the binding site hydrophobic core, respectively. Binding of the compound to N-Myc-Max was nonetheless energetically more favorable as reflected by larger negative free energy of binding of -40.3 kcal/mol relative to -33.1 kcal/mol affinity for c-Myc-Max (Figure 4.3) due to additional H-bonding interactions with the side chain of Lys43 of N-Myc (equivalent of Lys939 of c-Myc). All the above-mentioned basic residues form critical H-bonds with the phosphate groups of the DNA backbone. Moreover, mutation of Arg215 abrogated c-Myc-Max and N-Myc-Max binding to DNA.   177  Figure 4.2 Binding poses and interactions of VPC-70551 within the c-Myc-Max (A) and N-Myc-Max (B) DBD pockets. 178  Figure 4.3 VPC-70551 interactions with c-Myc-Max (A) and N-Myc-Max (B) obtained from MM/GBSA simulations in implicit solvent. Binding of VPC-70551 to N-Myc-Max is energetically more favorable due to additional H-bonding interactions with the side chain of Lys43 of N-Myc.  Further experimental validation demonstrated that VPC-70551 inhibited the transcriptional activity of N-Myc-Max in IMR32 neuroblastoma cell line in which N-Myc is amplified and actively expressed340 with identical potency, IC50 = 4 μM, as that demonstrated in inhibiting c-Myc-Max activity in LNCaP cells (Figure 4.4A). Moreover, treatment with VPC-70551 showed a dose-dependent reduction of apoptosis in IMR32 cells induced by caspase 3/7 pathway as 179 measured in real-time using the caspase 3/7 green live cell analysis on IncuCyte®, ruling out nonspecific toxicity (Figure 4.4B).  Figure 4.4 Effects of lead compound VPC-70551 on N-Myc-Max transcriptional activity in IMR32 N-Myc amplified cell line (A), on apoptosis (B) and downstream target genes (C) in the same cell line.  Furthermore, in IMR32 cells, treatment with 5 μM of 70551 showed downregulation of N-Myc target genes, such as LMO3341, TFAP4342 and HK2343 involved in proliferation and metabolism as measured by qRT-PCR after 24 hours treatment with the compound (Figure 4.4C). LMO3 is part of an N-Myc-driven signature able to identify patients with poor prognosis in amplified N-Myc molecular subtype settings.341 LMO3 is a neuronal-specific transcription factor considered an oncogene in neuroblastoma. It interacts with the tumor suppressor p53 and inhibits its activity.344 Moreover, N-Myc promotes neuroblastoma by establishing a regulatory circuit with TFAP4, a bHLHLZ transcription factor, positively regulated by N-Myc, which maintains cells in a proliferative state by blocking p21 expression that normally promotes cell cycle arrest.342,345 N-Myc plays a critical role in aerobic glycolysis in neuroblastoma by 180 activating the transcription of multiple glycolytic genes. Of note, expression of HK2 was significantly higher in N-Myc amplified tumors than in tumors without N-Myc amplifications.346  Notably, the lead compound, VPC-70551, effectively stopped the growth of N-Myc driven IMR32 cells in a dose-dependent manner while it did not affect the viability of N-Myc independent cell lines, SK-N-AS and NB-16 (neuroblastoma cell lines expressing mutant TP53 and lacking N-Myc) confirming its specific mode of action (Figure 4.5).  Figure 4.5 Effect of lead compound VPC-70551 on viability of N-Myc-driven IMR32 cells (A), and N-Myc independent SK-N-AS and NB-16 neuroblastoma cell lines (B).  Importantly, VPC-70551 strongly affected the viability of LASCPC-01347 cell line having N-Myc as the driver for NEPC (Figure 4.6) as measured in real-time using IncuCyte® technology.  181  Figure 4.6 Effect of lead compound VPC-70551 on viability of N-Myc-driven LASCPC-01 NEPC cell line.   4.3 Discussion Transformation of castration-resistant prostate cancer (CRPC) towards androgen signaling independence has emerged as a resistance mechanism in a subset of metastatic CRPC following exposure to AR-targeted therapies such as abiraterone or enzalutamide. Under sustained therapeutic inhibition, CRPC can progress to its ultimate stage, NEPC. This highly aggressive stage of the disease has limited treatment options as even cytotoxic chemotherapy offers little survival benefit and NEPC patients succumb to the disease in less than a year. Hence, the development of novel therapeutic approaches for patients with NEPC represents a clinical unmet need.  N-Myc overexpression has been identified as an important driver of NEPC as it confers poorly differentiated, highly proliferative and invasive features of this lethal phenotype. Therefore, N-Myc is a most wanted therapeutic target for potential treatment of NEPC. 182 Given the high structural and functional similarity between c-Myc and N-Myc, we hypothesized that the developed c-Myc-Max inhibitors would similarly affect the transforming activity of N-Myc-Max. In the absence of a publicly available X-ray structure of the N-Myc-Max complex and given the high homology between c-Myc and N-Myc in their DNA binding domains, an N-Myc-Max model was built at high resolution using the crystal structure of c-Myc-Max as a template. An equivalent site at the DNA interface was then identified using in silico techniques. Docking experiments with c-Myc-Max inhibitors in the N-Myc-Max site produced similar binding affinities. Simulations in implicit solvent suggested tighter binding of the compounds to the N-Myc-Max complex. Experimental validation of the current c-Myc-Max inhibitor, VPC-70551, demonstrated identical potency of the compound in N-Myc driven IMR32 cell line. VPC-70551 strongly affected the viability of IRM32 neuroblastoma with no effect on N-Myc independent neuroblastoma cell lines suggesting its specific mode of action. Importantly, VPC-70551 strongly affected the growth of N-Myc driven LASCPC-01 NEPC cell line.  Moreover, treatment with 70551 at low concentrations showed downregulation of N-Myc target genes in the IMR32 cell line. Optimization of in vitro assays, such as BLI and EMSA with purified N-Myc-Max complex is currently underway to assess the specificity of binding and disruption of protein-DNA interactions upon treatment with lead compound. Further optimization of VPC-70551 potency (IC50 = 4 μM) and stability (t1/2 = 140 min) in ongoing SAR, and future in vivo PK/PD profiling may lead to the development of the first N-Myc-Max directed drug to effectively treat NEPC. 183 Chapter 5: Development of hnRNP A1 Inhibitors  5.1 Background The heterogeneous nuclear ribonucleoprotein A1 (hnRNP A1) is a multifunctional RNA-binding protein that regulates alternative pre-mRNA splicing, transcription, nucleocytoplasmic shuttling, miRNA processing, and telomere elongation maintenance, as well as translation of cellular transcripts both in physiological and pathological conditions.348,349 Overexpression of hnRNP A1 in various cancer types, including prostate67,350, lung351, stomach352, and breast353 cancers, Burkitt lymphoma354, multiple myeloma355, leukemia356 and neuroblastoma357, has been associated with tumorigenesis, cancer progression and drug resistance. The molecular mechanisms by which hnRNP A1 supports malignant transformation, either directly or through interplay with well-established cancer drivers, include regulation of cell survival, alteration of cell cycle, invasion and metastasis, altered metabolism and stress adaptation (i.e., to hypoxia, starvation, and response to DNA damage), all of which are recognized hallmarks of cancer.348,349,358 In prostate cancer, alternative splicing (AS) plays a prominent role as it represents a mechanism of resistance to therapy.350 There is accumulating evidence on hnRNP A1 splicing regulatory interconnectivity with the androgen receptor (AR) and c-Myc oncogenic signaling pathways. Indeed, emerging data have implicated hnRNP A1 as a central player in a splicing regulatory circuit involving its direct transcriptional control by c-Myc and the production of the constitutively active ligand-independent alternative splice variant of androgen receptor, AR-V7, which promotes castration-resistant prostate cancer (CRPC). It has been demonstrated that under direct transcriptional control of c-Myc, in CRPC, hnRNP A1 promotes the production of the AR-184 V7 splice variant33,67,139,358, which in turn directly regulates the expression level of key proteins involved in cell-cycle progression and proliferation, in particular the ubiquitin conjugating enzyme 2C (UBE2C) in androgen-deprived 22Rv1 cells.68,359 In CRPC, knockdown of hnRNP A1 with short-interfering RNA (siRNA) suppressed AR-V7 levels and growth of 22Rv1 cells.67 Furthermore, in cancer, hnRNP A1 contributes to aerobic glycolysis by promoting the expression of the pyruvate kinase PKM2 isoform, a hallmark of tumor growth, as well as ∆Max, a Myc-associated factor X (Max) isoform, enhancer of Myc-dependent transformation and mediator of Myc-dependent tumor metabolism.360,361 Moreover, it is notable that while hnRNP A1 is transcriptionally upregulated by c-Myc, it is also one of the factors known to act on the c-Myc promoter to regulate c-Myc transcription.362,363 In addition, hnRNP A1 directly promotes translation of c-Myc protein itself by binding to internal ribosome entry sites (IRES) found in the 5′-UTR (untranslated region) of c-Myc mRNA.364 How hnRNP A1 discerns among its various functions thus depends on recognition and binding to specific RNA sequences (also, DNA for functions such as transcription or telomere elongation). While hnRNP A1 was characterized early as a splicing repressor binding to exonic or intronic splicing silencers, recent evidence indicates that hnRNP A1 more globally affects alternative splicing outcomes by binding directly to 3′-splice sites (3′ss) that contain consensus 5′-YAG-3′ motifs.365-367 It has been demonstrated that the AR-V7 splice variant is generated via splicing at the alternative 3′ss next to the cryptic exon 3B which includes a stop codon, rather than the 3′ss next to AR exon 4, resulting in the translation of the AR-V7 C-terminal truncated form of AR33,368. Importantly, emerging data suggest that the recruitment of hnRNP A1 to the AR-V7 3′ss in AR pre-mRNA is increased in 22Rv1 cells with acquired enzalutamide 185 resistance67, and that the specificity of hnRNP A1–RNA binding, as determined by X-ray crystallography, is dictated by the 5′-AG-3′ dinucleotide.300 Structurally, hnRNP A1 has a modular domain organization (Figure 5.1) consisting of tandem N-terminal RNA recognition motifs (RRM1 and RRM2, respectively), collectively referred to as unwinding protein 1 (UP1), and an intrinsically disordered Glycine-rich C-terminal region. The C-terminus mediates homologous and heterogeneous protein–protein interactions. The Gly rich C-terminal region contains an Arg-Gly-Gly (RGG) box, which is involved in non-specific RNA binding, and a nuclear targeting sequence, termed M9, which is responsible for bidirectional nucleocytoplasmic shuttling. UP1 is the primary RNA-binding domain (RBD) involved in specific protein–RNA interactions and alternative splicing events.300,348  Figure 5.1 hnRNP A1 protein domains organization.  To date, inhibition of hnRNP A1 has been demonstrated with the naturally occurring flavonoid quercetin (Figure 5.2), a rather promiscuous, poorly soluble and poorly bioavailable compound.369-371   Figure 5.2 Chemical structure of quercetin, literature hnRNP A1 inhibitor. 186 Identified as an hnRNP A1 binder through chemical proteomics, quercetin exerted nonetheless anti-cancer effects in PC3 AR-independent PCa cell line by a mechanism of action that involves binding to the C-terminal region of hnRNP A1, impairment of hnRNP A1 ability to shuttle between the nucleus and cytoplasm, resulting in its cytoplasmic retention and accumulation, driving cells toward apoptosis.372 More recently and pertinent to this work, quercetin has been shown to downregulate hnRNP A1 and thus AR-V7 expression in CRPC cell lines, and to re-sensitize enzalutamide-resistant 22Rv1-injected mouse xenografts in vivo.373  5.2 Results 5.2.1 Binding site identification on hnRNP A1 RBD domain  The published 1.92 Å X-ray structure of the UP1 RNA-binding domain (RBD) of hnRNP A1 bound to its RNA 5′-AGU-3′ trinucleotide target sequence (PDB ID: 4YOE)300 was utilized for identification of plausible pockets where small molecule inhibitors could bind and specifically disrupt hnRNP A1 binding to RNA and, therefore, to interfere with the subsequent splicing activities. The 4YOE X-ray structure provided the first insights into the specificity of hnRNP A1–RNA recognition, which indicated that UP1 specifically interacts with the 5′-AG-3′ ribodinucleotide within a nucleobase cavity formed upon folding of the RRM1 motif and the inter-RRM linker. RRM2 made no contact with the RNA sequence. Both A (adenine) and G (guanine) purines engaged in stereospecific contacts within the pocket thus explaining the preference for the 5′-AG-3′ sequence.300 Their replacement for pyrimidines, having smaller ring size, were deemed unlikely to satisfy the favorable interactions of the purines with similar energetics. For instance, single or double substitutions for cytosine resulted in a ~20-fold reduction in UP1 RNA-binding affinity, rate limiting for complex formation.300,365,374 187 Based on the prior crystallographic evidence regarding the structural determinants of hnRNP A1–RNA recognition specificity, we selected the RRM1 and the inter-RRM linker regions of UP1 for binding site identification. We employed the Site Finder module of the Molecular Operating Environment (MOE) suite of programs303, which predicted a binding site (Figure 5.3) significantly matching the X-ray defined nucleobase pocket.  Figure 5.3 In silico model of the UP1 domain of the hnRNP A1 splicing factor bound to the 5’-AG-3’ recognition sequence constructed based on the 1.92 Å X-ray structure of RNA-bound UP1 (PDB ID:4YOE). The RNA bases are shown as sticks and colored in green. The predicted binding site on hnRNP A1 at the RNA recognition interface is represented as a grey solid surface. Virtual atoms utilized to probe the protein surface are shown as brown spheres within the identified pocket.  The predicted binding site is shaped by residues: Gln12, Lys15, Phe17, Met46, Arg55, Phe57, Phe59, Lys87, Arg88 from the RRM1 motif, and Ala89, Val90, Ser91, Arg92, Ser95 and His101 188 from the inter-RRM linker. The functional relevance of these residues has been well described (Figure 5.4).300   Figure 5.4 hnRNP A1 interactions with 5’-AG-3’ RNA recognition sequence within the pocket. Hydrogen-bonding interactions formed between the nucleobases with the backbone and side chains of Val90, Arg88, Gln12 residues, as well as those formed by the sugars of the RNA backbone with the side chains of Arg92 and Ser95 are indicated with red dashed lines. Hydrophobic interactions with Phe17, His101, Phe59 and aliphatic side chains are indicated with dashed green lines.  