Assays and Inhibitors for Jumonjic Domain-Containing Histone Demethylases As Epigenetic Regulators

Assays and Inhibitors for JumonjiC Domain-Containing Histone Demethylases as Epigenetic Regulators

Martin Roatsch

Assays and Inhibitors for JumonjiC Domain-Containing Histone Demethylases as Epigenetic Regulators

Inauguraldissertation

zur Erlangung des akademischen Grades eines Doctor rerum naturalium (Dr. rer. nat.)

der FakultätfürChemie und Pharmazie der Albert-Ludwigs-UniversitätFreiburg

vorgelegt von Martin Roatsch geboren in Berlin, Deutschland

2016 Eidesstattliche Versicherung

gemäß §7, Absatz 1, Satz 3, Nr. 8 der Promotionsordnung der Albert-Ludwigs-Universitätfürdie FakultätfürChemie und Pharmazie

1. Bei der eingereichten Dissertation zu dem Thema Assays and Inhibitors for JumonjiC Domain-Containing Histone Demethylases as Epigenetic ” Regulators“ handelt es sich um meine eigenst¨andigerbrachte Leistung.

2. Ich habe nur die angegebenen Quellen und Hilfsmittel benutzt und mich keiner unzulässigen Hilfe Dritter bedient. Insbesondere habe ich wörtlich oder sinngemäßaus anderen Werken übernommene Inhalte als solche kenntlich gemacht.

3. Die Dissertation oder Teile davon habe ich bislang nicht an einer Hochschule des In- oder Auslands als Bestandteil einer Pr¨ufungs-oder Qualiﬁkationsleistung vorgelegt.

4. Die Richtigkeit der vorstehenden Erkl¨arungen best¨atigeich.

5. Die Bedeutung der eidesstattlichen Versicherung und die strafrechtlichen Folgen einer unrichtigen oder unvollst¨andigeneidesstattlichen Versicherung sind mir bekannt.

Ich versichere an Eides statt, dass ich nach bestem Wissen die reine Wahrheit erkl¨artund nichts ver- schwiegen habe.

Freiburg i. Br., 2. Juni 2016 Unterschrift

Dekan der Fakultät: Prof. Dr. Manfred Jung Vorsitzender des Promotionsausschusses: Prof. Dr. Stefan Weber Referent: Prof. Dr. Manfred Jung Koreferent: Prof. Dr. Bernhard Breit Datum der mündlichen Prüfung: 29. Juni 2016 Datum der Verpflichtung: 7. Juli 2016 Zusammenfassung

JumonjiC-Domänen-enthaltende Histon-Demethylasen (JMJC-Demethylasen) sind epigeneti- sche Radierer“-Enzyme, welche Methylgruppen von den ε-Aminogruppen der Histon-Lysin- ” reste in einem oxidativen und Eisen(II)-/α-Ketoglutarat-abhängigenMechanismus entfernen. Ihre Beteiligung an einer Vielzahl von Krankheiten, insbesondere Krebs, wurde beschrieben. Eine Screening-Plattform mit zwei orthogonalen In-vitro-Enzymaktivitäts-Messsystemen wurde fürJMJD2A (KDM4A) etabliert und die LANCEUltra-Methode auf die verwandten Demethylasen JARID1A (KDM5A) und JMJD3 (KDM6B) übertragen. Mithilfe dieser Plattform wurden mehrere Screening-Kampagnen fürHemmstoff-Leitstrukturen durchgeführt, basierend sowohl auf rationalem Design, fragment-basiertem und analogie-basiertem Screening, virtuellem Screening sowie der Untersuchung fokussierter Substanz-Bibliotheken. Hiermit konnten mehrere neue Leitstrukturen fürHemmstoffe identifiziert und iterativ in Rückkopplung mit biologischen Testungen optimiert werden. Dies gilt für Naturstoffe und synthetische Derivative der Tropolon-Klasse sowie fragment-artige Tetrazolylhydrazide mit bemerkenswerter KDM4A-Selektivität. Mithilfe von virtuellem Screening wurden Aminopyrimidylpyridine mit submikromolarer Potenz und KDM4- und KDM5-Subfamilien- Selektivitätentdeckt. Eine Syntheseplattform fürAnaloga und Prodrugs wurde etabliert und diese Prodrugs hemmten das Zellwachstum von KYSE-150-Speiseröhrenkrebs-Zellen.Ein Hydroxamat-basierter Hemmstoff von Zink-abhängigenHiston-Desacetylasen wurde so modi- fiziert, dass seine Enzymklassen-SelektivitätfürJMJC-Demethylasen umgekehrt wurde. Die Struktur-Aktitiväts-Beziehungen wurden eingehend untersucht und Derivate hemmten das Zellwachstum von Leukämie-Zellenin JMJC-Demethylase-abhängigerArt. Klinisch einge- setzte Eisen-Chelatoren wie Deferasirox und neue synthetisierte Analoga hemmten potent die Demethylasen sowohl in vitro als auch in vivo in einer Reihe von Speiseröhrenkrebs-Zellenund direkte Effekte auf die zelluläreHiston-Methylierung konnten nachgewiesen werden. Diese neuen Leitstrukturen stellen wertvolle Werkzeug-Verbindungen dar, mit denen die physiologische Rolle der JMJC-Demethylasen untersucht werden kann, und dienen als Ausgangspunkt fürdie Entwicklung neuartiger Therapiekandidaten.

i Summary

JumonjiC domain-containing histone demethylases (JMJC demethylases) are epigenetic “eraser” enzymes that oxidatively remove methyl marks from the ε-amino groups of histone lysine residues in an iron(II)- and 2-oxoglutarate-dependent mechanism. Their involvement in a number of diseases and, in particular, cancer has been recognized. A screening platform using two orthogonal in vitro enzyme activity assays has been established for JMJD2A (KDM4A) and the LANCEUltra system also expanded to be used with related demethylases JARID1A (KDM5A) and JMJD3 (KDM6B). This platform has been used in several lead structure screening campaigns using fragment-based drug discovery, rational design, virtual screening, and analogy-based and focused library screening. Thus, several novel inhibitor lead structures were discovered and optimized iteratively in feedback with biological testing. This includes a natural product and its analogs of the tropolone class and fragment-like tetrazolyl hydrazides with remarkable KDM4A selectivity. Virtual screening identified aminopyrimidylpyridines with sub-micromolar potency and KDM4 and KDM5 subfamily selectivity. A synthetic platform towards analogs and prodrugs was established and prodrugs inhibited cell proliferation of KYSE-150 esophageal cancer cells. A hydroxamate-based inhibitor of zinc-dependent HDACs could be modified and its enzyme class selectivity reverted towards JMJC demethylases. Its structure-activity relationship was studied extensively and derivatives inhibited cell proliferation of a leukemia cell line in a JMJC demethylase-dependent manner. Clinically used iron chelators like deferasirox and novel synthetic analogs potently inhibited demethylases in vitro and in vivo in a broad range of esophageal cancer cells with on-target effects on cellular histone methylation levels. These novel lead structures are valuable tool compounds to study the physiological role of JMJC demethylases and provide starting points for therapeutic candidates.

ii The experimental parts of this research project as well as the production of the present thesis were undertaken between March 2012 and May 2016 in the group of Prof. Dr. Manfred Jung at the Institute of Pharmaceutical Sciences, Albert-Ludwigs-University Freiburg.