Phe17 and His101 contribute to the majority of binding free energy, originating from favorable van der Waals and π-π stacking interactions with the rings of the adenine at first position in the RNA 5‘-AG-3‘ sequence. Moreover, Val90 and Arg88 backbone amides make specific H-bonding interactions with the adenine nitrogen atoms. Additional H-bonding and 189 hydrophobic packing are contributed from the hydroxyl group of Ser95 and Phe57, respectively. At the second 5′-AG-3′ RNA position, amino groups of Gln12 and Lys15 primarily and selectively recruit the guanine through H-bonding via their amino group. In addition, Val90 and Arg92 provide H-bonding capabilities via their backbone carbonyl and guanidinium groups. Cation-π and π-π interactions with Arg92 and Phe59, respectively, further enhance the interactions with the guanine.300 Small molecule inhibitors targeting this site are expected to block UP1 binding to RNA by interfering with the most functionally relevant protein–RNA interactions within the site, and to alter hnRNP A1′s splicing activity.  5.2.2 In silico identification of hit compounds targeting the hnRNP A1 binding site  Subsets of drug-like molecules deposited in the ZINC open chemical repository308,309,317 having in-stock availability from selected vendor catalogues, filtered by logP (octanol-water partition coefficient) parameter, charge and chemical reactivity, were virtually screened against the identified hnRNP A1–RNA pocket utilizing Glide docking software220,221 in standard precision (SP). Compounds with the best docking scores were prioritized for subsequent in silico scoring based on calculated pKi315 and RMSD (root mean square deviation) between poses obtained from various docking programs, including ICM263 and Hybrid264. As a result, 139 chemicals were selected for purchase based on two or more of the abovementioned binding affinity indicators and satisfaction of at least three essential π-π or H-bonding interactions with the residues shaping the pocket, in particular contacts with Phe17, His101, Val90 and Phe59. These compounds were then subjected to experimental evaluation through the UBE2C reporter assay for assessing their effect on AR-V7 driven luciferase-detected UBE2C activity in 190 22Rv1 cells in androgen-deprived conditions, with quercetin, the previously reported hnRNP A1 binder used as positive control. A dozen compounds demonstrated more than 50% inhibition of UBE2C activity at a 25 μM concentration in this assay.  5.2.2.1 VPC-80021  VPC-80021 was the earliest identified small molecule hit predicted to disrupt hnRNP A1–RNA interactions, which demonstrated ability to reduce AR-V7 levels in the UBE2C assay.  Mode of binding of VPC-80021 to hnRNP A1 RBD site The predicted binding pose of VPC-80021 is shown in Figure 5.5. The chemical structure VPC-80021 is composed of an 4chloro-1H-pyrrolo[2,3-b]pyridine (4chloro-7-azaindole) fused heterocycle at one end, a aminoethyl linker and a 3H-quinazolin-4-one ring at the other end. Within the docking site at the hnRNP A1–RNA binding interface, the 7-azaindole ring of VPC-80021 forms π-π stacking interactions with Phe17. Moreover, the 7-azaindole ring forms two hydrogen bonds: one with the carbonyl oxygen of Arg88 backbone via the NH of the pyrrole ring and a second with the Val90 backbone amide via the nitrogen of the pyridine ring. Two additional H-bonds are formed with the carbonyl oxygen of Val90 backbone by the amino group of VPC-80021 linker and by the NH of the quinazolin ring. Furthermore, the quinazolin ring of VPC-80021 makes significant hydrophobic interactions with the aromatic ring of Phe59, as well as with the aliphatic side-chains of Arg92 and Gln12.  All these interactions made significant contributions to the binding specificity of hnRNP A1 to the cognate 5′-AG-3′ RNA sequence, as described in Subsection 5.2.1. In the predicted docking pose, the 7-azaindole ring of VPC-80021 was positioned such that it significantly 191 overlapped that of the 5′ adenine in the X-ray structure, fully mimicking the purine in its hydrogen-bonding pattern. (Figure 5.5 right).   Figure 5.5 (Left) Predicted binding pose of VPC-80021 within the hnRNP A1 RBD pocket. VPC-80021 is shown in stick representation (orange = carbon, blue = nitrogen, red = oxygen, green = fluorine). Interacting residues are shown as sticks and colored in light blue-grey. The pocket surface is colored in light grey. Hydrogen bonds formed by VPC-80021 with the backbone of Val90 and Arg88 are indicated with red dashed lines. Hydrophobic interactions with aromatic rings of Phe17, His101 and Phe59 as well as with aliphatic side-chains of residues shaping the pocket are indicated with green dashed lines. (Right) Significant structural overlap between VPC-80021 and nucleobases of the 5′-AG-3′ RNA within the hnRNP A1 RBD pocket.  Similarly, there is significant overlap between the quinazolin ring and the guanine at the second position in the 5′-AG-3′ recognition sequence. The Glide score of VPC-80021 was established as −9.3 kcal/mol, the most significant among all hits. Its binding affinity was dominated by hydrophobic interactions as per calculated pKi (4 out of 6.8) with a good balance of H-bonding interactions contributing with a 2.4 score to the overall pKi. Excellent RMSD 192 values calculated from consensus scoring with ICM and Hybrid docking programs as well as Glide “blind docking” in SP and XP modes were 0.37, 0.71, 0.11 and 0.19 Å, respectively. The estimated binding free energy of VPC-80021 obtained from MM/GBSA simulations in implicit solvent was −83.7 kcal/mol, ranking second among all purchased compounds and first among the identified 13 hits. In these simulations with protein flexibility within 5.0 Å from the ligand, VPC-80021 maintained the protein-ligand interactions predicted by rigid docking and, in addition, formed a strong H-bond between the keto group of the quinazolin ring and the side chain of Gln12 as well as cation-π interactions with the side chain of Arg92 (Figure 5.6).  Figure 5.6 hnRNP A1/VPC-80021 interactions obtained from MM/GBSA simulations with implicit solvent in two-dimensional representation. Gln12 further strengthened ligand binding via H-bonding interactions. Green lines indicate π-π interactions, red line indicates cation-π interactions, solid and dashed purple arrows represent backbone and side chain hydrogen-bonding interactions, respectively, and the curved contour represents the shape of the pocket around the ligand colored by the character of the interacting protein residues: (green = hydrophobic, blue = positively charged, cyan = polar). 193 The dynamic behavior of the hnRNP A1/VPC-80021 complex when fully subjected to force field in explicit solvent as opposed to the MM/GBSA implicit solvent approach was further investigated by classical molecular dynamics (MD) simulations. Analysis of the trajectory obtained from long 350 ns simulations showed that the complex was extremely stable with RMSD deviations of 2.3 Å and 2 Å for the protein backbone and for the ligand relative to the protein, respectively (Figure 5.7A). Throughout the 350 ns simulation time, VPC-80021 was engaged in five hydrogen-bonding interactions that occurred for more than 88% of simulation time: three with the backbone of Val90, a forth with the backbone of Arg88, and the fifth with Gln12 side chain. Additionally, VPC-80021 formed π-π interactions with Phe17 for 41% of simulation time and with Phe59 to a slightly lesser extent (Figure 5.7B and C). Overall, the in silico techniques indicated tight binding of VPC-80021 to the identified hnRNP A1 RBD pocket able to outcompete RNA by strongly taking over the essential hnRNP A1/5′-AG-3′ RNA interactions contributing to the majority of binding free energy (as described in section 5.2.1).  194  Figure 5.7 Dynamics of hnRNP A1-80021 complex during 350 ns MD simulations with explicit solvent. (A) RMSD displacement during the simulation trajectory of the protein backbone (blue) and of the ligand with respect to the protein (orange); (B) Protein-ligand contacts during the simulation time course. (C) Two-dimensional schematic of detailed interactions between VPC-80021 atoms with the protein residues. Binding of the ligand occurs via hydrogen bonding with Val90, Arg88, and Gln12 (purple arrows) and π-π interactions with Phe17 (green lines).  195 Development of VPC-80021  Given the excellent in silico profile of VPC-80021, supported by preliminary experimental validation on its inhibitory effects of AR-V7 levels (in the designed hnRNP A1/AR-V7/UBE2C monitoring assay), medicinal chemistry (medchem) was performed on VPC-80021. Multiple derivatives of VPC-80021 were designed in silico to enhance known interactions of hnRNP A1 RBD with the 5′-AG-3′ RNA sequence. As such, fused heterocycles of 5- or 6-membered rings were considered to maximize π-π stacking interactions as those formed by the RNA purines. Moreover, H-bonding acceptors or donors were introduced to engage additional residues in the pocket, such as Lys15. All in silico generated compounds were then docked in the RBD pocket using Glide both in standard precision (SP) and extra precision (XP) modes. Blind docking was further performed in XP mode. RMSD-based consensus was then calculated not only between the poses obtained from Glide in various modes but also with eHiTS predicted poses. The designed derivatives were further scored based on indicators including pKi, ligand efficiency, and ADMET risk. Importantly, free energy perturbation MD simulations were performed utilizing Desmond FEP+279,280, part of Schrӧdinger’s suite of programs, to derive relative binding free energy differences between morphed derivatives as the most robust and accurate ranking method predictive of their relative potency. The chemical structures and scores of selected in silico medchem compounds are shown in Figure 5.8 and Table 5.1, respectively. 196  Figure 5.8 Chemical structures of representative in silico designed derivatives of VPC-80021.  Enhanced protein-ligand interactions, such as additional hydrogen bonding interactions between derivatives with the side chain of Lys15 account for the increase in pKi, ligand efficiency (LE) and, importantly, relative binding free energy differences, as exemplified in Figure 5.9 on the perturbation of derivative 80021-A6 from the 80021-A1 intermediate, obtained from FEP+ MD simulations. 80021-A1 is the first analog of the parental compound 80021 with the low-risk halogenure removed. 197   Figure 5.9 hnRNP A1-ligand interactions diagram obtained from FEP+ MD simulations of derivative 80021-A6 morphed from 80021-A1 intermediate analog of parental VPC-80021. Binding of 80021-A6 is energetically favored due to additional strong H-bonding with Lys15 side chain. Improved potency is expected.   Table 5.1 In silico scores and synthesis cost of representative VPC-80021 medchem derivatives. VPC ID Glide pKi LE tPSA ADMET RMSD* ∆∆G FEP+ $$$ 80021 -9.32 6.78 -0.39 86.5 2.1   $900 80021-A5 -9.78 6.88 -0.44 94.5 1.1 1.17 -0.85 $12,900 80021-A5-M4 -10.34 7.31 -0.44 114.8 2.9 1.13 -0.94 $12,580 80021-A7 -10.32 7.16 -0.46 108.2 3.1 1.58 -1.42 $13,080 80021-A5-M2 -10.20 6.70 -0.44 114.8 0.2 2.52 -2.04 $12,580 80021-A6 -9.98 7.03 -0.45 110.9 1.6 1.04 -2.21 $7,500 80021-A7-A2 -9.90 6.17 -0.45 133.9 4.5 2.07 -2.47 $14,900 * eHiTS RMSD 198 As the in silico data was convincing as predictive of potentially improved potency, a request for synthesis quote was placed in view of purchasing and experimental testing of designed derivatives. Unfortunately, the prices for chemical synthesis were extremely high in the $7,500-$14,900 range (Table 5.1), bringing the development effort of VPC-80021 to a full stop. Similar rounds of in silico design of potential hnRNP A1 binders will be conducted in future work, using simpler starting medchem templates to reduce the potential high cost of medicinal chemistry campaign.  5.2.2.2 VPC-80051  As the development of VPC-80021 came to a halt due to prohibitive synthesis costs of the compound itself and, importantly, derivatives predicted to be more potent, hit VPC-80051 was the next compound considered for in-depth investigation.  Mode of binding of VPC-80051 to hnRNP A1 RBD site The predicted binding pose of VPC-80051 is shown in Figure 5.10A. The molecule is composed of an indazole bicyclic aromatic moiety at one end, a carboxamide linker and a difluorophenyl ring at the other end. Within the docking site at the hnRNP A1–RNA binding interface, the benzene ring of the indazole group of VPC-80051 forms π-π stacking interactions with both Phe17 and His101. Moreover, the pyrazole ring of the indazole group forms three hydrogen bonds with the backbone of Val90: two between the Val90 backbone amide with the pyrazole nitrogen atoms and a third with Val90 backbone carbonyl. In addition, a fourth hydrogen bond is formed between the amine hydrogen of the carboxamide linker and the backbone carbonyl of Val90.  199 All these interactions made significant contributions to the binding specificity of hnRNP A1 to the adenine in the first position of the cognate 5′-AG-3′ RNA sequence, as described in Subsection 5.2.1. In the predicted docking pose, the indazole moiety of VPC-80051 was positioned such that it significantly overlapped that of the 5′ adenine in the X-ray structure with a calculated RMSD of 2.85 Å, taking over these specific interactions within the cavity (Figure 5.10A right). Furthermore, the difluorophenyl ring of VPC-80051 makes significant hydrophobic interactions with the aromatic ring of Phe59, as well as with the aliphatic side-chains of Arg92 and Lys15, overlapping to a good extent with the guanine at the second position in the 5′-AG-3′ recognition sequence with RMSD of 2.82 Å (Figure 5.10A right). The Glide score of VPC-80051 was −8.2 kcal/mol and its binding affinity was dominated by hydrophobic interactions as per calculated pKi (4.5 out of 6.8).  For comparison, the Glide score of the literature compound quercetin was −6.2 kcal/mol and its predicted pKi was 6. Quercetin’s predicted binding pose is shown in Figure 5.10B. Quercetin forms two hydrogen bonds: one between a hydroxyl group attached to the chromen-4-one moiety at first position with the backbone carbonyl of Arg88 and a second between a hydroxyl of its catechol moiety at second position with the backbone carbonyl of Val90. Additionally, π-π interactions are formed between the chromen-4-one moiety of quercetin with Phe17 and His101. While the overlap between the trihydroxy-chromen-4-one group with the 5′ adenine of RNA is significant with an RMSD of 2.7 Å (Figure 5.10B right), there is little overlap at the second position between the catechol (dihydroxyphenyl) ring and the guanine base (RMSD = 4.3 Å). 