I would like to express my gratitude to all the individuals who have made it possible for me to undertake this fascinating research project and who have made our time together so very rewarding. First and foremost to my supervisor Prof. Dr. Manfred Jung for inviting me into his group, sharing this interesting and compelling topic of our project with me, and his many useful hints and valuable techniques I was allowed to learn from him. I would also like to acknowledge the other members of my thesis advisory committee Prof. Dr. Bernhard Breit and Dr. Wolfgang Hüttelfor their time and dedication and many inspiring discussions of this project. I am also grateful for the help I received from ‘Team Jumonji’ Dr. Inga Hoffmann and Dr. Ludovica Morera, with whom I enjoyed a very fruitful collaboration over so many years and without whom this project would not have advanced so far. I would like to thank all present and former members of the Jung group: Bright Asamoah, Johannes Bacher, Tobias Borgmann, Teresa Burgahn, Dr. Alokta Chakrabarti, Julia Eib, Dr. Silviya Furdas, Sören Hammelmann, Mirjam Hau, Dr. Alexander Hauser, Daniel Herp, Eva-Maria Herrlinger, Dr. Kristina Keller, Andreas Kürner,Laura Münninghoff,Dr. Benjamin Maurer, Daria Monaldi, Dr. Tobias Rumpf, Dr. Matthias Schiedel, Karin Schmidtkunz, Dr. Martin Schmitt, Johannes Schulz-Fincke, Johanna Senger, Roman Simon, Dr. Diana Stolfa, SörenSwyter, Dr. Tobias Wagner, Alexan- dra Walter, Sandra Wenzler, and Dr. Sarah Zähringer.They have created a truly wonderful work atmosphere in and out of the lab, making it very worthwhile to achieve such riveting results together with such an enthusiastic team. I am also thankful to all thesis and intern students who chose to join the Jung group and who contributed to this project: Marina Auth, Paul Disse, Ahmed Hassan, Eva-Maria Herrlinger, BjörnKlaiber, Juliane Mietz, Soma Mohammadamin, Nicole Steck, and Katharina Wolf. I am grateful to the Collaborative Research Center 992 ‘Medical Epigenetics’ for funding of this project and to its integrated research training group. I have been allowed to meet many dedicated young scientists from different fields and benefit from their various expertises. This enthusiastic team of young students has supported me both in and outside of official meetings.

iii In particular, I want to point out Dr. Theresa Ahrens, Dr. Inga Hoffmann, C. Felix Krombholz, and Dr. Xavier Lucas, who have become not only collaborators, but great friends. I am equally grateful for the support I received through scholarships from the Studienstiftung des deutschen Volkes both during my undergraduate as well as my doctoral studies and the many inspiring people from all walks of life I was allowed to meet. I am thankful to the group of Prof. Dr. Wolfgang Sippl (MLU Halle-Wittenberg) and in particular Dr. Martin Pippel and Dr. Dina Robaa for their contributions in molecular docking and virtual screening, which have greatly advanced many aspects of this project. I would like to thank Dr. Henriette Franz and the entire team of Prof. Dr. Roland Schüle (University Medical Center Freiburg) for kindly producing the enzyme JMJD2A required for all in vitro experiments, for a wonderful and harmonic collaboration, and for welcoming me into their group during a lab rotation. I would like to extend my thanks to all collaborators who have contributed to this project either by synthesis of compounds or by conducting additional experiments, which were valuable to the advancement of the research project discussed in this thesis. This includes, among many others, Nicole Rügerand Dr. Thomas Emmrich (Link group, EMAU Greifswald), Michael Fürst(Heinrich group, FAU Erlangen-Nuremberg), Dr. Ralf Flaig (DLS Oxford), Kerstin Serrer (Schleicher group, ALU Freiburg), and Dr. Theresa Ahrens (Laßmann group, University Med- ical Center Freiburg). Lastly, I would like to thank Sandra Wenzler, Dr. Ludovica Morera, and Dr. Theresa Ahrens for proofreading of parts of this thesis and their many valuable suggestions. “We can only see a short distance ahead, but we can see plenty there that needs to be done.” Alan M. Turing (1912 – 1954)

Contents

1. Introduction 1 1.1. Epigenetics ...... 2 1.1.1. DNA Modifications ...... 4 1.1.2. Histone Modifications ...... 5 1.2. Histone Demethylases ...... 10 1.2.1. Lysine-Specific Demethylase 1 ...... 10 1.2.2. JumonjiC Domain-Containing Histone Demethylases ...... 11 1.2.2.1. Overview ...... 11 1.2.2.2. Mechanism ...... 11 1.2.2.3. Structural Aspects ...... 14 1.2.2.4. Relevance in Disease ...... 17 1.3. Assays for Histone Demethylases ...... 20 1.3.1. Assays Based on Formaldehyde Quantification ...... 21 1.3.2. Antibody-Based Assays ...... 23 1.3.3. Mass Spectrometry-Based Assays ...... 25 1.3.4. Other in vitro Assays ...... 26 1.4. Inhibitors of Histone Demethylases ...... 28 1.4.1. Co-Substrate Analogs as Inhibitors ...... 28 1.4.2. Catechol-Based Inhibitors ...... 31 1.4.3. Hydroxyquinoline-Based Inhibitors ...... 32 1.4.4. Hydroxamate-Based Inhibitors ...... 33 1.4.5. N,N ’-Biheterocyclic Inhibitors ...... 35 1.4.6. Other Mechanism-Based Small Molecule Inhibitors ...... 38 1.4.7. Inhibitors Based on Zinc Ejection ...... 40

vii 1.4.8. Peptidic Inhibitors ...... 40 1.4.9. Discussion ...... 41

2. Project Approach 43

3. Assay Development for JMJC Histone Demethylases 47 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays ...... 47 3.1.1. Assay Development and Optimization for JMJD2A ...... 47 3.1.1.1. Calibration of Fluorescence Intensity ...... 47 3.1.1.2. Enzyme Activity Measurements ...... 49 3.1.1.3. Substrate Solution Components ...... 50 3.1.1.4. DMSO Tolerance ...... 55 3.1.1.5. Microtiter Plates ...... 56 3.1.1.6. Assay Buﬀer ...... 57 3.1.2. Assay Validation – Statistical Robustness ...... 58 3.1.3. Assay Validation – Reference Inhibitors ...... 59 3.1.4. Possible Adaptation to other Enzymes and Substrates ...... 61 3.2. FDH-Counterscreening Assay ...... 64 3.3. LANCEUltra Assays ...... 68 3.3.1. Assay Development and Optimization for JMJD2A ...... 68 3.3.2. Assay Validation – Statistical Robustness ...... 71 3.3.3. Other Assay Parameters ...... 72 3.3.4. Assay Validation – Reference Inhibitors ...... 74 3.3.5. Comparison of JMJD2A Activity Assays ...... 75 3.3.6. Adaptation to other Enzymes and Substrates ...... 77 3.3.6.1. JMJD3 (KDM6B) ...... 77 3.3.6.2. JARID1A (KDM5A) ...... 80 3.4. DELFIA-Based Assay for JMJD2A ...... 83 3.5. Iron Chelation Assay Based on Ferrozine Displacement ...... 85

4. Novel Inhibitors of JMJC Histone Demethylases 91 4.1. Co-Substrate Analogs as Inhibitors ...... 91

viii 4.2. Tetrazolyl Hydrazide-Based Inhibitors ...... 94 4.2.1. Discovery ...... 94 4.2.2. Competitivity of Inhibition Towards 2-Oxoglutarate ...... 96 4.2.3. Structural Binding Model ...... 98 4.2.4. Selectivity ...... 99 4.2.5. Elaboration of the Lead Structure ...... 100 4.2.5.1. Acylhydrazone Derivatives ...... 100 4.2.5.2. Rigidification of the Spacer ...... 103 4.2.5.3. Branched Derivatives ...... 104 4.3. Tropolone-Based Inhibitors ...... 106 4.3.1. Discovery ...... 106 4.3.2. Competitivity of Inhibition Towards 2-Oxoglutarate ...... 107 4.3.3. Selectivity ...... 111 4.3.4. Cellular Activity ...... 111 4.4. Hydroxamic Acid-Based Inhibitors ...... 113 4.4.1. Discovery ...... 113 4.4.2. Elaboration of the Lead Structure – SAR ...... 116 4.4.2.1. Modifications to the Cap Group ...... 116 4.4.2.2. Modifications to the Spacer ...... 120 4.4.2.3. Modifications to the Hydroxamate Warhead – Propanoates . . 121 4.4.2.4. Modifications to the Hydroxamate Warhead – Phosphonates . 122 4.4.2.5. Other Hydroxamic Acid-Based Inhibitors ...... 125 4.4.3. Selectivity among JMJC Demethylases ...... 127 4.4.4. Selectivity over HDACs ...... 128 4.4.5. Mode of Action of JMJD2A Inhibition ...... 129 4.4.6. Cellular Activity ...... 131 4.5. Aminopyrimidylpyridine-Based Inhibitors ...... 134 4.5.1. Discovery ...... 134 4.5.2. Structure-Activity Relationship (SAR) ...... 138 4.5.3. Modifications to the General Lead Structure ...... 143 4.5.4. Structural Binding Model – X-Ray Crystallography ...... 148

ix 4.5.5. Competitivity of Inhibition Towards 2-Oxoglutarate ...... 151 4.5.6. Synthesis of Derivatives and Prodrugs ...... 153 4.5.7. Cellular Effects ...... 157 4.5.8. Selectivity ...... 159 4.6. Clinically Used Iron Chelators ...... 161 4.6.1. Discovery ...... 161 4.6.2. Elucidation of the Mode of Action ...... 163 4.6.2.1. Variation of Iron Content ...... 163 4.6.2.2. Competitivity Towards 2-Oxoglutarate ...... 164 4.6.2.3. EPR Spectroscopy ...... 166 4.6.3. Structural Binding Model ...... 168 4.6.4. Selectivity ...... 169 4.6.5. Synthesis of Deferasirox Analogs ...... 170 4.6.6. Elaboration of the Deferasirox Lead Structure ...... 172 4.6.7. Cellular Effects of Deferasirox ...... 173 4.6.7.1. Effects on Cell Proliferation ...... 173 4.6.7.2. On-Target Effects ...... 176 4.6.8. Discussion ...... 179

5. Outlook 181

6. Conclusion 183

7. Experimental Section 185 7.1. Biochemical Assays ...... 185 7.1.1. General Remarks ...... 185 7.1.1.1. Reagents and Chemicals ...... 185 7.1.1.2. Peptides ...... 187 7.1.1.3. Antibodies ...... 187 7.1.1.4. Buﬀers ...... 188 7.1.1.5. Machines and Software ...... 188 7.1.1.6. Enzymes ...... 188

x 7.1.2. FDH-Coupled Enzyme Activity Assays ...... 190 7.1.2.1. Inhibition Curves ...... 190 7.1.2.2. Kinetic Activity Measurements ...... 192 7.1.3. FDH-Counterscreening Assay ...... 192 7.1.4. LANCEUltra Assays ...... 194 7.1.4.1. Inhibition Curves for JMJD2A ...... 194 7.1.4.2. Inhibition Curves for JARID1A ...... 196 7.1.4.3. Inhibition Curves for JMJD3 ...... 196 7.1.4.4. Competitivity Investigations against 2-Oxoglutarate ...... 196 7.1.5. Ferrozine Assay ...... 197 7.2. Chemical Synthesis ...... 198 7.2.1. General Remarks ...... 198 7.2.2. Synthesis of Deferasirox Analogs ...... 199 7.2.2.1. Synthesis of Starting Materials ...... 199 7.2.2.2. Couplings ...... 202 7.2.3. Synthesis of Aminopyrimidylpyridine-Based Inhibitors ...... 207 7.2.3.1. Synthesis of Starting Materials ...... 207 7.2.3.2. Synthesis of Guanidines ...... 209 7.2.3.3. Couplings ...... 212 7.2.3.4. Saponiﬁcations ...... 218 7.2.3.5. Transesteriﬁcation ...... 220

References 222

A. List of Publications 243

B. List of Abbreviations 247

C. List of Compound Synonyms 251

List of Figures

1.1. Schematic representation of the different levels of compaction of genetic material3 1.2. Crystal structure of a nucleosome particle ...... 3 1.3. Modifications of the cytosine residue in DNA ...... 5 1.4. Chemical structures of modified lysine and arginine residues in histone tails . .6 1.5. Known post-translational modifications in histone H3 ...... 8 1.6. Reaction sequence in the demethylation reaction catalyzed by LSD1 ...... 10 1.7. Mechanism for the demethylation reaction catalyzed by JumonjiC domain-containing histone demethylases ...... 13 1.8. Domain architecture of selected JumonjiC domain-containing histone demethylases ...... 15 1.9. Overlay of crystal structures of JMJC domains ...... 16 1.10. In vitro assays for the quantification of formaldehyde (HCHO) ...... 22 1.11. Schematic outline of the LANCEUltra assay principle ...... 24 1.12. Structure of the enzyme’s co-substrate 2-oxoglutarate and first co-substrate analog inhibitors ...... 29 1.13. Structures of catechol-based inhibitors ...... 31 1.14. Structures of hydroxyquinoline-based inhibitors ...... 32 1.15. Structures of hydroxamate-based inhibitors ...... 34 1.16. Structures of N,N ’-biheterocyclic inhibitors ...... 36 1.17. Structures of other reported JMJC demethylase inhibitors ...... 38 1.18. Structures of zinc-ejecting JMJC demethylase inhibitors ...... 40

2.1. Drug discovery workﬂow and strategies for lead structure generation ...... 44

3.1. Calibration curve for ﬂuorescence of NADH ...... 48

xiii 3.2. Example of a kinetic FDH-coupled enzyme activity assay ...... 49 3.3. Michaelis-Menten kinetic curves for two peptide substrates ...... 52 3.4. Determination of optimal assay concentrations for 2-oxoglutarate (2-OG) and ferrous sulfate ...... 53

3.5. Comparison of FDH-coupled activity assays with FeSO4 and (NH4)2Fe(SO4)2 . 54 3.6. Eﬀect of DMSO content on performance of the FDH-coupled assay ...... 56 3.7. Determination of statistical robustness of the FDH-coupled assay for JMJD2A 59 3.8. Validation of the FDH-coupled activity assay for JMJD2A using reference inhibitors ...... 60 3.9. Example of a kinetic FDH-coupled activity assay using full-length histones as substrate ...... 62 3.10. Calibration experiments incubating formaldehyde dehydrogenase (FDH) with exogenous formaldehyde and NAD+ ...... 65 3.11. Examples of compounds tested in the FDH-counterscreening assay ...... 66

3.12. Speciﬁcity of the α-H3K9me2 antibody used in LANCEUltra assays ...... 69 3.13. Kinetic experiments using the LANCEUltra assay ...... 70 3.14. Determination of statistical robustness of the LANCEUltra assay for JMJD2A 72 3.15. Validation of the LANCEUltra activity assay for JMJD2A using reference inhibitor 2,4-PDCA ...... 74

3.16. Development of a LANCEUltra-based assay for JMJD3 and H3K27me3 .... 78

3.17. Development of a LANCEUltra-based assay for JARID1A and H3K4me3 .... 81 3.18. Validation of the LANCEUltra activity assay for JARID1A using reference inhibitor 2,4-PDCA ...... 82 3.19. Representative example of data obtained in DELFIA measurements ...... 84 3.20. Principle of the Ferrozine-based detection of iron and UV/vis spectra ...... 86 3.21. Calibration curve for iron(II) in the Ferrozine-based photometric assay . . . . . 88 3.22. Displacement of Ferrozine in the iron(II)-binding assay by test compounds . . . 89

4.1. Overview over screened 2-substituted acetohydrazides against JMJD2A . . . . 95 4.2. Control compounds to verify the SAR for tetrazolyl hydrazide inhibitors . . . . 96 4.3. Competitivity of inhibition of JMJD2A by tetrazolyl hydrazide 44k ...... 97

xiv 4.4. Proposed binding mode of tetrazolyl hydrazide inhibitor 44k in JMJD2A . . . 99 4.5. Docking pose of compounds 53c and 53e in the active site of JMJD2A . . . . 108 4.6. Competitivity of inhibition of JMJD2A by tropolone-based inhibitors ...... 110 4.7. Focused library screening of hydroxamate-based HDAC inhibitors against JMJD2A114 4.8. Control compounds to validate the relevance of the hydroxamic acid in 57a .. 115 4.9. Competitivity of inhibition of JMJD2A by hydroxamate-based inhibitors 57a and 63c ...... 131 4.10. Virtual screening campaign towards novel JMJC demethylase inhibitors . . . . 135 4.11. Co-crystal structure of 86a in the active site of JMJD2A ...... 149 4.12. Comparison of the binding mode of 86a in JMJD2A to that of previously published 2,4-PDCA 6 and 22b ...... 150 4.13. Competitivity of inhibition of JMJD2A by optimized pyrimidylpyridine inhibitor 86ag ...... 152 4.14. Competitivity of inhibition of JMJD2A by pyrimidylpyridine inhibitors 86a and 86b ...... 153 4.15. Synthesis of methyl ester prodrugs and other derivatives of pyrimidylpyridine inhibitors like 86 ...... 154 4.16. Synthesis of the N -morpholinoethyl ester prodrug of inhibitor 86ag ...... 156 4.17. Effect of pyrimidylpyridine-based inhibitors 86 and their prodrugs on proliferation of KYSE-150 cells ...... 158 4.18. Chemical structures of clinically used iron chelators ...... 162 4.19. Competitivity of inhibition of JMJD2A by metal ion chelators ...... 165 4.20. EPR spectra of iron in complex with deferasirox 109a and the enzyme JMJD2A 167 4.21. Proposed binding mode of deferasirox 109a in the active site of JMJD2A . . . 168 4.22. Synthesis of analogs of deferasirox 109a ...... 171 4.23. Cellular potency of deferasirox 109a and its methyl ester 109c on a variety of esophageal cancer cell lines ...... 175 4.24. Effect of deferasirox-based inhibitors on histone methylation in KYSE-150 cells 177 4.25. Effect of deferasirox-based inhibitors on histone methylation in KYSE-150 cells (cont’d.) ...... 178