200  Figure 5.10 (a) Left. Predicted binding pose of VPC-80051 within the hnRNP A1 RBD pocket. VPC-80051 is shown in stick representation (magenta = carbon, blue = nitrogen, red = oxygen, green = fluorine). Interacting residues are shown as sticks and colored in light blue-grey. The pocket surface is colored in light grey. Hydrogen bonds formed by VPC-80051 with the backbone amide and carbonyl of Val90 are indicated with red dashed lines. Hydrophobic interactions with aromatic rings of Phe17, His101 and Phe59 as well as with aliphatic side-chains of residues shaping the pocket are indicated with green dashed lines; Right. Significant structural overlap between VPC-80051 and nucleobases of the 5′-AG-3′ RNA within the hnRNP A1 RBD pocket; (b) Left. Predicted binding pose of quercetin. Hydrogen bonds formed with Val90 and Arg88 backbone carbonyls are indicated with red dashed lines; Right. Structural overlap between quercetin and the 5′-AG-3′ RNA. Minimal overlap with the guanine base. VPC-80051 chemical name: (N-[(1S)-1-(3,4-difluorophenyl)ethyl]-1H-indazole-3-carboxamide). Quercetin chemical name: 2-(3,4-dihydroxyphenyl)-3,5,7-trihydroxy-4H-chromen-4-one. 201 Further confidence in the binding pose of VPC-80051 was provided by calculated RMSD values of 1.4 and 1.1 Å between the Glide pose and the poses obtained from ICM and Hybrid docking programs, respectively. Of note, the RMSD with the best pose generated by a “blind docking” Glide SP calculation, in which conformational sampling occurs within a grid large enough to cover the entire UP1 domain of the hnRNP A1 protein and not only the smaller grid centered on selected RRM1 and inter-RRM residues of the predicted binding site, was 1.7 Å. This provided good evidence on the accuracy of the binding site model as well as on the affinity of VPC-80051 for the site having a large number of equivalent interactions to those formed by 5′-AG-3′ cognate RNA with the X-ray nucleobase pocket. The docked structure was further minimized in implicit solvent and the binding free energies was calculated with the MM/GBSA method. The estimated binding free energy of VPC-80051 obtained from MM/GBSA simulations was −60.7 kcal/mol relative to −57.9 kcal/mol for quercetin control. In these simulations, in which the binding site residues were allowed to relax (i.e., protein flexibility within 5.0 Å from the ligand) during the minimization procedure of the complexes, VPC-80051 and quercetin maintained the protein-ligand interactions predicted by rigid docking (Figure 5.11). 202  Figure 5.11 hnRNP A1-ligand interactions obtained from MM/GBSA simulations with implicit solvent for: (a) VPC-80051 and (b) quercetin in two-dimensional representation. Green lines indicate π-π interactions, purple arrow pointed lines represent backbone hydrogen-bonding interactions, and the curved contour represents the shape of the pocket around the ligand colored by the character of the interacting protein residues: (green = hydrophobic, blue = positively charged, cyan = polar).  Extensive molecular dynamics (MD) simulations were subsequently carried out to gain further insights into induced fit motions and solvent effects, and thus, the dynamic behavior of the hnRNP A1/VPC-80051 complex when fully subjected to force field in explicit solvent as opposed to the MM/GBSA implicit solvent approach. Analysis of the trajectory obtained from MD simulations showed that VPC-80051 interacted favorably with the pocket residues for more than 80% of the 100 ns simulation time mainly by hydrogen-bonding between its indazole moiety with Val90 backbone and π-π interactions with Phe17 (Figure 5.12). Interactions with His101 and Phe57 are seconding. Superimposition of conformations of the simulated complex taken at different time steps revealed that the side chain of Arg92 underwent larger fluctuations 203 relative to other side chains in the pocket (Figure 5.12C), opening and making the binding site more exposed to the solvent. As such, with more translational and rotational freedom, the difluorophenyl ring of VPC-80051 is no longer engaged in hydrophobic interactions with Arg92 side chain or π-π interactions with Phe59, and as such does not contribute substantially to the binding of the ligand.   In vitro characterization of VPC-80051 To quantify the ability of hit compounds to interact directly with hnRNP A1 RBD, the Bio-Layer Interferometry (BLI) technique was employed with purified UP1 domain of the hnRNP A1 protein (N-terminal residues 1-196) immobilized on a streptavidin biosensor. Compound VPC-80051, as well as quercetin control, demonstrated direct binding to hnRNP A1 in a dose-dependent manner (Figure 5.13A). Future NMR spectroscopy or X-ray crystallography structural studies may unequivocally prove the binding mode of VPC-80051 to the hnRNP A1 RBD pocket.  Quantitative RT-PCR (qRT-PCR) was further employed to assess the effect of hits on the mRNA levels of AR-V7 in 22Rv1 cells in androgen-deprived conditions. Treatment with VPC-80051 at 25 and 10 μM concentrations resulted in a reduction of AR-V7 levels comparable to that of quercetin. The ratios of measured cycle threshold differences (∆CT) of levels of AR-V7 mRNA versus DMSO control normalized to actin, in percentage, were 66.20% and 79.55% for VPC-80051 at 25 and 10 μM, respectively, compared to 62.25% and 71.15% for quercetin at same concentrations. The reduction of AR-V7 levels observed in qRT-PCR upon 24 h treatment with VPC-80051 and quercetin was corroborated by Western blot analysis (Figure 5.13B). 204 Furthermore, the decrease in AR-V7 levels correlated with a reduction in viability of 22Rv1 cells treated with VPC-80051 at tested concentrations (Figure 5.13C).  Figure 5.12 Dynamics of hnRNP A1-80051 complex during 100 ns MD simulations with explicit solvent. (a) RMSD displacement during the simulation trajectory of the protein side chains (blue) and of the ligand with respect to the protein (orange). (b) Protein-ligand contacts during the simulation time course. Val90 and Phe17 are main interacting residues. (c) Three-dimensional superimposition of hnRNP A1 binding site and VPC-80051 at the beginning of the simulation and after 88.4 ns simulation time. Light grey and light magenta indicate the protein residues and the ligand in the reference frame while dark grey and dark magenta indicate the residues and the ligand in the selected 88.4 ns snapshot. The blue curved arrow indicates the larger movement of Arg92 side chain during the simulation. (d) Two-dimensional schematic of detailed interactions between VPC-80051 atoms with the protein residues. Binding of the ligand occurs via hydrogen bonding and water bridges with Val90 (purple arrow pointed lines) and π-π interactions with Phe17 (green lines).  205  Figure 5.13 In vitro characterization of VPC-80051 hnRNP A1 inhibitor. (a) Dose-dependent direct binding of VPC-80051 and control quercetin (QRCT) to purified UP1 domain of hnRNP A1 protein quantified by bio-layer interferometry (BLI). (b) Reduction of levels of AR-V7 splice variant upon treatment with 10 and 25 μM of VPC-80051 and QRCT as analyzed by Western blotting. (c) Reduction of 22Rv1 cell viability upon treatment with VPC-80051 and QRCT at greater than 10 μM doses where decrease in AR-V7 levels was observed.  5.3 Discussion Alternative splicing (AS) is an essential mechanism in gene expression, and has important clinical implications as it represents a mechanism of resistance to therapy. The heterogeneous nuclear ribonucleoprotein A1 (hnRNP A1) is a versatile RNA-binding protein playing a critical role in alternative pre-mRNA splicing regulation in cancer. Emerging data have implicated hnRNP A1 as a central player in a splicing regulatory circuit involving its direct transcriptional 206 control by c-Myc oncoprotein and the production of the constitutively active ligand-independent alternative splice variant of androgen receptor, AR-V7, which promotes enzalutamide-resistance. As there is an urgent need for effective CRPC drugs, targeting hnRNP A1 could serve a dual purpose of preventing AR-V7 generation as well as reducing c-Myc transcriptional output.  This work provides first evidence that targeting the RNA splicing activity of hnRNP A1 with drug-like small molecule inhibitors discovered by CADD approaches is a promising strategy to reduce the levels of AR-V7 and to overcome enzalutamide-resistance in prostate cancer.  The inhibitors were developed to target the RNA-binding domain of hnRNP A1. By employing the CADD platform, a binding site was identified utilizing the X-ray of hnRNP A1 bound to its RNA recognition sequence. Large-scale virtual screening combined with experimental validation yielded a number of hits predicted to interfere with the most functionally relevant protein–RNA interactions within the site. Development of a first promising hit VPC-80021 by in silico medicinal chemistry and application of the most elaborate and accurate scoring functions, including relative free binding energies differences obtained from FEP+ MD simulations to prioritize estimated most potent derivatives for chemical synthesis and experimental validation came to a halt due to prohibitive costs of synthesis.  We thus reported VPC-80051 as the first drug-like inhibitor of hnRNP A1 splicing activity discovered to date by using a computer-aided drug discovery approach targeting hnRNP A1 binding to RNA. 375 Unlike the literature-reported hnRNP A1 inhibitor quercetin, a well-known aggregator376 and containing the catechol_A moiety, one of the worst offenders enlisted in PAINS (Pan-Assay Interference Compounds)377, annotated in more than 46 catalogs for promiscuous binding and various inhibitory activities against multiple targets378 and as such 207 rejected by FAF-Drugs4 ADME-Tox filtering tool299 for inclusion in any drug discovery campaign, VPC-80051 contains no PAINS moieties and has no documented experimental activities against any protein target. Moreover, the quantitative estimate of druglikeness (QED) score336 of VPC-80051 as calculated by FAF-QED web service379 is 0.78 compared to 0.5 of quercetin. On a scale from 0 to 1, only compounds with a QED score above 0.65 (i.e., the median score for the current orally bioavailable drugs) are considered as having a desirable drug likeliness profile336. On the binding mode of VPC-80051 to the hnRNP A1 identified pocket, good agreement was obtained between various in silico techniques, including rigid docking with consensus scoring and flexible MD simulations with both implicit and explicit solvent. Furthermore, VPC-80051 showed evidenced ability to bind directly to the UP1 RNA-binding domain of hnRNP A1 in vitro and to alter hnRNP A1 alternative splicing activity in castrate-resistant 22Rv1 cells by reducing expression of the AR-V7 splice variant that promotes drug-resistance in CRPC.  While exploratory, this study provides a foundation for future computational efforts that, when combined with rigorous biological validation, may lead to the discovery of novel, more potent and selective hnRNP A1 inhibitors. Combination studies with current anti-AR drugs and/or c-Myc inhibitors may provide synergistic or additive responses, in reducing AR-V7 levels and resenzitizing cells to enzalutamide. Targeting the c-Myc/hnRNP A1/AR-V7 axis may be a viable strategy for treatment of CRPC. Targeting alternative splicing may offer future opportunities for development of next generation cancer-specific therapeutics.  208 Chapter 6: Conclusion 6.1 Summary of the Study Significant progress has been made in recent years in the treatment of prostate cancer (the second leading cause of cancer-related death in men) by adoption of novel AR inhibitors both in hormone-sensitive and castrate-resistant settings. However, despite the prolonged survival benefit for treatment-responsive CRPC patients, development of resistance is common. Moreover, administration of increasingly potent anti-androgen therapy eventually leads to neuroendocrine transdifferentiation resulting in the most aggressive AR-independent neuroendocrine phenotype of prostate cancer (NEPC). NEPC has an extremely poor prognosis and its only treatment option relies on decades old chemotherapy, which carries a short-lived response at the cost of significant toxicity. Overall, the current therapeutic landscape of prostate cancer solely directed at targeting AR is insufficient. As CRPC and NEPC are currently incurable, novel strategies are required to meet the urgent clinical need for management of these rapidly lethal forms of the disease. Encoded by major oncogenes, Myc transcription factors have been for decades the most sought after drug targets in several cancer types. Deregulation of Myc has been linked to nearly all human cancers making this bona fide oncogenic transcription factor a high-value target for therapeutic intervention. Successful targeting the oncogenic transformation of Myc with small molecules, although considered high-risk, represents the high-reward opportunity long awaited to advance prostate cancer treatment.  In CRPC and NEPC, the exacerbated expression of c-Myc and N-Myc, respectively, plays critical roles in disease progression and treatment-resistance. c-Myc has been demonstrated to promote AR gene transcription and enhance the stability of both AR and AR-V7 proteins in 209 CRPC cells and PDX models. Importantly, c-Myc knockdown by shRNA as well as inhibition with 10058-F4, a c-Myc small molecule prototype inhibitor, alleviated enzalutamide-resistance. N-Myc, on the other hand, is a primary inducer of NEPC.  The selective disruption of protein-protein or, importantly, protein-nucleic acid (i.e. DNA, RNA) interactions is a pinnacle of cancer therapy yet is a challenging endeavor especially when targeting intrinsically disordered proteins or their complexes. As best illustrated by Myc, its inherent disorder in the dimerization and DNA binding regions and lack of any obvious targetable pockets on its unstructured monomeric form, hampered the application of conventional structure-based drug design approaches.  Targeting the functionally mandatory Myc-Max complex proved similarly challenging as the Myc-Max/DNA interface is very large and surface exposed. Moreover, the lack of high-quality structures of the Myc-Max complex ligated with small molecules to evidence key protein interacting residues and serve as guides for rational drug design further restrained therapeutic development efforts. The majority of literature reported small molecule prototype inhibitors, that bind and alter the disordered Myc monomer and subsequently block Myc-Max dimerization, as well as the few Myc-Max/DNA disruptors, emerged mainly from high-throughput screens of limited chemical libraries unlikely to contain clinically optimized structures with only three Myc inhibitors discovered through computational approaches. Experimental evidence of their binding sites on the disordered Myc monomer is limited to a handful of inhibitors. While these small molecule inhibitors affected growth in cancer-specific cell lines with few showing effectiveness in animal models, they generally demonstrated suboptimal in vivo safety and efficacy profiles. As such, for many years, the pessimistical view in the cancer research field was that Myc is “undruggable”. 210 Application of computer-aided drug discovery methods involving novel in silico binding sites identification, large-scale virtual screening to maximize the throughput of structure-based drug design, hit identification and lead optimization techniques, both structure- and ligand-based, and ADMET profiling, represents the novel accelerated strategy employed in this work to target Myc-Max transforming activity in advanced prostate cancer. The strategy led to the discovery of novel small molecule inhibitors of both c-Myc-Max and N-Myc-Max function. Their development involved specific disruption of protein-nucleic acid interactions by targeting their DNA-binding domain.  On Myc-Max, in silico techniques led to the identification of a novel druggable site at the dimer/DNA interface, which is unique in terms of amino acid composition, size and contiguity relative to the few binding regions reported in the literature on the Myc monomer alone, or on the Mad-Max interface with DNA. Targeting both c-Myc-Max and N-Myc-Max complexes at the equivalent DBD sites, all modeled in silico, with same inhibitors developed by CADD techniques described in this work is unprecedented. This is the first computational study that employed in silico screening of millions of chemicals that when combined with pharmacophore modeling, chemical similarity searches, MD simulations and ADMET predictions yielded in a relative short time the promising Myc-Max lead inhibitor VPC-70551 that disrupts Myc-Max binding to DNA and is more potent and metabolically stable than existing compounds from literature. Among the spectrum of target genes upregulated by oncogenic levels of Myc is the hnRNP A1 splicing factor with similar transforming activity observed in many cancers. In CRPC, overexpressed hnRNP A1 has been shown to selectively stimulate AR-V7 alternative splicing while its knockdown with siRNA suppressed AR-V7 levels and growth of CRPC cells. 211 Moreover, its inhibition with a natural occurring flavonoid, quercetin, resensitized enzalutamide-resistant cell lines and mouse xenografts to enzalutamide in vivo. Thus, hnRNP A1 represents a prospective target in advanced prostate cancer. The same CADD strategy was employed in this work to develop novel small molecule inhibitors that specifically disrupt the interactions of hnRNP A1 with RNA. On hnRNP A1, to our knowledge, VPC-80051 is the first drug-like small molecule inhibitor discovered to date by using a computer-aided drug discovery approach targeting hnRNP A1 binding to RNA. VPC-80051 may serve as a basis for future structure- and ligand-based development of more potent and selective small molecule inhibitors of hnRNP A1–RNA interactions aimed at altering the production of cancer-specific alternative splice isoforms. Protein-nucleic acid recognition is structurally determined by specific atomic contacts with H-bonds being statistically the most frequent interactions in crystallographic data, followed by van der Waals, hydrophobic, and electrostatic interactions. DNA and RNA behave differently in recognizing their protein targets. While most protein-DNA interactions involve phosphates atoms, protein-RNA interactions involve most frequently base edge and sugar atoms.380 The research described in this thesis showed that the novel developed inhibitors are able to compete with the native DNA/RNA ligands of their respective protein targets by taking over such interactions. VPC-70551 mimics the H-bonds formed by Arg214, Lys939, and Arg239 Myc-Max residues with the phosphate groups of the DNA backbone. Similarly, VPC-80051 takes over the H-bonds formed by Val90 of hnRNP A1 with the edges of the two bases of the RNA recognition sequence. The discontinued VPC-80021 and medchem derivatives additionally took over the H-bonding interaction of Arg88, Gln12 and Lys15 of hnRNP A1 with the RNA base edges. Hydrophobic interactions are furthermore required for binding of small molecule inhibitors to 212 their respective sites at protein-nucleic acid interfaces. To substantiate the results obtained from MD simulations and in vitro binding assays, mutations of key binding residues, NMR spectroscopy or X-ray crystallography structural studies may unequivocally prove the binding mode of discovered inhibitors to the intended targeted sites. As an endnote, although important progress has been made recently in targeting the challenging complexes of intrinsically disordered proteins, with Myc-Max standing as a primary example, the success of future structure-based drug discovery efforts relies on emerging technologies that are capable of solving protein structures complexed with small molecule inhibitors at atomic resolution. One such promising technique is cryo-Electron Microscopy (cryo-EM), which may enable the structure determination of protein targets intractable to X-ray analysis as well as helping to identify key features of protein-drug interactions at high-resolution.381 Armed with such knowledge, a rational approach to Myc-Max inhibition, combining the most advanced and accurate computational techniques for drug discovery, medicinal chemistry, appropriate formulations, suitable delivery systems, and powerful and reproducible preclinical studies, is highly-likely to provide in the near future the much desired drug candidate to be used alone or in combination for treatment of Myc-driven cancers, and to irrevocably change the paradigm that Myc is “undruggable”.  6.2 Future Directions In this study, we presented the successful application of computational drug design methods integrated with biological validation for the discovery and development of novel small molecule inhibitors targeting the potent transforming activity of Myc-Max in advanced prostate cancer. Specifically, direct targeting of c- and N-Myc-Max complexes and Myc-upregulated 213 hnRNP A1 splicing factor by disruption of protein-nucleic acid interactions, at novel identified sites within their nucleic acid-binding domains. Albeit promising leads, future work is required to transform these small molecules into drug candidates for treatment of CRPC and NEPC. Achieving the optimum combination of potency and stability of a drug-like candidate is a key component of the drug discovery process. The current lead inhibitor VPC-70551 demonstrated a 4 μM potency in inhibiting the transcriptional activity of c- and N-Myc-Max complexes and good in vitro metabolic stability (t1/2 = 140 min), but still needs to be optimized for improved potency and enhanced microsomal stability. SAR by catalog (i.e. ENAMINE-REAL) is currently underway to identify more potent and stable analogs. Pharmacokinetics and anti-tumor growth studies will follow. Dependent on the suitability of their pharmacokinetic profiles, additional medicinal chemistry efforts may be required to develop drug candidates for future preclinical in vivo efficacy and safety studies. Oral bioavailability is the ultimate goal in developing any drug candidate. Therefore, studies directed towards developing oral formulations and dose schedules of these compounds represents an important aspect of future investigations.  Albeit the developed Myc-Max inhibitors could make a major clinical impact when delivered as single agents, their possible synergistic effects when used in combination with classical anti-androgens, in particular enzalutamide, could be evaluated in vitro and in vivo models for their potential to maximize the inhibitory effect and reduce the toxicity related to each treatment. Such combination therapies may ultimately overcome drug-resistance mechanisms and neuroendocrine differentiation.  As an alternative to Myc-Max direct inhibition described in this study, direct targeting of Myc transcription with G-quadruplex stabilizers at the NHEIII1 cis-element, although quite complex, is gaining tremendous speed and is a very promising strategy as it already yielded drug 214 candidates that entered clinical trials. “Drug-likeliness” and therapeutic selectivity remain nonetheless major concerns for future development efforts. Yet, the indirect strategy of targeting the protein-DNA interactions of NHEIII1/NM23-H2382 or FUSE/FBP383 complexes that regulate the expression of Myc itself at the promoter region, while therapeutically promising and approachable by CADD techniques, is currently underexplored. The FUSE/FBP system may be of particular interest for future investigations as preliminary in silico modeling and screening suggest. Targeting alternative splicing for therapeutic benefit in advanced prostate cancer is still in its infancy. Besides hnRNP A1, other splicing factors, including U2AF65, ASF/SF2368 and hnRNP F384, as well as the RNA-binding protein Sam68385, the transcription factor YB-1386, and the molecular chaperone HSP90387 have been shown to selectively regulate the production of the AR-V7 variant. As such, contingent to their druggability, targeting these proteins may provide new therapeutic opportunities in castrate-resistant settings. On hnRNP A1, this study showed that disruption of hnRNP A1-RNA interactions is a feasible strategy to reduce the levels of AR-V7. It would be worth exploring the effects of hnRNP A1 inhibition on reduction of the PKM2 cancer-specific isoform. Moreover, disruption of hnRNP A1 binding to internal ribosome entry sites (IRES) by application of CADD could be a novel strategy previously unpursued for blocking the translation of Myc protein itself. The role of hnRNP A1 in NEPC is, to our knowledge, unknown, albeit in high-risk metastatic neuroblastoma N-Myc controls the splicing pattern of hnRNP A1 by direct binding to the hnRNP A1 promoter region, thus regulating its expression and that of the downstream target PKM2 isoform357. This is rather similar to the c-Myc role in upregulating hnRNP A1 and its subsequent downstream functions (i.e. AR-V7, PKM2, ∆Max isoforms) in CRPC. 215 On NEPC, VPC-70551 is a promising lead that pending further optimization and preclinical development may result in a first-in-class N-Myc-Max drug. Dual inhibition with improved VPC-70551-based and AURKA inhibitors may provide a viable solution for NEPC treatment. Targeting SRRM4 splicing factor, a driver of NEPC388, with small molecules is another big challenge for CADD due to the highly disordered nature of the protein worth pursuing in the near future. 216 Bibliography 1. http://www.prostatecancer.ca/Prostate-Cancer/About-Prostate-Cancer/Statistics. 2. Fradet, Y., Klotz, L., Trachtenberg, J. & Zlotta, A. The burden of prostate cancer in Canada. Canadian Urological Association journal = Journal de l'Association des urologues du Canada 3, S92-S100 (2009). 3. De Marzo, A.M., Haffner, M.C., Lotan, T.L., Yegnasubramanian, S. & Nelson, W.G. Premalignancy in Prostate Cancer: Rethinking What we Know. Cancer prevention research 9, 648-656 (2016). 4. Huggins, C. & Hodges, C.V. Studies on prostatic cancer. I. The effect of castration, of estrogen and androgen injection on serum phosphatases in metastatic carcinoma of the prostate. CA: a cancer journal for clinicians 22, 232-240 (1972). 5. HUGGINS, C., STEVENS, R.E., Jr. & HODGES, C.V. STUDIES ON PROSTATIC CANCER: II. THE EFFECTS OF CASTRATION ON ADVANCED CARCINOMA OF THE PROSTATE GLAND. JAMA Surgery 43, 209-223 (1941). 6. Goldenberg, S.L.P., T.; Chi, K.N.;. The Intelligent Patient Guide to Prostate Cancer; 4th ed. edition; 2014. 7. Litwin, M.S. & Tan, H.J. The Diagnosis and Treatment of Prostate Cancer: A Review. Jama 317, 2532-2542 (2017). 8. Mohler, J.L., et al. Prostate Cancer, Version 1.2016. Journal of the National Comprehensive Cancer Network : JNCCN 14, 19-30 (2016). 9. Scher, H.I., et al. Design and end points of clinical trials for patients with progressive prostate cancer and castrate levels of testosterone: recommendations of the Prostate Cancer Clinical Trials Working Group. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 26, 1148-1159 (2008). 10. Dardenne, E., et al. N-Myc Induces an EZH2-Mediated Transcriptional Program Driving Neuroendocrine Prostate Cancer. Cancer cell 30, 563-577 (2016). 11. Perlmutter, M.A. & Lepor, H. Androgen deprivation therapy in the treatment of advanced prostate cancer. Reviews in urology 9 Suppl 1, S3-8 (2007). 12. Harris, W.P., Mostaghel, E.A., Nelson, P.S. & Montgomery, B. Androgen deprivation therapy: progress in understanding mechanisms of resistance and optimizing androgen depletion. Nature clinical practice. Urology 6, 76-85 (2009). 13. Crawford, E.D. & Rove, K.O. Advanced prostate cancer: therapeutic sequencing, outcomes, and cost implications. The American journal of managed care 18, SP250-253 (2012). 14. Ingrosso, G., et al. Current therapeutic options in metastatic castration-resistant prostate cancer. Seminars in oncology 45, 303-315 (2018). 15. Tran, C., et al. Development of a second-generation antiandrogen for treatment of advanced prostate cancer. Science 324, 787-790 (2009). 16. Scher, H.I., et al. Increased survival with enzalutamide in prostate cancer after chemotherapy. The New England journal of medicine 367, 1187-1197 (2012). 17. Takayama, K. & Inoue, S. Transcriptional network of androgen receptor in prostate cancer progression. International journal of urology : official journal of the Japanese Urological Association 20, 756-768 (2013). 18. Gelmann, E. Molecular biology of the androgen receptor. Journal of Clinical Oncology 20, 3001-3015 (2002). 217 19. Lallous, N., Dalal, K., Cherkasov, A. & Rennie, P.S. Targeting alternative sites on the androgen receptor to treat castration-resistant prostate cancer. International journal of molecular sciences 14, 12496-12519 (2013). 20. Prescott, J. & Coetzee, G.A. Molecular chaperones throughout the life cycle of the androgen receptor. Cancer letters 231, 12-19 (2006). 21. Heinlein, C.A. & Chang, C. Androgen receptor (AR) coregulators: an overview. Endocrine reviews 23, 175-200 (2002). 22. Vander Ark, A., Cao, J. & Li, X. Mechanisms and Approaches for Overcoming Enzalutamide Resistance in Prostate Cancer. Frontiers in oncology 8, 180 (2018). 23. Claessens, F., et al. Emerging mechanisms of enzalutamide resistance in prostate cancer. Nature Reviews Urology 11, 712 (2014). 24. Grasso, C.S., et al. The mutational landscape of lethal castration-resistant prostate cancer. Nature 487, 239-243 (2012). 25. Robinson, D., et al. Integrative clinical genomics of advanced prostate cancer. Cell 161, 1215-1228 (2015). 26. Gottlieb, B., Beitel, L.K., Nadarajah, A., Paliouras, M. & Trifiro, M. The androgen receptor gene mutations database: 2012 update. Human mutation 33, 887-894 (2012). 27. Lorente, D., et al. Switching and withdrawing hormonal agents for castration-resistant prostate cancer. Nature reviews. Urology 12, 37-47 (2015). 28. Forootan, S.S., Hussain, S., Aachi, V., Foster, C.S., Ke, Y. . Molecular Mechanisms Involved in the Transition of Prostate Cancer Cells from Androgen Dependant to Castration Resistant State. J Androl Gynaecol 2, 9 (2014). 29. Jernberg, E., Bergh, A. & Wikstrom, P. Clinical relevance of androgen receptor alterations in prostate cancer. Endocrine connections 6, R146-R161 (2017). 30. Lallous, N., et al. Functional analysis of androgen receptor mutations that confer anti-androgen resistance identified in circulating cell-free DNA from prostate cancer patients. Genome biology 17, 10 (2016). 31. Paul, N., et al. Cheminformatics Modeling of Adverse Drug Responses by Clinically Relevant Mutants of Human Androgen Receptor. Journal of chemical information and modeling 56, 2507-2516 (2016). 32. Watson, P.A., Arora, V.K. & Sawyers, C.L. Emerging mechanisms of resistance to androgen receptor inhibitors in prostate cancer. Nature reviews. Cancer 15, 701-711 (2015). 