List of Tables

1.1. Overview over reported JMJC demethylases and their substrate speciﬁcities . . 12 1.2. Selection of JumonjiC domain-containing histone demethylases linked to cancer 18

3.1. Optimized composition of the substrate solution for FDH-coupled JMJD2A assays 55 3.2. Comparison of diﬀerent microtiter plates for the FDH-coupled JMJD2A assay . 57 3.3. Composition of the substrate solution for the FDH-counterscreening assay . . . 66 3.4. Composition of the substrate solution for LANCEUltra JMJD2A assays . . . . 73 3.5. Composition of the detection mix for LANCEUltra JMJD2A assays ...... 73 3.6. Comparison of the FDH-coupled and LANCEUltra JMJD2A in vitro assays . . 76

4.1. Inhibition data of co-substrate analogs as potential inhibitors of JMJD2A . . . 92

4.2. IC50 values of tetrazolyl hydrazide-based inhibitors against JMJD2A ...... 95 4.3. Kinetic data of inhibition experiments with tetrazolyl hydrazide inhibitor 44k . 97 4.4. Selectivity data of the most potent tetrazolyl hydrazide-based inhibitor 44k against JMJD2A, JMJD3, and JARID1A ...... 100 4.5. Potency of tetrazolyl acylhydrazone-based test compounds against JMJD2A . . 102

4.6. IC50 values of rigidiﬁed tetrazolyl hydrazide-based inhibitors against JMJD2A . 103

4.7. IC50 values of branched tetrazolyl hydrazide-based inhibitors against JMJD2A 105

4.8. IC50 values of tropolone-based inhibitors against JMJD2A ...... 107 4.9. Selectivity data of tropolone-based inhibitors against JMJD2A, JMJD3, and JARID1A ...... 111 4.10. Cellular potency of tropolone-based inhibitors on KYSE-150 cells ...... 112 4.11. Potency of substituted phenylalanine-based hydroxamic acids as inhibitors of JMJD2A ...... 117 4.12. Potency of 3-substituted tyrosine-based hydroxamic acids as JMJD2A inhibitors 118

xvii 4.13. Potency of 3-substituted tyrosine-based hydroxamic acids lacking the ester group as JMJD2A inhibitors ...... 119 4.14. Potency of hydroxamate-based JMJD2A inhibitors with modiﬁed spacers . . . 120 4.15. Potency of biphenylalanine hydroxamic acids with propanoate side chains as JMJD2A inhibitors ...... 122 4.16. Potency of hydroxamates with phosphonate side chains as JMJD2A inhibitors . 124 4.17. Potency of crebinostat-based JMJD2A inhibitors ...... 126 4.18. Selectivity of hydroxamic acid-based inhibitors against JMJD2A, JMJD3, and JARID1A ...... 128 4.19. Selectivity of hydroxamic acid-based inhibitors over histone deacetylases . . . . 129 4.20. Cellular potency of hydroxamate-based inhibitors on KYSE-150 and HL-60 cells 132 4.21. Results of in vitro testing of virtual screening hits ...... 136

4.22. IC50 values of aminopyrimidylpyridine-based inhibitors against JMJD2A . . . . 141

4.23. IC50 values of control compounds to validate the lead structure 86 ...... 144

4.24. IC50 values of synthesized analogs and prodrugs of lead structure 86 ...... 145

4.25. IC50 values of further analogs with deviations from the lead structure 86 .... 147 4.26. Selectivity of pyrimidylpyridine-based inhibitors and their prodrugs against JMJD2A, JMJD3, and JARID1A ...... 159 4.27. Inhibition data of clinically used iron chelators as potential inhibitors of JMJD2A162 4.28. Variation of potency of clinically used iron chelators with iron content in JMJD2A assays ...... 163 4.29. Selectivity of clinically used iron chelators against JMJD2A, JMJD3, and JARID1A170 4.30. In vitro characterization of deferasirox analogs as inhibitors of JMJD2A . . . . 172 4.31. Cellular potency of deferasirox-based inhibitors on KYSE-150 and HL-60 cells . 174

xviii 1. Introduction

Over the past decades, tremendous efforts have been undertaken to conceive innovative therapeutics to combat cancer, a group of diseases that is linked to one in eight deaths worldwide.[1] This has led to the development of outstanding pharmacological tools interfering with key cellular processes involved in the pathogenesis of cancer including anti-hormone therapy, inhibition of growth factor signaling and survival signaling cascades, or inhibition of angiogenesis.[2] Since the first publication of the human genetic code[3] and completion of the human genome project in 2003, it was believed that once the entire genome was sequenced, it would be possible to determine unambiguously the genetic causes responsible for the development of cancer and, consequently, for its therapeutic targeting. This has even led to the notion of “personalized medicine” – the idea that every patient could have their tumor tissue material sequenced, the key genetic factors for their disease determined, and the appropriate treatment tailored to their particular kind of cancer prescribed. Knowing beforehand, which patients would respond to which kind of therapy and if and when a tumor might become resistant, would certainly help avoid unnecessary treatment burdens and side effects. This concept has, however, still not held its promises to date.[4] Aside from ethical concerns, one of the main reasons may be that the simple notion that “all cancers arise as a result of changes that have occurred in the DNA sequence of the genomes of cancer cells”[1] and that cancer is, thus, essentially a disease of the genome, may be too simplified and overlooking critical regulatory mechanisms like epigenetics.

1 Chapter 1. Introduction

1.1. Epigenetics

Virtually every cell in the human body contains the complete and identical genetic material that is required to describe the organism. However, over 200 different cell types can be identified in humans, which all perform various different functions and differ dramatically in their size and morphology. This is possible since there are additional mechanisms at play that can regulate, which part of the genetic information is relevant for a given cell type and which is not. These cell type-specific expression patterns are also inherited from one cell generation to the next to ensure tissue fidelity across several cell divisions. These mechanisms, which are so to speak above or in addition to the genetically encoded information are termed Epigenetics (from Greek:επ ´ ´ι – upon, above, in addition to and γενεσις´ – origin, birth).[5, 6] More precisely, epigenetic phenomena are defined as

a change in phenotype that is heritable but does not involve DNA mutation.[5]

The question whether this applies solely to inheritance across cell divisions, i. e. mitotic inheritance, or can also include transgenerational inheritance through meiosis is still under debate.[7] In the case of mice, there is evidence, however, that certain traits can be passed on to future generations without changes in DNA sequence.[8, 9] Furthermore, histone methylation has been reported to be heritable transgenerationally in fission yeast.[10] One fundamental aspect of these regulatory mechanisms lies in the organization of the human genome within the cell. The genetic information of every cell is encoded in the base pair sequence of the deoxyribonucleic acid (DNA). The very distinct base pairing by hydrogen bonds (cytosine–guanine and thymine–adenine) and π-π-base stacking leads to the formation of a DNA double helical structure. In this form, however, the human genetic material would reach a length of roughly 2 meters, which is clearly more than the size of an individual cell, thus requiring mechanisms of compaction.[11] The different levels of compaction are summarized schematically in Figure 1.1. The first level consists in the formation of nucleosomes, the simplest repeating subunit of chromatin. In eu- karyotic cells, the DNA double helix does not exist loosely, but rather, it is wrapped around a complex of highly basic proteins, the histones. These are small (10–20 kDa) proteins containing an unusually high amount of lysine and arginine residues. Under physiological pH conditions,

2 1.1. Epigenetics

Figure 1.1.: Schematic representation of the diﬀerent levels of compaction of genetic material within the cell. Modiﬁed under license from Ref. 12.

Figure 1.2.: Crystal structure of a nucleosome particle. View through the axis, around which DNA is wrapped. Yellow – histone H2A, red – histone H2B, blue – histone H3, green – histone H4. Modiﬁed from Ref. 13 and based on PDB code 1EQZ.[14]

3 Chapter 1. Introduction the many basic residues of the histones are positively charged and can, thus, engage in strong ionic interactions with the negatively charged phosphodiester backbone of DNA. A stretch of ca. 147 base pairs of DNA is wound around the core histones for almost two turns. This complex of DNA with the histone proteins is termed the nucleosome. A crystal structure of a nucleosome has been published and is depicted in Fig. 1.2, showing how DNA is wrapped around the histone complex. There are four core histones, the highly conserved histones H2A, H2B, H3, and H4. One octameric histone complex contains one H3/H4 tetramer and two H2A/H2B dimers. Addi- tionally, a so-called linker histone H1 binds to the DNA already wrapped around the core and holds the complex in place. This arrangement results in a “beads on a string”-type structure, approximately 10 nm in size, which is packed relatively loosely (cf. Fig. 1.1). These beads on a string are then further spiralized and packaged to a more compact structure termed the 30 nm ﬁber and further to a very small and compact structure, the chromo- some. Quite simply speaking, regions of the genetic material that are more loosely packed (euchromatin) are more accessible to DNA replication and transcription than more densely packed heterochromatin. This is, thus, the ﬁrst governing mechanism, which controls, which portions of the genetic code are expressed and which are not.

1.1.1. DNA Modiﬁcations

In addition to the fact that DNA is wrapped around the histones and, thus, compacted, it can also carry chemical modifications of its own. The best studied is methylation in position C5 of the base cytosine. This is effected by DNA methyltransferases (DNMTs) using the co-factor S-adenosyl methionine (SAM). Methylation occurs primarily in regions, which are rich in cytosine-guanine base pairs, so-called CpG islands. Many genes in the human genome contain CpG islands in the promoter regions of the gene, which are, however, unmethylated and expressed. If methylation occurs at the promoter or around the transcription start site, this usually leads to gene repression, whereas methylation within the gene body does not block transcriptional elongation.[5, 6, 15] As a further level of complexity, enzymes are known, which can modify the already modified methylcytosine residues in DNA, the so-called ten-eleven translocation (TET) enzymes, of

4 1.1. Epigenetics

Figure 1.3.: Chemical structures of the potential modiﬁcations of cytosine in DNA. mC – methylcytosine, hmC – hydroxymethylcytosine, fC – formylcytosine, caC – carboxylcytosine.

which three are known: TET1, TET2, and TET3. These three enzymes likely have redundant functions. As iron(II)-dependent dioxygenases, they oxidize the methylcytosine residues further, as depicted in Fig. 1.3.[16, 17] The function of these oxidized variants is still under debate, but it is speculated that they play a role in the reversal of DNA methylation. This can occur either through spontaneous decarboxylation of carboxylcytosine (caC) or if the other oxidized variants function as triggers that can be recognized by DNA repair mechanisms such as base or nucleotide excision repair, which would then remove the methylated cytosine residue and have it replaced by unmethylated cytosine.[17–19] Recently, it has been reported that TET enzymes can also oxidize the DNA base thymine to hydroxyuracil in mouse embryonic stem cells.[20]

1.1.2. Histone Modiﬁcations

As can be seen from the crystal structure of the nucleosome (cf. Fig. 1.2), there are unordered N-terminal chains of the histones, which protrude out of the nucleosome particle. In recent years, it has become evident that these histone tails are highly modified in their amino acid side chains. Such post-translational modifications (PTMs) have attracted attention as they are believed to contain relevant information for the next level of epigenetic encoding.[21–24] Lysine residues are prone to acetylation on their ε-amino group. Similarly, lysine and arginine residues can be methylated. Here, another level of complexity is achieved as there are several possible methylation states. These include mono-, di-, and trimethylation for lysine residues and monomethylation and symmetric and asymmetric dimethylation for arginine residues, respectively. The chemical structures of the differently modified lysine and arginine residues

5 Chapter 1. Introduction

(a) Potential modiﬁcations of lysine residues in histone tails. Kac – acetyl-

lysine, Kme1 – methyllysine, Kme2 – dimethyllysine, Kme3 – trimethyllysine

(b) Potential modiﬁcations of arginine residues in histone tails. Rme1 – methylarginine, Rme2s – sym-

metrical dimethylarginine, Rme2a – asymmetrical dimethylarginine

Figure 1.4.: Chemical structures of potential post-translational modiﬁcations of lysine and arginine residues in histone tails.

6 1.1. Epigenetics are summarized in Fig. 1.4. In addition, arginine residues can also be transformed into the unnatural amino acid citrulline by deimination.[25] Furthermore, serine and threonine residues can be phosphorylated at their terminal OH groups. Ubiquitination of histone proteins is also known.[21, 22] The histone tail post-translational modiﬁcations are subject to a system of epigenetic proteins. These can be categorized into the three classes[26]

• Writers: enzymes, which transfer a histone modiﬁcation to the amino acid side chain;

• Erasers: enzymes, which remove histone modiﬁcations from the amino acid side chain;

• Readers: proteins, which specifically recognize histone modifications and recruit further effector proteins.

The writers include enzymes such as histone acetyltransferases (HATs), histone methyltransferases (HMTs), and protein arginine methyltransferases (PRMTs), which transfer acetyl and methyl groups, respectively. HATs use acetyl-CoA as co-factor to transfer the acetyl group onto lysine whereas HMTs require S-adenosyl methionine (SAM) as methyl source. Eraser enzymes are histone deacetylases (HDACs) and histone demethylases (HDMs), which remove these modifications. Histone lysine demethylases are discussed in detail in Section 1.2. To date, no arginine demethylases are known (cf. Fig. 1.4). The system of modifications can only have an effect if there are other proteins, which bind to the modified histone tails and can recruit further effector proteins. These include proteins with bromodomains specifically binding to acetylated lysine residues[26] as well as methyllysine readers. Examples contain chromo- domains, Tudor domains, or PHD domains. A common feature is the aromatic cage structure in their binding pocket, which can stabilize the methylated lysine residues and recognize the different degrees of methylation.[26–28] A summary of some of the known modifications of the first residues in histone H3 is given in Fig. 1.5. This large number of possible modifications including their combinations has led to the hypothesis of a so-called histone code,[29] where these modifications represent a coding language of their own similar to that of the DNA bases A, C, G, and T. Indeed, there seems to be a certain level of cross-talk between different types of modifications,[21] even though the meaning of this code is far from understood. This is further complicated by the fact that many

7 Chapter 1. Introduction

Figure 1.5.: Selection of the known post-translational modifications in histone H3. Only the first 20 residues are shown. Ac – acetylation, cit – citrullination, me – methylation, P – phosphorylation. Note that ‘me’ can denote any of the different methylation states for lysine and arginine (cf. Fig. 1.4). Note that lysine acetylation and methylation are mutually exclusive. Adapted from Refs. 6 and 21.

of the epigenetic enzymes also have non-histone targets and these post-translational modifications occur in several other protein networks as well.[30, 31] In one recent example of histone modification cross-talk,[21] it was reported that chemical inhibition of histone deacetylases (HDACs) also affected histone methylation globally. Anal- ysis of histones extracted from cells that were subjected to HDAC inhibitors revealed not only the expected histone hyperacetylation, but also increased levels of histone lysine methylation and reduced levels of arginine methylation. Further, the expression of some histone demethylases was down-regulated.[32] Seeing as the different histone modifications and their combinations represent a complex and context-dependent language[33, 34] rather than a simple code,[29] this may also represent compensatory mechanisms within the cell. In general, acetylation of lysines has a dramatic effect on the charge of this histone residue and its ability to interact with DNA. Acetylated lysine residues are, therefore, connected to a more relaxed chromatin state and higher transcription rates, whereas deacetylated regions are a hallmark of heterochromatin and repressed expression.[21, 33] In the case of methylation, this is more complicated and context-dependent. Trimethylation of lysine residues 4 and 36 in histone H3 (i. e. H3K4me3 and H3K36me3) is generally connected with increased transcrip- [21, 26, 33] tion, whereas H3K9me3 and H3K27me3 are considered repressive marks.

8 1.1. Epigenetics

As discussed above, for many of the histone modifications, both writers and erasers are known. This means that these regulatory marks are generally transient and reversible. This opens up the concept of epigenetics as a way for the cell to adapt to different environmental cues and to maintain a particular expression program as long as the cue is present, but also to stop it once expression of this genetic information is no longer required.[35] The same plasticity in histone modifications is also discussed to play a role in ageing.[36] If one considers DNA with all its genetic information as the ‘book of life,’ then epigenetics and, in particular, histone modifications can be viewed as ‘bookmarks in the book of life,’[23] regulating which parts of the genome are relevant to which cell type and at what times. Given the strong influence of these bookmarks on gene expression profiles, it becomes evident that aberrant histone modifications in gene regions where they do not belong can have detrimental effects and can lead to diseases. This can arise in the case where epigenetic enzymes and readers are mutated and thereby lose their function, are not expressed, or are overexpressed in certain cell types. Especially in the case of neoplastic diseases like cancer, there are numerous examples where epigenetic modifications are mis-written, mis-erased, or mis-read.[37] Again, the reversibility of such epi-mutations as opposed to permanent damages like DNA mutations offers a promising novel approach for treatment. Therefore, many of these epigenetic proteins have become interesting drug targets.[6, 15, 24, 26, 37] The need to develop specific epigenetic inhibitors as biological tool compounds and potential therapeutics has recently been recognized.[26, 38–41] This is particularly true for histone deacetylases, for which inhibitors are already in clinical use,[42] but increasingly more so also for histone acetyltransferases,[43] histone methyltransferases,[44] and histone demethylases.[45] Recently, also reader proteins have attracted more attention.[28] Several small-molecule drugs targeting epigenetic mechanisms, so-called epi-drugs, are already in clinical trials.[46, 47]