33. Cao, S., Zhan, Y. & Dong, Y. Emerging data on androgen receptor splice variants in prostate cancer. Endocrine-related cancer 23, T199-T210 (2016). 34. Antonarakis, E.S., et al. AR-V7 and resistance to enzalutamide and abiraterone in prostate cancer. The New England journal of medicine 371, 1028-1038 (2014). 35. Antonarakis, E.S., et al. Clinical Significance of Androgen Receptor Splice Variant-7 mRNA Detection in Circulating Tumor Cells of Men With Metastatic Castration-Resistant Prostate Cancer Treated With First- and Second-Line Abiraterone and Enzalutamide. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 35, 2149-2156 (2017). 36. Arora, V.K., et al. Glucocorticoid receptor confers resistance to antiandrogens by bypassing androgen receptor blockade. Cell 155, 1309-1322 (2013). 37. Taylor, B.S., et al. Integrative genomic profiling of human prostate cancer. Cancer cell 18, 11-22 (2010). 218 38. Beltran, H., et al. Molecular characterization of neuroendocrine prostate cancer and identification of new drug targets. Cancer discovery 1, 487-495 (2011). 39. Aggarwal, R., Zhang, T., Small, E.J. & Armstrong, A.J. Neuroendocrine prostate cancer: subtypes, biology, and clinical outcomes. Journal of the National Comprehensive Cancer Network : JNCCN 12, 719-726 (2014). 40. Santoni, M., et al. Neuroendocrine differentiation in prostate cancer: novel morphological insights and future therapeutic perspectives. Biochimica et biophysica acta 1846, 630-637 (2014). 41. Epstein, J.I., et al. Proposed morphologic classification of prostate cancer with neuroendocrine differentiation. The American journal of surgical pathology 38, 756-767 (2014). 42. Wyatt, A.W. & Gleave, M.E. Targeting the adaptive molecular landscape of castration-resistant prostate cancer. EMBO molecular medicine 7, 878-894 (2015). 43. Akamatsu, S., Inoue, T., Ogawa, O. & Gleave, M.E. Clinical and molecular features of treatment-related neuroendocrine prostate cancer. International journal of urology : official journal of the Japanese Urological Association 25, 345-351 (2018). 44. Carabet, L.A., Rennie, P.S. & Cherkasov, A. Therapeutic Inhibition of Myc in Cancer. Structural Bases and Computer-Aided Drug Discovery Approaches. International journal of molecular sciences 20(2018). 45. Kalkat, M., et al. MYC Deregulation in Primary Human Cancers. Genes 8(2017). 46. Dalla-Favera, R., et al. Human c-myc onc gene is located on the region of chromosome 8 that is translocated in Burkitt lymphoma cells. Proceedings of the National Academy of Sciences of the United States of America 79, 7824-7827 (1982). 47. Varmus, H.E. The molecular genetics of cellular oncogenes. Annual review of genetics 18, 553-612 (1984). 48. Schwab, M., et al. Amplified DNA with limited homology to myc cellular oncogene is shared by human neuroblastoma cell lines and a neuroblastoma tumour. Nature 305, 245-248 (1983). 49. Schwab, M., et al. Enhanced expression of the human gene N-myc consequent to amplification of DNA may contribute to malignant progression of neuroblastoma. Proceedings of the National Academy of Sciences of the United States of America 81, 4940-4944 (1984). 50. Nau, M.M., et al. L-myc, a new myc-related gene amplified and expressed in human small cell lung cancer. Nature 318, 69-73 (1985). 51. Dang, C.V. MYC on the path to cancer. Cell 149, 22-35 (2012). 52. Dang, C.V., et al. Function of the c-Myc oncogenic transcription factor. Experimental cell research 253, 63-77 (1999). 53. Dang, C.V. MYC, metabolism, cell growth, and tumorigenesis. Cold Spring Harbor perspectives in medicine 3(2013). 54. McMahon, S.B. MYC and the control of apoptosis. Cold Spring Harbor perspectives in medicine 4, a014407 (2014). 55. Stine, Z.E., Walton, Z.E., Altman, B.J., Hsieh, A.L. & Dang, C.V. MYC, Metabolism, and Cancer. Cancer discovery 5, 1024-1039 (2015). 56. Sears, R.C. The life cycle of C-myc: from synthesis to degradation. Cell cycle 3, 1133-1137 (2004). 219 57. Hanahan, D. & Weinberg, R.A. Hallmarks of cancer: the next generation. Cell 144, 646-674 (2011). 58. Cerami, E., et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer discovery 2, 401-404 (2012). 59. Gao, J., et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Science signaling 6, pl1 (2013). 60. Koh, C.M., et al. MYC and Prostate Cancer. Genes & cancer 1, 617-628 (2010). 61. Boutros, P.C., et al. Spatial genomic heterogeneity within localized, multifocal prostate cancer. Nature genetics 47, 736-745 (2015). 62. Beltran, H., et al. Aggressive variants of castration-resistant prostate cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 20, 2846-2850 (2014). 63. Beltran, H., et al. Divergent clonal evolution of castration-resistant neuroendocrine prostate cancer. Nature medicine 22, 298-305 (2016). 64. Bernard, D., Pourtier-Manzanedo, A., Gil, J. & Beach, D.H. Myc confers androgen-independent prostate cancer cell growth. The Journal of clinical investigation 112, 1724-1731 (2003). 65. Kim, J.H., et al. Integrative analysis of genomic aberrations associated with prostate cancer progression. Cancer research 67, 8229-8239 (2007). 66. Bai, S., et al. A positive role of c-Myc in regulating androgen receptor and its splice variants in prostate cancer. Oncogene (2019). 67. Nadiminty, N., et al. NF-kappaB2/p52:c-Myc:hnRNPA1 Pathway Regulates Expression of Androgen Receptor Splice Variants and Enzalutamide Sensitivity in Prostate Cancer. Molecular cancer therapeutics 14, 1884-1895 (2015). 68. Hu, R., et al. Distinct transcriptional programs mediated by the ligand-dependent full-length androgen receptor and its splice variants in castration-resistant prostate cancer. Cancer research 72, 3457-3462 (2012). 69. Conacci-Sorrell, M., McFerrin, L. & Eisenman, R.N. An overview of MYC and its interactome. Cold Spring Harbor perspectives in medicine 4, a014357 (2014). 70. Meyer, N. & Penn, L.Z. Reflecting on 25 years with MYC. Nature reviews. Cancer 8, 976-990 (2008). 71. Wierstra, I. & Alves, J. The c-myc promoter: still MysterY and challenge. Advances in cancer research 99, 113-333 (2008). 72. Gonzalez, V. & Hurley, L.H. The c-MYC NHE III(1): function and regulation. Annual review of pharmacology and toxicology 50, 111-129 (2010). 73. Brooks, T.A. & Hurley, L.H. Targeting MYC Expression through G-Quadruplexes. Genes & cancer 1, 641-649 (2010). 74. Yang, D. & Hurley, L.H. Structure of the biologically relevant G-quadruplex in the c-MYC promoter. Nucleosides, nucleotides & nucleic acids 25, 951-968 (2006). 75. Neidle, S. Quadruplex nucleic acids as targets for anticancer therapeutics. Nature Reviews Chemistry 1, 0041 (2017). 76. Ambrus, A., Chen, D., Dai, J., Jones, R.A. & Yang, D. Solution structure of the biologically relevant G-quadruplex element in the human c-MYC promoter. Implications for G-quadruplex stabilization. Biochemistry 44, 2048-2058 (2005). 220 77. Mathad, R.I., Hatzakis, E., Dai, J. & Yang, D. c-MYC promoter G-quadruplex formed at the 5'-end of NHE III1 element: insights into biological relevance and parallel-stranded G-quadruplex stability. Nucleic acids research 39, 9023-9033 (2011). 78. Brooks, T.A. & Hurley, L.H. The role of supercoiling in transcriptional control of MYC and its importance in molecular therapeutics. Nature reviews. Cancer 9, 849-861 (2009). 79. Jones, T.R. & Cole, M.D. Rapid cytoplasmic turnover of c-myc mRNA: requirement of the 3' untranslated sequences. Molecular and cellular biology 7, 4513-4521 (1987). 80. Brewer, G. & Ross, J. Poly(A) shortening and degradation of the 3' A+U-rich sequences of human c-myc mRNA in a cell-free system. Molecular and cellular biology 8, 1697-1708 (1988). 81. Bernstein, P.L., Herrick, D.J., Prokipcak, R.D. & Ross, J. Control of c-myc mRNA half-life in vitro by a protein capable of binding to a coding region stability determinant. Genes & development 6, 642-654 (1992). 82. Ioannidis, P., et al. CRD-BP: a c-Myc mRNA stabilizing protein with an oncofetal pattern of expression. Anticancer research 23, 2179-2183 (2003). 83. Bahram, F., von der Lehr, N., Cetinkaya, C. & Larsson, L.G. c-Myc hot spot mutations in lymphomas result in inefficient ubiquitination and decreased proteasome-mediated turnover. Blood 95, 2104-2110 (2000). 84. Gregory, M.A. & Hann, S.R. c-Myc proteolysis by the ubiquitin-proteasome pathway: stabilization of c-Myc in Burkitt's lymphoma cells. Molecular and cellular biology 20, 2423-2435 (2000). 85. Hemann, M.T., et al. Evasion of the p53 tumour surveillance network by tumour-derived MYC mutants. Nature 436, 807-811 (2005). 86. Akhoondi, S., et al. FBXW7/hCDC4 is a general tumor suppressor in human cancer. Cancer research 67, 9006-9012 (2007). 87. McMahon, S.B., Van Buskirk, H.A., Dugan, K.A., Copeland, T.D. & Cole, M.D. The novel ATM-related protein TRRAP is an essential cofactor for the c-Myc and E2F oncoproteins. Cell 94, 363-374 (1998). 88. Itzen, F., Greifenberg, A.K., Bosken, C.A. & Geyer, M. Brd4 activates P-TEFb for RNA polymerase II CTD phosphorylation. Nucleic acids research 42, 7577-7590 (2014). 89. Kim, S.Y., Herbst, A., Tworkowski, K.A., Salghetti, S.E. & Tansey, W.P. Skp2 regulates Myc protein stability and activity. Molecular cell 11, 1177-1188 (2003). 90. von der Lehr, N., et al. The F-box protein Skp2 participates in c-Myc proteosomal degradation and acts as a cofactor for c-Myc-regulated transcription. Molecular cell 11, 1189-1200 (2003). 91. Farrell, A.S. & Sears, R.C. MYC degradation. Cold Spring Harbor perspectives in medicine 4(2014). 92. Garcia-Sanz, P., et al. Sin3b interacts with Myc and decreases Myc levels. The Journal of biological chemistry 289, 22221-22236 (2014). 93. Kurland, J.F. & Tansey, W.P. Myc-mediated transcriptional repression by recruitment of histone deacetylase. Cancer research 68, 3624-3629 (2008). 94. Thomas, L.R., et al. Interaction with WDR5 promotes target gene recognition and tumorigenesis by MYC. Molecular cell 58, 440-452 (2015). 95. Hydbring, P., Castell, A. & Larsson, L.G. MYC Modulation around the CDK2/p27/SKP2 Axis. Genes 8(2017). 221 96. Conacci-Sorrell, M., Ngouenet, C. & Eisenman, R.N. Myc-nick: a cytoplasmic cleavage product of Myc that promotes alpha-tubulin acetylation and cell differentiation. Cell 142, 480-493 (2010). 97. Conacci-Sorrell, M., Ngouenet, C., Anderson, S., Brabletz, T. & Eisenman, R.N. Stress-induced cleavage of Myc promotes cancer cell survival. Genes & development 28, 689-707 (2014). 98. Anderson, S., et al. MYC-nick promotes cell migration by inducing fascin expression and Cdc42 activation. Proceedings of the National Academy of Sciences of the United States of America 113, E5481-5490 (2016). 99. Dang, C.V. & Lee, W.M. Identification of the human c-myc protein nuclear translocation signal. Molecular and cellular biology 8, 4048-4054 (1988). 100. Grandori, C., Cowley, S.M., James, L.P. & Eisenman, R.N. The Myc/Max/Mad network and the transcriptional control of cell behavior. Annual review of cell and developmental biology 16, 653-699 (2000). 101. Baudino, T.A. & Cleveland, J.L. The Max Network Gone Mad. Molecular and cellular biology 21, 691 (2001). 102. Zhang, Q., et al. Domain-specific c-Myc ubiquitylation controls c-Myc transcriptional and apoptotic activity. Proceedings of the National Academy of Sciences of the United States of America 110, 978-983 (2013). 103. Tu, W.B., et al. Myc and its interactors take shape. Biochimica et biophysica acta 1849, 469-483 (2015). 104. Hann, S.R. MYC cofactors: molecular switches controlling diverse biological outcomes. Cold Spring Harbor perspectives in medicine 4, a014399 (2014). 105. Luscher, B. Function and regulation of the transcription factors of the Myc/Max/Mad network. Gene 277, 1-14 (2001). 106. Carroll, P.A., Freie, B.W., Mathsyaraja, H. & Eisenman, R.N. The MYC transcription factor network: balancing metabolism, proliferation and oncogenesis. Frontiers of medicine 12, 412-425 (2018). 107. Eilers, A.L., Billin, A.N., Liu, J. & Ayer, D.E. A 13-amino acid amphipathic alpha-helix is required for the functional interaction between the transcriptional repressor Mad1 and mSin3A. The Journal of biological chemistry 274, 32750-32756 (1999). 108. Sun, Q.Y., et al. Ordering of mutations in acute myeloid leukemia with partial tandem duplication of MLL (MLL-PTD). Leukemia 31, 1-10 (2017). 109. Nilsson, J.A., et al. Mnt loss triggers Myc transcription targets, proliferation, apoptosis, and transformation. Molecular and cellular biology 24, 1560-1569 (2004). 110. Hurlin, P.J., et al. Deletion of Mnt leads to disrupted cell cycle control and tumorigenesis. The EMBO journal 22, 4584-4596 (2003). 111. Popov, N., Wahlstrom, T., Hurlin, P.J. & Henriksson, M. Mnt transcriptional repressor is functionally regulated during cell cycle progression. Oncogene 24, 8326-8337 (2005). 112. Kretzner, L., Blackwood, E.M. & Eisenman, R.N. Myc and Max proteins possess distinct transcriptional activities. Nature 359, 426 (1992). 113. Yin, X., Grove, L. & Prochownik, E.V. Lack of transcriptional repression by max homodimers. Oncogene 16, 2629 (1998). 114. Cogliati, T., Dunn, B.K., Bar-Ner, M., Cultraro, C.M. & Segal, S. Transfected wild-type and mutant max regulate cell growth and differentiation of murine erythroleukemia cells. Oncogene 8, 1263-1268 (1993). 222 115. Lindeman, G.J., Harris, A.W., Bath, M.L., Eisenman, R.N. & Adams, J.M. Overexpressed max is not oncogenic and attenuates myc-induced lymphoproliferation and lymphomagenesis in transgenic mice. Oncogene 10, 1013-1017 (1995). 116. Yuza, Y., Kawakami, M., Takagi, K., Yamazaki, Y. & Urashima, M. Max protein expression is associated with survival of children with lymphoblastic lymphoma. Pediatrics international : official journal of the Japan Pediatric Society 41, 637-640 (1999). 117. Blackwood, E.M. & Eisenman, R.N. Max: a helix-loop-helix zipper protein that forms a sequence-specific DNA-binding complex with Myc. Science 251, 1211-1217 (1991). 118. Makela, T.P., Koskinen, P.J., Vastrik, I. & Alitalo, K. Alternative forms of Max as enhancers or suppressors of Myc-ras cotransformation. Science 256, 373-377 (1992). 119. O'Shea, J.M. & Ayer, D.E. Coordination of nutrient availability and utilization by MAX- and MLX-centered transcription networks. Cold Spring Harbor perspectives in medicine 3, a014258 (2013). 120. Grinberg, A.V., Hu, C.D. & Kerppola, T.K. Visualization of Myc/Max/Mad family dimers and the competition for dimerization in living cells. Molecular and cellular biology 24, 4294-4308 (2004). 121. Nair, S.K. & Burley, S.K. Structural Aspects of Interactions Within the Myc/Max/Mad Network. in The Myc/Max/Mad Transcription Factor Network (ed. Eisenman, R.N.) 123-143 (Springer Berlin Heidelberg, Berlin, Heidelberg, 2006). 122. Ferre-D'Amare, A.R., Prendergast, G.C., Ziff, E.B. & Burley, S.K. Recognition by Max of its cognate DNA through a dimeric b/HLH/Z domain. Nature 363, 38-45 (1993). 123. Nair, S.K. & Burley, S.K. X-ray structures of Myc-Max and Mad-Max recognizing DNA. Molecular bases of regulation by proto-oncogenic transcription factors. Cell 112, 193-205 (2003). 124. O'Shea, E.K., Klemm, J.D., Kim, P.S. & Alber, T. X-ray structure of the GCN4 leucine zipper, a two-stranded, parallel coiled coil. Science 254, 539-544 (1991). 125. Davis, L.J. & Halazonetis, T.D. Both the helix-loop-helix and the leucine zipper motifs of c-Myc contribute to its dimerization specificity with Max. Oncogene 8, 125-132 (1993). 126. Fisher, F. & Goding, C.R. Single amino acid substitutions alter helix-loop-helix protein specificity for bases flanking the core CANNTG motif. The EMBO journal 11, 4103-4109 (1992). 127. Dang, C.V., Dolde, C., Gillison, M.L. & Kato, G.J. Discrimination between related DNA sites by a single amino acid residue of Myc-related basic-helix-loop-helix proteins. Proceedings of the National Academy of Sciences of the United States of America 89, 599-602 (1992). 128. Jung, L.A., et al. OmoMYC blunts promoter invasion by oncogenic MYC to inhibit gene expression characteristic of MYC-dependent tumors. Oncogene 36, 1911-1924 (2017). 129. Soucek, L., et al. Design and properties of a Myc derivative that efficiently homodimerizes. Oncogene 17, 2463-2472 (1998). 130. Savino, M., et al. The action mechanism of the Myc inhibitor termed Omomyc may give clues on how to target Myc for cancer therapy. PloS one 6, e22284 (2011). 131. Horiuchi, D., Anderton, B. & Goga, A. Taking on challenging targets: making MYC druggable. American Society of Clinical Oncology educational book. American Society of Clinical Oncology. Meeting, e497-502 (2014). 223 132. Soucek, L., et al. Modelling Myc inhibition as a cancer therapy. Nature 455, 679-683 (2008). 133. Soucek, L., et al. Inhibition of Myc family proteins eradicates KRas-driven lung cancer in mice. Genes & development 27, 504-513 (2013). 134. Jin, F., Yu, C., Lai, L. & Liu, Z. Ligand clouds around protein clouds: a scenario of ligand binding with intrinsically disordered proteins. PLoS computational biology 9, e1003249 (2013). 135. Yu, C., et al. Structure-based Inhibitor Design for the Intrinsically Disordered Protein c-Myc. Scientific reports 6, 22298 (2016). 136. Amati, B., et al. Oncogenic activity of the c-Myc protein requires dimerization with Max. Cell 72, 233-245 (1993). 137. Blackwell, T.K., et al. Binding of myc proteins to canonical and noncanonical DNA sequences. Molecular and cellular biology 13, 5216-5224 (1993). 138. Whitfield, J.R., Beaulieu, M.E. & Soucek, L. Strategies to Inhibit Myc and Their Clinical Applicability. Frontiers in cell and developmental biology 5, 10 (2017). 139. Koh, C.M., Sabo, A. & Guccione, E. Targeting MYC in cancer therapy: RNA processing offers new opportunities. BioEssays : news and reviews in molecular, cellular and developmental biology 38, 266-275 (2016). 140. Huang, M. & Weiss, W.A. Neuroblastoma and MYCN. Cold Spring Harbor perspectives in medicine 3, a014415 (2013). 141. Andrieu, G., Belkina, A.C. & Denis, G.V. Clinical trials for BET inhibitors run ahead of the science. Drug discovery today. Technologies 19, 45-50 (2016). 142. Shi, X., et al. JQ1: a novel potential therapeutic target. Die Pharmazie 73, 491-493 (2018). 143. https://www.cancer.gov/about-cancer/treatment/clinical-trials/intervention/bet-bromodomain-inhibitor-zen-3694 (last accessed Nov 1, 2018). 144. Berthon, C., et al. Bromodomain inhibitor OTX015 in patients with acute leukaemia: a dose-escalation, phase 1 study. The Lancet. Haematology 3, e186-195 (2016). 145. Shapiro, G.I., et al. Abstract A49: Clinically efficacy of the BET bromodomain inhibitor TEN-010 in an open-label substudy with patients with documented NUT-midline carcinoma (NMC). Molecular cancer therapeutics 14, A49 (2015). 146. Poole, C., et al. Targeting the MYC Oncogene in Burkitt Lymphoma through HSP90 Inhibition. Cancers 10(2018). 147. Bayliss, R., Burgess, S.G., Leen, E. & Richards, M.W. A moving target: structure and disorder in pursuit of Myc inhibitors. Biochemical Society transactions 45, 709-717 (2017). 148. Gustafson, W.C., et al. Drugging MYCN through an allosteric transition in Aurora kinase A. Cancer cell 26, 414-427 (2014). 149. Brockmann, M., et al. Small molecule inhibitors of aurora-a induce proteasomal degradation of N-myc in childhood neuroblastoma. Cancer cell 24, 75-89 (2013). 150. Lu, J.J., et al. Dihydroartemisinin accelerates c-MYC oncoprotein degradation and induces apoptosis in c-MYC-overexpressing tumor cells. Biochemical pharmacology 80, 22-30 (2010). 151. McKeown, M.R. & Bradner, J.E. Therapeutic strategies to inhibit MYC. Cold Spring Harbor perspectives in medicine 4(2014). 224 152. Wang, H., et al. Structurally diverse c-Myc inhibitors share a common mechanism of action involving ATP depletion. Oncotarget 6, 15857-15870 (2015). 153. Paulo, A., Castillo, C.C. & Neidle, S. Targeting Promoter Quadruplex Nucleic Acids for Cancer Therapy.  (2017). 154. Duarte, A.R., Cadoni, E., Ressurreicao, A.S., Moreira, R. & Paulo, A. Design of Modular G-quadruplex Ligands. ChemMedChem 13, 869-893 (2018). 155. Chen, B.J., Wu, Y.L., Tanaka, Y. & Zhang, W. Small molecules targeting c-Myc oncogene: promising anti-cancer therapeutics. International journal of biological sciences 10, 1084-1096 (2014). 156. Dexheimer, T.S., et al. NM23-H2 may play an indirect role in transcriptional activation of c-myc gene expression but does not cleave the nuclease hypersensitive element III(1). Molecular cancer therapeutics 8, 1363-1377 (2009). 157. Thakur, R.K., et al. Metastases suppressor NM23-H2 interaction with G-quadruplex DNA within c-MYC promoter nuclease hypersensitive element induces c-MYC expression. Nucleic acids research 37, 172-183 (2009). 158. Seenisamy, J., et al. The dynamic character of the G-quadruplex element in the c-MYC promoter and modification by TMPyP4. Journal of the American Chemical Society 126, 8702-8709 (2004). 159. Seenisamy, J., et al. Design and synthesis of an expanded porphyrin that has selectivity for the c-MYC G-quadruplex structure. Journal of the American Chemical Society 127, 2944-2959 (2005). 160. Ou, T.M., et al. Inhibition of cell proliferation by quindoline derivative (SYUIQ-05) through its preferential interaction with c-myc promoter G-quadruplex. Journal of medicinal chemistry 54, 5671-5679 (2011). 161. Wang, J., et al. A Pt(II)-Dip complex stabilizes parallel c-myc G-quadruplex. Chemical communications 49, 4758-4760 (2013). 162. Wu, P., et al. Stabilization of G-quadruplex DNA with platinum(II) Schiff base complexes: luminescent probe and down-regulation of c-myc oncogene expression. Chemistry 15, 13008-13021 (2009). 163. Zhang, J., et al. Recent progress and future potential for metal complexes as anticancer drugs targeting G-quadruplex DNA. Current medicinal chemistry 19, 2957-2975 (2012). 164. https://clinicaltrials.gov/ct2/show/NCT00780663; (last accessed, May 26, 2019). 165. Duan, W., et al. Design and synthesis of fluoroquinophenoxazines that interact with human telomeric G-quadruplexes and their biological effects. Molecular cancer therapeutics 1, 103-120 (2001). 166. Drygin, D., et al. Anticancer activity of CX-3543: a direct inhibitor of rRNA biogenesis. Cancer research 69, 7653-7661 (2009). 167. Drygin, D., et al. Targeting RNA polymerase I with an oral small molecule CX-5461 inhibits ribosomal RNA synthesis and solid tumor growth. Cancer research 71, 1418-1430 (2011). 168. Xu, H., et al. CX-5461 is a DNA G-quadruplex stabilizer with selective lethality in BRCA1/2 deficient tumours. Nature communications 8, 14432 (2017). 169. https://clinicaltrials.gov/ct2/show/NCT02719977; (last accessed, May 26, 2019). 170. Felsenstein, K.M., et al. Small Molecule Microarrays Enable the Identification of a Selective, Quadruplex-Binding Inhibitor of MYC Expression. ACS chemical biology 11, 139-148 (2016). 225 171. Calabrese, D.R., et al. Chemical and structural studies provide a mechanistic basis for recognition of the MYC G-quadruplex. Nature communications 9, 4229 (2018). 172. Dai, J., Carver, M., Hurley, L.H. & Yang, D. Solution structure of a 2:1 quindoline-c-MYC G-quadruplex: insights into G-quadruplex-interactive small molecule drug design. Journal of the American Chemical Society 133, 17673-17680 (2011). 173. Hu, M.H., et al. Discovery of a New Four-Leaf Clover-Like Ligand as a Potent c-MYC Transcription Inhibitor Specifically Targeting the Promoter G-Quadruplex. Journal of medicinal chemistry 61, 2447-2459 (2018). 174. Chen, S.B., et al. Pharmacophore-based discovery of triaryl-substituted imidazole as new telomeric G-quadruplex ligand. Bioorganic & medicinal chemistry letters 21, 1004-1009 (2011). 175. Schmidt, A.W., Reddy, K.R. & Knolker, H.J. Occurrence, biogenesis, and synthesis of biologically active carbazole alkaloids. Chemical reviews 112, 3193-3328 (2012). 176. Stump, S., Mou, T.C., Sprang, S.R., Natale, N.R. & Beall, H.D. Crystal structure of the major quadruplex formed in the promoter region of the human c-MYC oncogene. PloS one 13, e0205584 (2018). 177. Weaver, M.J., et al. AIMing towards improved antitumor efficacy. Bioorganic & medicinal chemistry letters 25, 1765-1770 (2015). 178. Han, X., et al. Design, synthesis and biological evaluation of a novel class of anticancer agents: anthracenylisoxazole lexitropsin conjugates. Bioorganic & medicinal chemistry 17, 1671-1680 (2009). 179. Monsen, R.C. & Trent, J.O. G-quadruplex virtual drug screening: A review. Biochimie 152, 134-148 (2018). 180. Ma, D.L., et al. Discovery of a natural product-like c-myc G-quadruplex DNA groove-binder by molecular docking. PloS one 7, e43278 (2012). 181. Kuntz, I.D., Blaney, J.M., Oatley, S.J., Langridge, R. & Ferrin, T.E. A geometric approach to macromolecule-ligand interactions. Journal of molecular biology 161, 269-288 (1982). 182. Phan, A.T., Kuryavyi, V., Gaw, H.Y. & Patel, D.J. Small-molecule interaction with a five-guanine-tract G-quadruplex structure from the human MYC promoter. Nature chemical biology 1, 167-173 (2005). 183. Kang, H.J. & Park, H.J. In silico identification of novel ligands for G-quadruplex in the c-MYC promoter. Journal of computer-aided molecular design 29, 339-348 (2015). 184. Jain, A.N. Surflex: fully automatic flexible molecular docking using a molecular similarity-based search engine. Journal of medicinal chemistry 46, 499-511 (2003). 185. Hou, J.Q., et al. Identification of a selective G-quadruplex DNA binder using a multistep virtual screening approach. Chemical communications 51, 198-201 (2015). 186. Rocca, R., et al. Hit Identification of a Novel Dual Binder for h-telo/c-myc G-Quadruplex by a Combination of Pharmacophore Structure-Based Virtual Screening and Docking Refinement. ChemMedChem 11, 1721-1733 (2016). 187. Wolber, G. & Langer, T. LigandScout: 3-D pharmacophores derived from protein-bound ligands and their use as virtual screening filters. Journal of chemical information and modeling 45, 160-169 (2005). 188. Bhat, J., Mondal, S., Sengupta, P. & Chatterjee, S. In Silico Screening and Binding Characterization of Small Molecules toward a G-Quadruplex Structure Formed in the Promoter Region of c-MYC Oncogene. ACS omega 2, 4382-4397 (2017). 226 189. Berg, T., et al. Small-molecule antagonists of Myc/Max dimerization inhibit Myc-induced transformation of chicken embryo fibroblasts. Proceedings of the National Academy of Sciences of the United States of America 99, 3830-3835 (2002). 190. Shi, J., Stover, J.S., Whitby, L.R., Vogt, P.K. & Boger, D.L. Small molecule inhibitors of Myc/Max dimerization and Myc-induced cell transformation. Bioorganic & medicinal chemistry letters 19, 6038-6041 (2009). 191. Yin, X., Giap, C., Lazo, J.S. & Prochownik, E.V. Low molecular weight inhibitors of Myc-Max interaction and function. Oncogene 22, 6151-6159 (2003). 192. Wang, H., et al. Disruption of Myc-Max heterodimerization with improved cell-penetrating analogs of the small molecule 10074-G5. Oncotarget 4, 936-947 (2013). 193. Guo, J., et al. 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 chemotherapy and pharmacology 63, 615-625 (2009). 194. Clausen, D.M., et al. In vitro cytotoxicity and in vivo efficacy, pharmacokinetics, and metabolism of 10074-G5, a novel small-molecule inhibitor of c-Myc/Max dimerization. The Journal of pharmacology and experimental therapeutics 335, 715-727 (2010). 195. Fletcher, S. & Prochownik, E.V. Small-molecule inhibitors of the Myc oncoprotein. Biochimica et biophysica acta 1849, 525-543 (2015). 196. Wang, H., et al. Improved low molecular weight Myc-Max inhibitors. Molecular cancer therapeutics 6, 2399-2408 (2007). 197. Yap, J.L., et al. Pharmacophore identification of c-Myc inhibitor 10074-G5. Bioorganic & medicinal chemistry letters 23, 370-374 (2013). 198. Chauhan, J., et al. Discovery of methyl 4'-methyl-5-(7-nitrobenzo[c][1,2,5]oxadiazol-4-yl)-[1,1'-biphenyl]-3-carboxylate, an improved small-molecule inhibitor of c-Myc-max dimerization. ChemMedChem 9, 2274-2285 (2014). 199. Wanner, J., et al. Reversible linkage of two distinct small molecule inhibitors of Myc generates a dimeric inhibitor with improved potency that is active in myc over-expressing cancer cell lines. PloS one 10, e0121793 (2015). 200. Soodgupta, D., et al. Small Molecule MYC Inhibitor Conjugated to Integrin-Targeted Nanoparticles Extends Survival in a Mouse Model of Disseminated Multiple Myeloma. Molecular cancer therapeutics 14, 1286-1294 (2015). 201. Kiessling, A., Wiesinger, R., Sperl, B. & Berg, T. Selective inhibition of c-Myc/Max dimerization by a pyrazolo[1,5-a]pyrimidine. ChemMedChem 2, 627-630 (2007). 202. Kiessling, A., Sperl, B., Hollis, A., Eick, D. & Berg, T. Selective inhibition of c-Myc/Max dimerization and DNA binding by small molecules. Chemistry & biology 13, 745-751 (2006). 203. Stellas, D., et al. Therapeutic effects of an anti-Myc drug on mouse pancreatic cancer. Journal of the National Cancer Institute 106(2014). 204. Hart, J.R., et al. Inhibitor of MYC identified in a Krohnke pyridine library. Proceedings of the National Academy of Sciences of the United States of America 111, 12556-12561 (2014). 205. Choi, S.H., et al. Targeted Disruption of Myc-Max Oncoprotein Complex by a Small Molecule. ACS chemical biology 12, 2715-2719 (2017). 206. Castell, A., et al. A selective high affinity MYC-binding compound inhibits MYC:MAX interaction and MYC-dependent tumor cell proliferation. Scientific reports 8, 10064 (2018). 227 207. Fujimori, T., Wirsching, P. & Janda, K.D. Preparation of a Kröhnke Pyridine Combinatorial Library Suitable for Solution-Phase Biological Screening. Journal of Combinatorial Chemistry 5, 625-631 (2003). 208. Lin, C.Y., et al. Transcriptional amplification in tumor cells with elevated c-Myc. Cell 151, 56-67 (2012). 209. Zeller, K.I., Jegga, A.G., Aronow, B.J., O'Donnell, K.A. & Dang, C.V. An integrated database of genes responsive to the Myc oncogenic transcription factor: identification of direct genomic targets. Genome biology 4, R69 (2003). 210. Schuhmacher, M., et al. The transcriptional program of a human B cell line in response to Myc. Nucleic acids research 29, 397-406 (2001). 211. Wang, H., et al. Direct inhibition of c-Myc-Max heterodimers by celastrol and celastrol-inspired triterpenoids. Oncotarget 6, 32380-32395 (2015). 212. Jung, K.Y., et al. Perturbation of the c-Myc-Max protein-protein interaction via synthetic alpha-helix mimetics. Journal of medicinal chemistry 58, 3002-3024 (2015). 