9 Chapter 1. Introduction

1.2. Histone Demethylases

As previously discussed (cf. Section 1.1.2), many of the histone modifications like lysine acetylation are reversible and exist in an equilibrium between their enzymatic writers and erasers. For a long time, however, lysine methylation was believed to be a permanent mark due to its slow turnover and the strong chemical nature of the C−N-bond.[48] Therefore, it was not until 2004 that this view was revised when the first lysine-specific demethylase was discovered.[49]

1.2.1. Lysine-Speciﬁc Demethylase 1

Lysine-specific demethylase 1 (LSD1) belongs to the class of monoamine oxidase-like enzymes and depends on flavin adenine dinucleotide (FAD) as its co-factor. The mechanism of demethylation is briefly depicted in Fig. 1.6. In an initial step, the terminal N -methyl group is de- hydrogenated by the enzyme reducing the redox factor FAD to FADH2. This results in the formation of an iminium intermediate. Upon addition of a molecule of water to this double bond, hydroxymethyllysine is formed. Due to its unstable nature as a hemiaminal, this intermediate spontaneously decomposes to yield formaldehyde and the remaining lysine residue carrying one less methyl group. The redox co-factor is regenerated with molecular dioxygen [49–52] resulting in the release of H2O2. As this mechanism proceeds via the iminium intermediate and, therefore, requires the presence of a double bond on the ε-N atom, LSD1 can only demethylate mono- and dimethylated lysine residues.[49, 50]

Figure 1.6.: Reaction sequence in the demethylation reaction catalyzed by LSD1.

10 1.2. Histone Demethylases

1.2.2. JumonjiC Domain-Containing Histone Demethylases

In 2006, only two years after the discovery of the first lysine-specific demethylase LSD1 (cf. Section 1.2.1), the much larger class of JumonjiC domain-containing histone demethylases (JMJC demethylases) was discovered independently by Tsukada et al.[53] and Whetstine et al.[54] These enzymes demethylate histones via a completely different catalytic mechanism (cf. Section 1.2.2.2) and are also able to demethylate trimethylated histone lysine residues.[54]

1.2.2.1. Overview

To date, about 30 proteins of the JumonjiC domain-containing histone demethylase class have been discovered in the human genome, 19 of which have been shown to possess enzymatic activity. These can be grouped into 7 families according to their sequence similarity.[45, 55, 57, 58] They each possess their own specificity with regard to which methylated lysine residues in the histone sequence they accept as substrate as well as to which degree of methylation is demethylated. Table 1.1 gives an overview over the different JMJC demethylases reported and their substrate preferences. Despite an effort to develop a systematic nomenclature for these enzymes,[59] the many different trivial names are still more commonly used and included in the table. In this systematic nomenclature, histone demethylases (HDMs) have been renamed lysine demethylases (KDMs) as some representatives also have non-histone substrates, e. g. p65 and NF-κB in the case of KDM2A.[60] In this system, LSD1 is termed KDM1A.[59] The distinct biological function of every demethylase is known for only a few of them. Given the similarity in their substrates, it is likely that many of them possess redundant functions.

1.2.2.2. Mechanism

JumonjiC domain-containing histone demethylases belong to the Cupin superfamily of dioxygenases[61] and demethylate histones in an entirely diﬀerent manner than the FAD-dependent enzyme LSD1 (cf. Fig. 1.6). JumonjiC domain-containing histone demethylases are metallo- enzymes with an Fe2+ ion in their active site and require 2-oxoglutarate (2-OG) as co-substrate. Their mechanism of action can be assumed to proceed analogous to that of other well characterized Fe2+-/2-OG-dependent enzymes.[45, 62, 63] Thus, the demethylation reaction proceeds via C−H bond activation and radical oxidation of the methyl group as depicted in Figure 1.7.

11 Chapter 1. Introduction

Table 1.1.: Overview over reported JMJC demethylases and their substrate speciﬁcities. Only those subtypes with proven enzymatic activity and known substrates are included. Adapted from Refs. 52, 55, and 56.

Class Systematic Name Synonyms Preferred Substrates

KDM2 KDM2A FBXL11, JHDM1A H3K36me2/1

KDM2B FBXL10, JHDM1B H3K4me3, H3K36me2/1

KDM3 KDM3A JMJD1A, JHDM2A H3K9me2/1

KDM3B JMJD1B, JHDM2B H3K9me2/1

KDM4 KDM4A JMJD2A, JHDM3A H3K9me3/2, H3K36me3/2

KDM4B JMJD2B, JHDM3B H3K9me3/2, H3K36me3/2

KDM4C JMJD2C, JHDM3C, GASC1 H3K9me3/2, H3K36me3/2

KDM4D JMJD2D, JHDM3D H3K9me3/2, H3K36me3/2

KDM4E JMJD2E H3K9me3/2

KDM5 KDM5A JARID1A, RBP2 H3K4me3/2

KDM5B JARID1B, PLU1 H3K4me3/2

KDM5C JARID1C, SMCX H3K4me3/2

KDM5D JARID1D, SMCY H3K4me3/2

KDM6 KDM6A UTX, MGC141941 H3K27me3/2

KDM6B JMJD3, KIAA0346 H3K27me3/2

KDM7 KDM7A KIAA1718 H3K9me2/1, H3K27me2/1

KDM7B KIAA1111, PHF8, ZNF422 H3K9me2/1, H4K20me1

KDM7C PHF2, CENP35 H3K9me2/1, H3K27me2/1

KDM8 KDM8 JMJD5 H3K36me2

In the active site, the ferrous ion is complexed by the imidazole groups of two histidine residues and the carboxylate group of an aspartate or glutamate residue. In the resting state, the three remaining coordination sites of the octahedrally coordinated Fe2+ are occupied by water molecules. In a ﬁrst step, the co-substrate 2-oxoglutarate and one molecule of dioxygen bind to the metal ion, displacing the water molecules. The redox-active ferrous ion center then

12 1.2. Histone Demethylases

Figure 1.7.: Mechanism for the demethylation reaction catalyzed by JumonjiC domain- containing histone demethylases. Modiﬁed under license from Ref. 52.

transfers a single electron onto the bound dioxygen molecule, resulting in the formation of an Fe3+ center and a highly reactive peroxide radical anion (step 2). Next, this reactive anion can integrate into the 2-oxoglutarate molecule forming a mixed anhydride (step 3). There is some debate on the nature of the iron species resulting from this. One option is an Fe3+ ion bound to a single hydroxyl radical anion as shown.[45, 52] Other publications, however, formulate this as a tetravalent oxoferryl FeIV−O species.[62, 64–66] Recently, computational

13 Chapter 1. Introduction eﬀorts have been made to resolve this question, but were essentially unable to rule out one or the other structure for JMJD2A (KDM4A).[67] Either way, this highly reactive oxygen species is able to activate the otherwise unreactive C−H bond in the terminal methyl group and to abstract a hydrogen atom. The resulting OH moiety is subsequently transferred back onto the methyl group resulting in the intermediate hydroxymethyllysine (step 4). As was the case in the demethylation reaction catalyzed by LSD1 (cf. Fig. 1.6), this unstable hemiaminal spontaneously decomposes to yield formaldehyde and the histone tail lacking one methyl group. Furthermore, the mixed anhydride bound to the metal center dissociates and readily fragments into CO2 and succinate (step 5). When three water molecules rebind the Fe2+ ion, the catalytic center is regenerated and can undergo another cycle of demethylation. This reaction sequence could, in principle, be repeated resulting in the removal of one methyl group after the other. The degree of demethylation of every JumonjiC demethylase is, however, determined by its intrinsic substrate speciﬁcity (cf. Table 1.1). In summary, the carbon atom of the methyl group is essentially oxidized to formaldehyde HCHO using one of the two oxygen atoms of the dioxygen molecule, while the other is used to cleave the 2-oxoglutarate co-substrate into succinate and CO2 (cf. Fig. 1.7). Because this mechanism proceeds via a radical oxidation step, it does not require the presence of a lone pair on the ε-N atom. Therefore, JumonjiC domain-containing histone demethylases are able to also demethylate trimethylated lysine residues (cf. Table 1.1).[26, 53–58, 65]