213. Mo, H. & Henriksson, M. Identification of small molecules that induce apoptosis in a Myc-dependent manner and inhibit Myc-driven transformation. Proceedings of the National Academy of Sciences of the United States of America 103, 6344-6349 (2006). 214. Mo, H., Vita, M., Crespin, M. & Henriksson, M. Myc overexpression enhances apoptosis induced by small molecules. Cell cycle 5, 2191-2194 (2006). 215. Jeong, K.C., Ahn, K.O. & Yang, C.H. Small-molecule inhibitors of c-Myc transcriptional factor suppress proliferation and induce apoptosis of promyelocytic leukemia cell via cell cycle arrest. Molecular bioSystems 6, 1503-1509 (2010). 216. Jeong, K.C., et al. Intravesical instillation of c-MYC inhibitor KSI-3716 suppresses orthotopic bladder tumor growth. The Journal of urology 191, 510-518 (2014). 217. Seo, H.K., et al. Antitumor activity of the c-Myc inhibitor KSI-3716 in gemcitabine-resistant bladder cancer. Oncotarget 5, 326-337 (2014). 218. Drews, J. Drug discovery: a historical perspective. Science 287, 1960-1964 (2000). 219. Amaro, R.E., et al. Ensemble Docking in Drug Discovery. Biophysical journal 114, 2271-2278 (2018). 220. Friesner, R.A., et al. Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. Journal of medicinal chemistry 47, 1739-1749 (2004). 221. Halgren, T.A., et al. Glide: a new approach for rapid, accurate docking and scoring. 2. Enrichment factors in database screening. Journal of medicinal chemistry 47, 1750-1759 (2004). 222. Pearlman, D.A.C., D.A.; Caldwell, J.W.; Ross, W.S.; Cheatham III, T.E.; DeBolt, S.; Fergusin, D.; Seibel, G; Kollman, P. AMBER, a package of computer programs for applying molecular mechanics, normal mode analysis, molecular dynamics and free energy calculations to simulate the structural and energetic properties of molecules. Computer Physics Communications 91, 1-41 (1995). 223. Yao, R., et al. Identification of a novel c-Myc inhibitor with anti-tumor effects on multiple myeloma cells. Bioscience reports (2018). 224. Holien, T., Vatsveen, T.K., Hella, H., Waage, A. & Sundan, A. Addiction to c-MYC in multiple myeloma. Blood 120, 2450-2453 (2012). 228 225. Szabo, A.G., et al. Overexpression of c-myc is associated with adverse clinical features and worse overall survival in multiple myeloma. Leukemia & lymphoma 57, 2526-2534 (2016). 226. Spitzer, R. & Jain, A.N. Surflex-Dock: Docking benchmarks and real-world application. Journal of computer-aided molecular design 26, 687-699 (2012). 227. Sybyl-X2.1 http://www.certara.com (last accessed Nov 1, 2018). 228. Franke, N.E., et al. Impaired bortezomib binding to mutant beta5 subunit of the proteasome is the underlying basis for bortezomib resistance in leukemia cells. Leukemia 26, 757-768 (2012). 229. Jiang, H., et al. Stabilizers of the Max homodimer identified in virtual ligand screening inhibit Myc function. Molecular pharmacology 76, 491-502 (2009). 230. Morris, G.M.G., D.S.; Halliday, R.S., Huey, R.; Hart, W.e.;Belew, R.K.; Olson, A.J. Automated docking using a Lamarckian genetic algorithm and an empirical binding free energy function. Journal of computational chemistry 19, 1639-1662 (1998). 231. Follis, A.V., Hammoudeh, D.I., Wang, H., Prochownik, E.V. & Metallo, S.J. Structural rationale for the coupled binding and unfolding of the c-Myc oncoprotein by small molecules. Chemistry & biology 15, 1149-1155 (2008). 232. Berjanskii, M.V., Neal, S. & Wishart, D.S. PREDITOR: a web server for predicting protein torsion angle restraints. Nucleic acids research 34, W63-69 (2006). 233. MacKerell, A.D., et al. All-atom empirical potential for molecular modeling and dynamics studies of proteins. The journal of physical chemistry. B 102, 3586-3616 (1998). 234. van der Lee, R., et al. Classification of intrinsically disordered regions and proteins. Chemical reviews 114, 6589-6631 (2014). 235. Muller, I., et al. Targeting of the MYCN protein with small molecule c-MYC inhibitors. PloS one 9, e97285 (2014). 236. Hammoudeh, D.I., Follis, A.V., Prochownik, E.V. & Metallo, S.J. Multiple independent binding sites for small-molecule inhibitors on the oncoprotein c-Myc. Journal of the American Chemical Society 131, 7390-7401 (2009). 237. Fisette, O., Lague, P., Gagne, S. & Morin, S. Synergistic applications of MD and NMR for the study of biological systems. Journal of biomedicine & biotechnology 2012, 254208 (2012). 238. Ambadipudi, S. & Zweckstetter, M. Targeting intrinsically disordered proteins in rational drug discovery. Expert Opinion on Drug Discovery 11, 65-77 (2016). 239. Michel, J. & Cuchillo, R. The impact of small molecule binding on the energy landscape of the intrinsically disordered protein C-myc. PloS one 7, e41070 (2012). 240. Yuan, Y., Pei, J. & Lai, L. Binding site detection and druggability prediction of protein targets for structure-based drug design. Current pharmaceutical design 19, 2326-2333 (2013). 241. Yu, C., et al. Structure-based Inhibitor Design for the Intrinsically Disordered Protein c-Myc. Scientific reports 6, 22298 (2016). 242. A, M. 2018 FDA drug approvals. Nature Reviews Drug Discovery 18(2019). 243. Cheng, Y., et al. Rational drug design via intrinsically disordered protein. Trends in biotechnology 24, 435-442 (2006). 244. Lambert, S.A., et al. The Human Transcription Factors. Cell 175, 598-599 (2018). 229 245. Lambert, M., Jambon, S., Depauw, S. & David-Cordonnier, M.H. Targeting Transcription Factors for Cancer Treatment. Molecules 23(2018). 246. Tsafou, K., Tiwari, P.B., Forman-Kay, J.D., Metallo, S.J. & Toretsky, J.A. Targeting Intrinsically Disordered Transcription Factors: Changing the Paradigm. Journal of molecular biology 430, 2321-2341 (2018). 247. Chen, X.P. & Du, G.H. Target validation: A door to drug discovery. Drug discoveries & therapeutics 1, 23-29 (2007). 248. Dulbecco, R. A turning point in cancer research: sequencing the human genome. Science 231, 1055-1056 (1986). 249. Gashaw, I., Ellinghaus, P., Sommer, A. & Asadullah, K. What makes a good drug target? Drug discovery today 16, 1037-1043 (2011). 250. Sioud, M. Main approaches to target discovery and validation. Methods in molecular biology 360, 1-12 (2007). 251. Broach, J.R. & Thorner, J. High-throughput screening for drug discovery. Nature 384, 14-16 (1996). 252. Bleicher, K.H., Bohm, H.J., Muller, K. & Alanine, A.I. Hit and lead generation: beyond high-throughput screening. Nature reviews. Drug discovery 2, 369-378 (2003). 253. Macalino, S.J., Gosu, V., Hong, S. & Choi, S. Role of computer-aided drug design in modern drug discovery. Archives of pharmacal research 38, 1686-1701 (2015). 254. Laurence L. Brunton, R.H.-D., Björn C. Knollmann. Goodman & Gilman's: The Pharmacological Basis of Therapeutics, 13e Eds. Laurence L. Brunton, et al. New York, NY: McGraw-Hill. (2018). 255. Neervannan, S. Preclinical formulations for discovery and toxicology: physicochemical challenges. Expert opinion on drug metabolism & toxicology 2, 715-731 (2006). 256. Tuntland, T., et al. Implementation of pharmacokinetic and pharmacodynamic strategies in early research phases of drug discovery and development at Novartis Institute of Biomedical Research. Frontiers in pharmacology 5, 174 (2014). 257. Ocana, A., Pandiella, A., Siu, L.L. & Tannock, I.F. Preclinical development of molecular-targeted agents for cancer. Nature reviews. Clinical oncology 8, 200-209 (2010). 258. Herter-Sprie, G.S., Kung, A.L. & Wong, K.K. New cast for a new era: preclinical cancer drug development revisited. The Journal of clinical investigation 123, 3639-3645 (2013). 259. Parulekar, W.R. & Eisenhauer, E.A. Phase I trial design for solid tumor studies of targeted, non-cytotoxic agents: theory and practice. Journal of the National Cancer Institute 96, 990-997 (2004). 260. El-Maraghi, R.H. & Eisenhauer, E.A. Review of phase II trial designs used in studies of molecular targeted agents: outcomes and predictors of success in phase III. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 26, 1346-1354 (2008). 261. Buyse, M. Phase III design: principles. Chinese clinical oncology 5, 10 (2016). 262. Meng, X.Y., Zhang, H.X., Mezei, M. & Cui, M. Molecular docking: a powerful approach for structure-based drug discovery. Current computer-aided drug design 7, 146-157 (2011). 263. Neves, M.A., Totrov, M. & Abagyan, R. Docking and scoring with ICM: the benchmarking results and strategies for improvement. Journal of computer-aided molecular design 26, 675-686 (2012). 230 264. McGann, M. FRED and HYBRID docking performance on standardized datasets. Journal of computer-aided molecular design 26, 897-906 (2012). 265. Friesner, R.A., et al. Extra precision glide: docking and scoring incorporating a model of hydrophobic enclosure for protein-ligand complexes. Journal of medicinal chemistry 49, 6177-6196 (2006). 266. Houston, D.R. & Walkinshaw, M.D. Consensus docking: improving the reliability of docking in a virtual screening context. Journal of chemical information and modeling 53, 384-390 (2013). 267. Kitchen, D.B., Decornez, H., Furr, J.R. & Bajorath, J. Docking and scoring in virtual screening for drug discovery: methods and applications. Nature reviews. Drug discovery 3, 935-949 (2004). 268. Huang, N., Shoichet, B.K. & Irwin, J.J. Benchmarking sets for molecular docking. Journal of medicinal chemistry 49, 6789-6801 (2006). 269. Onufriev, A., Bashford, D. & Case, D.A. Exploring protein native states and large-scale conformational changes with a modified generalized born model. Proteins 55, 383-394 (2004). 270. Onufriev, A.V. & Case, D.A. Generalized Born Implicit Solvent Models for Biomolecules. Annual review of biophysics (2019). 271. Prime, 2018-4; Schrödinger LLC: New York, NY, USA, 2018. 272. De Vivo, M., Masetti, M., Bottegoni, G. & Cavalli, A. Role of Molecular Dynamics and Related Methods in Drug Discovery. Journal of medicinal chemistry 59, 4035-4061 (2016). 273. Oren M. Becker, A.D.M.J., Benoit Roux, Masakatsu Watanabe. Computational Biochemistry and Biophysics CRC Press. (2001). 274. Harder, E., et al. OPLS3: A Force Field Providing Broad Coverage of Drug-like Small Molecules and Proteins. Journal of chemical theory and computation 12, 281-296 (2016). 275. Predescu, C., et al. Computationally efficient molecular dynamics integrators with improved sampling accuracy. Molecular Physics 110, 967-983 (2012). 276. Loup, V. Computer Experiments on Classical Fluids. PhysRev 159(1967). 277. Lippert, R.A., et al. Accurate and efficient integration for molecular dynamics simulations at constant temperature and pressure. J Chem Phys 139, 164106 (2013). 278. Wang, L., et al. Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field. Journal of the American Chemical Society 137, 2695-2703 (2015). 279. Schrödinger Release 2019-1: Desmond Molecular Dynamics System, D. E. Shaw Research, New York, NY, 2019. Maestro-Desmond Interoperability Tools, Schrödinger, New York, NY, 2019. 280. Schrödinger Release 2019-1: FEP+, Schrödinger, New York, NY, 2019. 281. Liu, P., Kim, B., Friesner, R.A. & Berne, B.J. Replica exchange with solute tempering: a method for sampling biological systems in explicit water. Proceedings of the National Academy of Sciences of the United States of America 102, 13749-13754 (2005). 282. Wang, L., Friesner, R.A. & Berne, B.J. Replica exchange with solute scaling: a more efficient version of replica exchange with solute tempering (REST2). The journal of physical chemistry. B 115, 9431-9438 (2011). 231 283. Wang, L., Berne, B.J. & Friesner, R.A. On achieving high accuracy and reliability in the calculation of relative protein-ligand binding affinities. Proceedings of the National Academy of Sciences of the United States of America 109, 1937-1942 (2012). 284. Abel, R., Wang, L., Harder, E.D., Berne, B.J. & Friesner, R.A. Advancing Drug Discovery through Enhanced Free Energy Calculations. Accounts of chemical research 50, 1625-1632 (2017). 285. Johnson, M., Maggiora G.M., Eds. Concepts and Applications of Molecular Similarity. John Wiley & Sons: New York (1990). 286. Bajorath, J. Representation and identification of activity cliffs. Expert Opin Drug Discov 12, 879-883 (2017). 287. Maggiora, G., Vogt, M., Stumpfe, D. & Bajorath, J. Molecular similarity in medicinal chemistry. Journal of medicinal chemistry 57, 3186-3204 (2014). 288. Lo, Y.C., Rensi, S.E., Torng, W. & Altman, R.B. Machine learning in chemoinformatics and drug discovery. Drug discovery today 23, 1538-1546 (2018). 289. Hawkins, P.C., Skillman, A.G. & Nicholls, A. Comparison of shape-matching and docking as virtual screening tools. Journal of medicinal chemistry 50, 74-82 (2007). 290. Yang, S.Y. Pharmacophore modeling and applications in drug discovery: challenges and recent advances. Drug discovery today 15, 444-450 (2010). 291. Wermuth, C.G., Ganellin, C.R., Lindberg, P. & Mitscher, L.A. Glossary of terms used in medicinal chemistry (IUPAC Recommendations 1998). in Pure and Applied Chemistry, Vol. 70 1129 (1998). 292. Cherkasov, A., et al. QSAR modeling: where have you been? Where are you going to? Journal of medicinal chemistry 57, 4977-5010 (2014). 293. Todeschini, R.a.C., V. Handbook of Molecular Descriptors. WILEY-VCH (2000). 294. Jagiello, K., et al. Advantages and limitations of classic and 3D QSAR approaches in nano-QSAR studies based on biological activity of fullerene derivatives. Journal of nanoparticle research : an interdisciplinary forum for nanoscale science and technology 18, 256 (2016). 295. Cherkasov, A. & Jankovic, B. Application of 'inductive' QSAR descriptors for quantification of antibacterial activity of cationic polypeptides. Molecules 9, 1034-1052 (2004). 296. Cherkasov, A. “Inductive” Descriptors. 10 Successful Years in QSAR. Curr. Comput. Aided. Drug Des., 21-42 (2005). 297. Kola, I. & Landis, J. Can the pharmaceutical industry reduce attrition rates? Nature reviews. Drug discovery 3, 711-715 (2004). 298. ADMET Predictor 8.5. http://www.simulations-plus.com/software/admetpredictor/ (last accessed May, 2019). 299. http://fafdrugs4.mti.univ-paris-diderot.fr/. 300. Morgan, C.E., et al. The First Crystal Structure of the UP1 Domain of hnRNP A1 Bound to RNA Reveals a New Look for an Old RNA Binding Protein. Journal of molecular biology 427, 3241-3257 (2015). 301. Sali, A. & Blundell, T.L. Comparative protein modelling by satisfaction of spatial restraints. Journal of molecular biology 234, 779-815 (1993). 302. Fiser, A. & Sali, A. Modeller: generation and refinement of homology-based protein structure models. Methods in enzymology 374, 461-491 (2003). 232 303. Molecular Operating Environment (MOE), 2013.08; Chemical Computing Group ULC, 1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2018. 304. Soga, S., Shirai, H., Kobori, M. & Hirayama, N. Use of amino acid composition to predict ligand-binding sites. Journal of chemical information and modeling 47, 400-406 (2007). 305. Schrödinger Release 2018-1: Maestro, Schrödinger, LLC, New York, NY, 2018. 306. Sastry, G.M., Adzhigirey, M., Day, T., Annabhimoju, R. & Sherman, W. Protein and ligand preparation: parameters, protocols, and influence on virtual screening enrichments. Journal of computer-aided molecular design 27, 221-234 (2013). 307. Schrödinger Release 2018-1: Schrödinger Suite 2018-1 Protein Preparation Wizard; Epik,  Schrödinger, LLC, New York, NY, 2016; Impact, Schrödinger, LLC, New York, NY, 2016; Prime, Schrödinger, LLC, New York, NY, 2018. 308. Irwin, J.J. & Shoichet, B.K. ZINC--a free database of commercially available compounds for virtual screening. Journal of chemical information and modeling 45, 177-182 (2005). 309. Irwin, J.J., Sterling, T., Mysinger, M.M., Bolstad, E.S. & Coleman, R.G. ZINC: a free tool to discover chemistry for biology. Journal of chemical information and modeling 52, 1757-1768 (2012). 310. Zsoldos, Z., Reid, D., Simon, A., Sadjad, B.S. & Johnson, A.P. eHiTS: an innovative approach to the docking and scoring function problems. Current protein & peptide science 7, 421-435 (2006). 311. McGann, M. FRED pose prediction and virtual screening accuracy. Journal of chemical information and modeling 51, 578-596 (2011). 312. McGann, M.R., Almond, H.R., Nicholls, A., Grant, J.A. & Brown, F.K. Gaussian docking functions. Biopolymers 68, 76-90 (2003). 313. Hawkins, P.C., Skillman, A.G., Warren, G.L., Ellingson, B.A. & Stahl, M.T. Conformer generation with OMEGA: algorithm and validation using high quality structures from the Protein Databank and Cambridge Structural Database. Journal of chemical information and modeling 50, 572-584 (2010). 314. mol_rmsd Calculate RMSD's for docking results, Scientific Vector Language (SVL) source code provided by Chemical Computing Group ULC, 1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2018. 315. scoring Analysis tool for non-bonded intermolecular interactions: H-bonds, transition metal, water bridges, hydrophobic, Scientific Vector Language (SVL) source code provided by Chemical Computing Group ULC, 1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2018. 316. OpenEye Scientific Software, Santa Fe, NM. http://www.eyesopen.com (last accessed Jan, 2019). 317. Sterling, T. & Irwin, J.J. ZINC 15--Ligand Discovery for Everyone. Journal of chemical information and modeling 55, 2324-2337 (2015). 318. Genheden, S. & Ryde, U. The MM/PBSA and MM/GBSA methods to estimate ligand-binding affinities. Expert Opin Drug Discov 10, 449-461 (2015). 319. Li, J., et al. The VSGB 2.0 model: a next generation energy model for high resolution protein structure modeling. Proteins 79, 2794-2812 (2011). 320. Martyna, G.J.K., M. L.; Tuckerman, M. Nosé–Hoover chains: The canonical ensemble via continuous dynamics. J Chem Phys 97, 2635-2643 (1992). 233 321. Martyna, G.J.T.D.J.K., M. L.;. Constant pressure molecular dynamics algorithms. J Chem Phys 101, 4177-4189 (1994). 322. Tuckerman, M.B., B. J.; Martyna, G. J. Reversible multiple time scale molecular dynamics. J Chem Phys 97, 1990-2001 (1992). 323. Abel, R., Wang, L., Harder, E.D., Berne, B.J. & Friesner, R.A. Advancing Drug Discovery through Enhanced Free Energy Calculations. Accounts of chemical research 50, 1625-1632 (2017). 324. Dalal, K., et al. Bypassing Drug Resistance Mechanisms of Prostate Cancer with Small Molecules that Target Androgen Receptor-Chromatin Interactions. Molecular cancer therapeutics 16, 2281-2291 (2017). 325. Mendez, J. & Stillman, B. Chromatin association of human origin recognition complex, cdc6, and minichromosome maintenance proteins during the cell cycle: assembly of prereplication complexes in late mitosis. Molecular and cellular biology 20, 8602-8612 (2000). 326. Huang, M.J., Cheng, Y.C., Liu, C.R., Lin, S. & Liu, H.E. A small-molecule c-Myc inhibitor, 10058-F4, induces cell-cycle arrest, apoptosis, and myeloid differentiation of human acute myeloid leukemia. Experimental hematology 34, 1480-1489 (2006). 327. Carabet, L.A., et al. Computer-aided drug discovery of Myc-Max inhibitors as potential therapeutics for prostate cancer. Eur J Med Chem 160, 108-119 (2018). 328. Axerio-Cilies, P., et al. Inhibitors of androgen receptor activation function-2 (AF2) site identified through virtual screening. Journal of medicinal chemistry 54, 6197-6205 (2011). 329. Lack, N.A., et al. Targeting the binding function 3 (BF3) site of the human androgen receptor through virtual screening. Journal of medicinal chemistry 54, 8563-8573 (2011). 330. Munuganti, R.S., et al. Targeting the binding function 3 (BF3) site of the androgen receptor through virtual screening. 2. development of 2-((2-phenoxyethyl) thio)-1H-benzimidazole derivatives. Journal of medicinal chemistry 56, 1136-1148 (2013). 331. Munuganti, R.S., et al. Identification of a potent antiandrogen that targets the BF3 site of the androgen receptor and inhibits enzalutamide-resistant prostate cancer. Chemistry & biology 21, 1476-1485 (2014). 332. Ban, F., et al. Discovery of 1H-indole-2-carboxamides as novel inhibitors of the androgen receptor binding function 3 (BF3). Journal of medicinal chemistry 57, 6867-6872 (2014). 333. Li, H., et al. Characterization of a new class of androgen receptor antagonists with potential therapeutic application in advanced prostate cancer. Molecular cancer therapeutics 12, 2425-2435 (2013). 334. Li, H., et al. Discovery of small-molecule inhibitors selectively targeting the DNA-binding domain of the human androgen receptor. Journal of medicinal chemistry 57, 6458-6467 (2014). 335. Dalal, K., et al. Selectively targeting the DNA-binding domain of the androgen receptor as a prospective therapy for prostate cancer. The Journal of biological chemistry 289, 26417-26429 (2014). 336. Bickerton, G.R., Paolini, G.V., Besnard, J., Muresan, S. & Hopkins, A.L. Quantifying the chemical beauty of drugs. Nature chemistry 4, 90-98 (2012). 337. crdd.osdd.net/oscadd/qed/; (last accessed, May, 2019). 234 338. Blee, A.M. & Huang, H. Lineage plasticity-mediated therapy resistance in prostate cancer. Asian journal of andrology 21, 241-248 (2019). 339. Beltran, H., et al. A Phase II Trial of the Aurora Kinase A Inhibitor Alisertib for Patients with Castration-resistant and Neuroendocrine Prostate Cancer: Efficacy and Biomarkers. Clinical cancer research : an official journal of the American Association for Cancer Research 25, 43-51 (2019). 340. Kubota, Y., Kim, S.H., Iguchi-Ariga, S.M. & Ariga, H. Transrepression of the N-myc expression by c-myc protein. Biochemical and biophysical research communications 162, 991-997 (1989). 341. Jung, M., et al. A Myc Activity Signature Predicts Poor Clinical Outcomes in Myc-Associated Cancers. Cancer research 77, 971-981 (2017). 342. Xue, C., et al. MYCN promotes neuroblastoma malignancy by establishing a regulatory circuit with transcription factor AP4. Oncotarget 7, 54937-54951 (2016). 343. Dang, C.V., Le, A. & Gao, P. MYC-induced cancer cell energy metabolism and therapeutic opportunities. Clinical cancer research : an official journal of the American Association for Cancer Research 15, 6479-6483 (2009). 344. Larsen, S., et al. LMO3 interacts with p53 and inhibits its transcriptional activity. Biochemical and biophysical research communications 392, 252-257 (2010). 345. Abbas, T. & Dutta, A. p21 in cancer: intricate networks and multiple activities. Nature reviews. Cancer 9, 400-414 (2009). 346. Qing, G., et al. Combinatorial regulation of neuroblastoma tumor progression by N-Myc and hypoxia inducible factor HIF-1alpha. Cancer research 70, 10351-10361 (2010). 347. Lee, J.K., et al. N-Myc Drives Neuroendocrine Prostate Cancer Initiated from Human Prostate Epithelial Cells. Cancer cell 29, 536-547 (2016). 348. Roy, R., Huang, Y., Seckl, M.J. & Pardo, O.E. Emerging roles of hnRNPA1 in modulating malignant transformation. Wiley interdisciplinary reviews. RNA 8(2017). 349. Jean-Philippe, J., Paz, S. & Caputi, M. hnRNP A1: the Swiss army knife of gene expression. International journal of molecular sciences 14, 18999-19024 (2013). 350. Lapuk, A.V., Volik, S.V., Wang, Y. & Collins, C.C. The role of mRNA splicing in prostate cancer. Asian journal of andrology 16, 515-521 (2014). 351. Liu, X., Zhou, Y., Lou, Y. & Zhong, H. Knockdown of HNRNPA1 inhibits lung adenocarcinoma cell proliferation through cell cycle arrest at G0/G1 phase. Gene 576, 791-797 (2016). 352. Chen, Y., et al. High expression of hnRNPA1 promotes cell invasion by inducing EMT in gastric cancer. Oncology reports 39, 1693-1701 (2018). 353. Loh, T.J., et al. CD44 alternative splicing and hnRNP A1 expression are associated with the metastasis of breast cancer. Oncology reports 34, 1231-1238 (2015). 354. Brockstedt, E., et al. Identification of apoptosis-associated proteins in a human Burkitt lymphoma cell line. Cleavage of heterogeneous nuclear ribonucleoprotein A1 by caspase 3. The Journal of biological chemistry 273, 28057-28064 (1998). 355. Shi, Y., et al. IL-6-induced stimulation of c-myc translation in multiple myeloma cells is mediated by myc internal ribosome entry site function and the RNA-binding protein, hnRNP A1. Cancer research 68, 10215-10222 (2008). 356. Iervolino, A., et al. hnRNP A1 nucleocytoplasmic shuttling activity is required for normal myelopoiesis and BCR/ABL leukemogenesis. Molecular and cellular biology 22, 2255-2266 (2002). 235 357. Zhang, S., et al. MYCN controls an alternative RNA splicing program in high-risk metastatic neuroblastoma. Cancer letters 371, 214-224 (2016). 358. David, C.J. & Manley, J.L. Alternative pre-mRNA splicing regulation in cancer: pathways and programs unhinged. Genes & development 24, 2343-2364 (2010). 359. Sun, F., et al. A novel prostate cancer therapeutic strategy using icaritin-activated arylhydrocarbon-receptor to co-target androgen receptor and its splice variants. Carcinogenesis 36, 757-768 (2015). 360. David, C.J., Chen, M., Assanah, M., Canoll, P. & Manley, J.L. HnRNP proteins controlled by c-Myc deregulate pyruvate kinase mRNA splicing in cancer. Nature 463, 364-368 (2010). 361. Babic, I., et al. EGFR mutation-induced alternative splicing of Max contributes to growth of glycolytic tumors in brain cancer. Cell metabolism 17, 1000-1008 (2013). 362. Takimoto, M., et al. Specific binding of heterogeneous ribonucleoprotein particle protein K to the human c-myc promoter, in vitro. The Journal of biological chemistry 268, 18249-18258 (1993). 363. Cogoi, S., Rapozzi, V., Cauci, S. & Xodo, L.E. Critical role of hnRNP A1 in activating KRAS transcription in pancreatic cancer cells: A molecular mechanism involving G4 DNA. Biochimica et biophysica acta. General subjects 1861, 1389-1398 (2017). 364. Jo, O.D., et al. Heterogeneous nuclear ribonucleoprotein A1 regulates cyclin D1 and c-myc internal ribosome entry site function through Akt signaling. The Journal of biological chemistry 283, 23274-23287 (2008). 365. Jain, N., Lin, H.C., Morgan, C.E., Harris, M.E. & Tolbert, B.S. Rules of RNA specificity of hnRNP A1 revealed by global and quantitative analysis of its affinity distribution. Proceedings of the National Academy of Sciences of the United States of America 114, 2206-2211 (2017). 366. Howard, J.M., et al. HNRNPA1 promotes recognition of splice site decoys by U2AF2 in vivo. Genome research 28, 689-698 (2018). 367. Tavanez, J.P., Madl, T., Kooshapur, H., Sattler, M. & Valcarcel, J. hnRNP A1 proofreads 3' splice site recognition by U2AF. Molecular cell 45, 314-329 (2012). 368. Liu, L.L., et al. Mechanisms of the androgen receptor splicing in prostate cancer cells. Oncogene 33, 3140-3150 (2014). 369. Xing, N., Chen, Y., Mitchell, S.H. & Young, C.Y. Quercetin inhibits the expression and function of the androgen receptor in LNCaP prostate cancer cells. Carcinogenesis 22, 409-414 (2001). 370. He, D., et al. Quercetin induces cell apoptosis of myeloma and displays a synergistic effect with dexamethasone in vitro and in vivo xenograft models. Oncotarget 7, 45489-45499 (2016). 371. Granato, M., et al. Concomitant reduction of c-Myc expression and PI3K/AKT/mTOR signaling by quercetin induces a strong cytotoxic effect against Burkitt's lymphoma. The international journal of biochemistry & cell biology 79, 393-400 (2016). 372. Ko, C.C., et al. Chemical proteomics identifies heterogeneous nuclear ribonucleoprotein (hnRNP) A1 as the molecular target of quercetin in its anti-cancer effects in PC-3 cells. The Journal of biological chemistry 289, 22078-22089 (2014). 373. Tummala, R., Lou, W., Gao, A.C. & Nadiminty, N. Quercetin Targets hnRNPA1 to Overcome Enzalutamide Resistance in Prostate Cancer Cells. Molecular cancer therapeutics 16, 2770-2779 (2017). 236 374. Rollins, C., Levengood, J.D., Rife, B.D., Salemi, M. & Tolbert, B.S. Thermodynamic and phylogenetic insights into hnRNP A1 recognition of the HIV-1 exon splicing silencer 3 element. Biochemistry 53, 2172-2184 (2014). 375. Carabet, L.A., et al. Computer-Aided Discovery of Small Molecules Targeting the RNA Splicing Activity of hnRNP A1 in Castration-Resistant Prostate Cancer. Molecules 24(2019). 376. McGovern, S.L. & Shoichet, B.K. Kinase inhibitors: not just for kinases anymore. Journal of medicinal chemistry 46, 1478-1483 (2003). 377. Baell, J. & Walters, M.A. Chemistry: Chemical con artists foil drug discovery. Nature 513, 481-483 (2014). 378. http://zinc15.docking.org/substances/ZINC000003869685/. 379. http://fafdrugs4.mti.univ-paris-diderot.fr/fafqed.html. 380. Lejeune, D., Delsaux, N., Charloteaux, B., Thomas, A. & Brasseur, R. Protein-nucleic acid recognition: statistical analysis of atomic interactions and influence of DNA structure. Proteins 61, 258-271 (2005). 381. Renaud, J.-P., et al. Cryo-EM in drug discovery: achievements, limitations and prospects. Nature Reviews Drug Discovery 17, 471 (2018). 382. Shan, C., et al. Chemical intervention of the NM23-H2 transcriptional programme on c-MYC via a novel small molecule. Nucleic acids research 43, 6677-6691 (2015). 383. Huth, J.R., et al. NMR-driven discovery of benzoylanthranilic acid inhibitors of far upstream element binding protein binding to the human oncogene c-myc promoter. Journal of medicinal chemistry 47, 4851-4857 (2004). 384. Fan, L., et al. Histone demethylase JMJD1A promotes alternative splicing of AR variant 7 (AR-V7) in prostate cancer cells. Proceedings of the National Academy of Sciences of the United States of America 115, E4584-E4593 (2018). 385. Stockley, J., et al. The RNA-binding protein Sam68 regulates expression and transcription function of the androgen receptor splice variant AR-V7. Scientific reports 5, 13426 (2015). 386. Shiota, M., et al. Potential Role for YB-1 in Castration-Resistant Prostate Cancer and Resistance to Enzalutamide Through the Androgen Receptor V7. Journal of the National Cancer Institute 108(2016). 387. Ferraldeschi, R., et al. Second-Generation HSP90 Inhibitor Onalespib Blocks mRNA Splicing of Androgen Receptor Variant 7 in Prostate Cancer Cells. Cancer research 76, 2731-2742 (2016). 388. Li, Y., et al. SRRM4 Drives Neuroendocrine Transdifferentiation of Prostate Adenocarcinoma Under Androgen Receptor Pathway Inhibition. European urology 71, 68-78 (2017).     237 Appendix  Figure A.1 Computer-aided drug discovery (CADD) Platform.   Figure A.2 VPC-70127 effect on growth of Myc-dependent LNCaP cells and Myc-independent HO15.19 cell line. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0380506/manifest

Comment

Related Items