1.2.2.3. Structural Aspects

Most JumonjiC domain-containing histone demethylases are large multi-domain proteins. The domain architecture for a small selection of demethylases is represented schematically in Fig. 1.8. The most important domain is without doubt the JumonjiC domain (JMJC). This is the domain that actually contains the catalytically active site required for the demethylation reaction (cf. Section 1.2.2.2).[53, 54] In addition to the highly conserved JMJC domain, many demethylases contain other domains relevant for protein-protein and protein-DNA interactions. This includes the N-terminal Jumonji domain (JMJN), which is also highly conserved in the KDM4 and KDM5 subfamilies. Enzymes

14 1.2. Histone Demethylases

Figure 1.8.: Domain architecture of selected JumonjiC domain-containing histone demethylases. Blue – JMJN domain, red – ARID domain, orange – JMJC domain, green – PHD domain, brown – Tudor domain. Based on sequence annotations in the uniprot.org database.

of the KDM5 subfamily contain a so called AT-rich interacting domain (ARID), a domain required for binding of DNA stretches rich in AT base pairs. This allows these enzymes to reach a certain sequence selectivity as to their DNA binding. The presence of Jumonji and ARID domains has given these enzymes their names as JARID1A and so forth. Demethylases of the KDM4, KDM5, and KDM7 subfamilies contain PHD domains, which are protein-protein interaction domains selectively binding methyllysine and methylarginine regions and, thus, required for substrate binding. In the KDM4 subfamily, many enzymes also contain two Tudor domains, which are also methyllysine binding domains. JMJD2E (KDM4E) is a shortened pseudogene product that contains only the JMJN and catalytic JMJC domain.[55] The presence of methyllysine-binding domains like Tudor can have profound impacts on the catalytic activity of a histone demethylase. Lohse et al. found that histone peptides carrying a double modiﬁcation like H3K4me3K9me3 were demethylated dramatically faster by JMJD2A than the H3K9me3 peptide of the same length even though H3K4me3 is not a substrate of this

15 Chapter 1. Introduction

Figure 1.9.: Overlay of crystal structures of the JMJC domains from KDM2A, KDM4A, KDM6A, and KDM7A. Modiﬁed under license from Ref. 52.

enzyme (cf. Table 1.1). This effect was particularly strong for full-length JMJD2A containing the reader domains compared to the catalytic core alone.[68] This can be seen as another example of histone modification cross-talk.[21] It has been shown that the domain architecture of the KDM4 subfamily is evolutionarily conserved across species, with KDM4A–C orthologs existing in all vertebrates with the same substrate specificity. All analyzed eukaryotes like earlier animals and even flagellates and yeast contain at least one KDM4 enzyme.[69] JMJC domains have been identified in prokaryotes like some strains of B. subtilis and E. coli, but based on their sequence similarity it is likely that they arose from horizontal gene transfer from metazoan eukaryotes.[61] It remains unclear whether they have enzymatic function in bacteria and what a potential substrate may be. The catalytic JMJC domain has been crystallized and analyzed by x-ray crystallography numerous times with and without co-substrate and inhibitors as recently reviewed.[70, 71] The first such publication by Couture et al., in which Fe2+ is replaced by Ni2+ revealed the coordination environment around the metal center and binding of the 2-oxoglutarate co-substrate.[72]

16 1.2. Histone Demethylases

An overlay of several published structures of the JMJC domains of KDM2A, KDM4A, KDM6A, and KDM7A is depicted in Fig. 1.9. This shows the high conservation of the JMJC domain across several demethylases and subfamilies. The JMJC domain folds into a jelly-roll-like all-β- barrel fold.[66] In its center, the Fe2+ required for catalysis is coordinated by two histidine and one aspartate or glutamate residue (cf. Fig. 1.7). In this image, NOG 2 is co-crystallized, an amide analog of the co-substrate 2-oxoglutarate 1 (cf. Section 1.4.1), which occupies the same coordination position around the central metal ion.[52] At the edge of the catalytic domain, a highly conserved lysine residue (Lys206 in KDM4A) is located such that it can stabilize the terminal carboxylate group of the co-substrate 2-oxoglutarate through an ionic interaction (not shown). Many inhibitors, thus, contain a negatively charged residue to interact with this particular amino acid (cf. Section 1.4). When the catalytic domain of JMJD2A was crystallized with shortened peptides of its known [73] substrates H3K9me3 and H3K36me3, this allowed for an understanding of how the substrate speciﬁcity of JumonjiC domain-containing histone demethylase arises. Substrate speci-

ficity and recognition of the respective methylation status (me3, me2, or me1) is determined by a methylammonium-binding pocket directly adjacent to the metal ion center. This binding pocket differs in size and amino acid residues from demethylase to demethylase giving them their distinct substrate profiles. The methylammonium-binding pocket positions the carbon atom of the methyl group into proximity to the Fe2+ site so that it can be hydroxylated.[64, 72]

1.2.2.4. Relevance in Disease

As discussed previously (cf. Section 1.1.2), aberrant histone modiﬁcations are increasingly recognized as a driver for the onset and maintenance of disease states. Although their physiological role in detail is still poorly understood, this is increasingly also becoming clear for JumonjiC domain-containing histone demethylases.[26, 39, 52, 74–78] Most of the studies have focused on neoplastic diseases such as cancer. However, involvement of the JMJD2 family of demethylases in herpesvirus infection and reactivation[79] and of JMJD2A in the regulation of Kaposi’s sarcoma-associated herpesvirus (KSHV) has also been demonstrated.[80] In mice, JMJD2A also promoted cardiac hypertrophy in response to stress stimuli.[81] KDM3A knock-out mice developed an adult obesity phenotype and, thus, link

17 Chapter 1. Introduction

Table 1.2.: Selection of expression levels of JumonjiC domain-containing histone demethylases linked to cancer. Adapted from Ref. 56.

Enzyme Up-regulated in Down-regulated in KDM2A lung, breast, and gastric cancer prostate cancer KDM2B leukemia, pancreatic adenocarcinoma, bladder cancer prostate cancer KDM3A breast, prostate, colorectal, renal cancer KDM4A breast cancer bladder cancer KDM4C breast, esophageal, and prostate cancer, lymphoma KDM5A lung cancer melanoma KDM5B breast, prostate, and bladder cancer KDM5C prostate cancer KDM6B leukemia, lung, liver, and prostate cancer KDM7B non-small-cell lung cancer KDM7C esophageal squamous cell carcinoma colorectal cancer

histone demethylases to metabolic diseases.[82] Furthermore, demethylases of the KDM6 subfamily like JMJD3 have been linked to inﬂammation and regulation of the NF-κB pathway.[83] As is the case for the JumonjiC domain-containing histone demethylases, many epigenetic enzymes use key intermediary metabolites as their co-substrates such as 2-oxoglutarate, but also SAM (DNA methyltransferases and histone methyltransferases) or acetyl-CoA (histone acetyltransferases). This raises the interesting question, to which extent epigenetic modiﬁers respond to environmental signals and the availability of nutrients and other metabolites. This connection has particularly been discussed in connection with cellular ageing.[35, 36, 84] The involvement of histone lysine demethylases in cancer has amply been reviewed in recent years[37, 40, 45, 52, 55, 56, 75, 85–89] and is summarized in Table 1.2. As becomes evident, JumonjiC domain-containing histone demethylases are mostly up-regulated and overexpressed in human cancers, making them attractive drug targets and this has spurred an interest in the development of small-molecule inhibitors for these enzymes.[56, 85]

18 1.2. Histone Demethylases

As a very prominent example, the histone demethylase JMJD2A (KDM4A) has been implicated in the initiation and manifestation of a number of cancer types. This is particularly true for hormone-dependent tumors such as breast[90, 91] and prostate[92] cancer, where it has been shown to be a co-activator of the estrogen and androgen receptor, respectively. Androgen receptor-dependent gene expression after demethylation has also been shown for JMJD1A[93] and for JMJD2C in cooperation with LSD1.[94] Furthermore, overexpression of JMJD2A has also been reported in other solid tumor types such as lung,[95] bladder,[96] and colorectal[97] cancer. On the other hand, reduced levels of JMJD2A have been found in other reports in malignantly transformed urothelium and were suggested as a marker for poor prognosis.[98] The contrasting roles of JMJD2A in diﬀerent cancers have recently been reviewed by Guerra-Calderas et al.[99] Regardless, there is a need for the development of inhibitors of this enzyme either as drug candidates or as biological tool compounds to elucidate the physiological role of KDM4A.[39, 52, 85] Another member of the KDM4 subfamily, KDM4C, was originally described as a putative oncogene and called ‘gene associated with squamous cell carcinoma 1’ (GASC1) before its catalytic demethylase function was even known.[100] It has since been suggested as a prognostic and predictive marker in invasive breast cancer.[101] The reason may be that KDM4C (GASC1) has been described as a co-activator of hypoxia-inducible factor 1 (HIF-1).[102] Hypoxia is a common feature of many solid tumors. KDM4D, yet another KDM4 subfamily demethylase, has been reported to regulate the tumor suppressor p53 in a colon cancer cell line.[103] Given the strong similarity in their substrate proﬁles (cf. Table 1.1) and domain make-up (cf. Fig. 1.8), it seems likely that many or all of the KDM4 demethylases may be involved in this process through their redundant enzymatic functions. In addition to the KDM4 subfamily, other demethylases like JARID1A (KDM5A) and JMJD3 (KDM6B) were studied as representatives of the KDM5 and KDM6 subfamily. Both have also been tightly linked to human cancers,[87, 104] although there are also some contrasting reports for the H3K27 demethylases of the KDM6 subfamily.[105] Ntziachristos et al. reported that despite their similar enzymatic activity, KDM6A and KDM6B had opposing roles in a leukemia cell line, with KDM6B being an oncogenic driver, while KDM6A rather acted as a tumor suppressor that is, however, frequently genetically inactivated.[105]

19 Chapter 1. Introduction

1.3. Assays for Histone Demethylases

Since the discovery of JumonjiC domain-containing histone demethylases in 2006 (cf. Section 1.2.2.1) and the realization of their therapeutic potential for medicinal chemistry, in particular for cancer therapy (cf. Section 1.2.2.4), a number of efforts have been undertaken to develop selective and highly potent inhibitors for these enzymes (cf. Section 1.4). As a basis for inhibitor screening, however, biochemical measurement tools (so-called assays) are required to evaluate enzymatic activity as well as the effect of a given chemical compound on said activity. This section briefly reviews available assay technologies for JMJC demethylases. For a more comprehensive description, the reader is directed to Refs. 106 and 107. Some efforts have been made to develop in vivo screening platforms that allow for a phenotypical read-out of the effect of demethylase inhibition[108] such as for H3K4-specific demethylases in S. cerevisiae[109] and for JMJD3 inhibitors by use of high-content imaging in live cells,[110] but most screening campaigns are carried out in vitro, using solely the isolated enzyme.[106] This dramatically facilitates assay design and set-up. However, to understand the relevance of a screening hit as it is translated into an in vivo setting is not always trivial. As a further simplification, in vitro assays usually measure the activity of a histone demethylase on a synthetic peptide that is derived from the natural amino acid sequence in the histone and contains their modifications included as unnatural amino acids. For LSD1, however, screening assays using full-length native histones have also been reported.[111] The available in vitro assays can be classified into three categories based on their principle:[106]

• quantiﬁcation of the coupled reaction product formaldehyde HCHO;

• antibody-based techniques, where a highly speciﬁc antibody is required that can distinguish between diﬀerent histone methylation states before and after demethylation such

as H3K9me3 and H3K9me2;

• mass spectrometry-based techniques, where the demethylation reaction can be followed by the mass change of a substrate peptide corresponding to one methyl group.

20 1.3. Assays for Histone Demethylases

1.3.1. Assays Based on Formaldehyde Quantiﬁcation

As was discussed in the mechanism of action of JumonjiC domain-containing histone demethylases (cf. Section 1.2.2.2), every demethylation reaction leads to the concomitant formation of one equivalent of formaldehyde HCHO in stoichiometric amounts. If this coupled product can be quantiﬁed, it can be used as a proxy for the activity of the demethylase in question. The most common option is the use of a second enzyme, formaldehyde dehydrogenase (FDH), which further oxidizes HCHO to formic acid under reduction of its co-factor NAD+ to NADH.

The reduced form is a fluorescent molecule at λex = 330 nm and λem = 460 nm and can, thus, easily be quantified even in continuous kinetic assays. The principle of this FDH-coupled assay is schematically outlined in Fig. 1.10(a). It was originally introduced by Couture et al.[72] and has since become one of the most widely used assays to determine JMJC histone demethylase activity. Thanks to the work of Sakurai et al., who redesigned, simplified, and miniaturized this assay,[112] it has become amenable even to high-throughput screening.[113] One caveat of this assay is that it can be prone to false positive results if the test compound in question inhibits FDH rather than the JumonjiC demethylase. Furthermore, test compounds that are themselves fluorescent at the used wavelengths can also interfere with the measurement.[106] Care must be taken with regard to the choice of substrate peptide. Not only must the histone modification be a substrate of the demethylase in question (cf. Table 1.1), but other modifications on the same peptide can have dramatic effects on the activity of a given enzyme if it is used as the full-length protein containing its reader domains.[68] Another option for the detection of formaldehyde is the in situ formation of a fluorescent di- hydropyridine dye as outlined in Fig. 1.10(b). This assay principle, based on a Hantzsch reaction of acetoacetanilide and formaldehyde in the presence of ammonia,[114] is marketed as a commercial JumonjiC demethylase assay kit by Cayman Chemical. In terms of inhibitor screening and characterization, there is, however, only one very questionable reference with an iridium(III) complex reported as a supposed JMJD2D inhibitor.[115] A third option is the use of a formaldehyde-selective reactive probe, Purpald, which, together with formaldehyde forms a purple heterocyclic compound in the presence of oxygen. The adduct can self-assemble on gold nanoparticles and be detected by surface-enhanced Raman

21 Chapter 1. Introduction

(a) FDH-coupled assay

(b) Acetoacetanilide assay Figure 1.10.: Different in vitro assays for the quantification of formaldehyde (HCHO). See text for details. Modified under license from Ref. 106.

scattering spectroscopy. While this principle has only been demonstrated for LSD1, it may very well be transferable to JumonjiC demethylases.[116] Very recently, two chemical compounds have been reported independently, which allow for detection of HCHO levels in live cells.[117, 118] These both rely on silicon-heterocyclic dyes with a homoallylic amine coupled to a fluorescence quencher. Upon reaction with formaldehyde to an iminium intermediate, these can undergo an aza-Cope sigmatropic rearrangement and subsequent hydrolysis to liberate the open fluorescent form of the dye.[117, 118] It has even been demonstrated that the subtle changes in intracellular HCHO levels upon LSD1 inhibition in living cells can be detected by continuous imaging as a decrease in fluorescence.[117]

22 1.3. Assays for Histone Demethylases

1.3.2. Antibody-Based Assays

An entirely different approach to measuring demethylase activity is the use of highly specific antibodies that can detect the subtle change in a histone peptide during the demethylation reaction. This is not trivial as the greatest part of the antibody’s epitope remains identical, i. e. the peptide sequence, the modified lysine, and the charge on the terminal ammonium ion. It is only the steric demand of one methyl group that changes during the demethylation and that needs to be recognized by the antibody. Therefore, it is no surprise that the quality of commercial antibodies can have dramatic effects on the assay outcome. In a large cross-check campaign, it was found that at least 25% of the tested commercial antibodies against epigenetic marks had “substantial problems of specificity or utility.”[119] Nonetheless, several antibody-based assay systems have been developed that can accurately measure JumonjiC histone demethylase activity that differ only in the way antibody binding is eventually quantified. This includes standard techniques like western or dot blot and enzyme- linked immunosorbent assay (ELISA). An ELISA approach has been used in the screening of a natural product library[120] and the characterization of a hit that was first identified in a phenotypical screening.[121] A more sophisticated approach to antibody quantification is the proprietary lanthanide chelate excite (LANCE) technology marketed by PerkinElmer.[122] The basic principle is schematically outlined in Fig. 1.11. Here, the highly specific antibody is labeled with a europium chelate complex, which relies on the intrinsic fluorescence of lanthanide ions. Additionally, the detection system contains a proprietary dye termed ULight that is covalently coupled to a biological anchor like streptavidin. If biotinylated peptides are used, this means that they will always be strongly bound by the dye. If and only if the labeled antibody recognizes its substrate, the Eu3+ fluorophor and the dye are brought into close proximity and a fluorescence resonance energy transfer (FRET) effect can be observed.[122] This has several advantages over direct measurements of fluorescence intensity as in the FDH-coupled assay. Firstly, the FRET effect results in quite a large Stokes shift, which is unlikely to occur in interfering autofluorescent organic compounds and, secondly, the half-life time of the excited state is exceptionally long for lanthanide complexes. This can be exploited if fluorescence emission is measured with a small time delay after excitation (i. e. time-resolved fluorescence (TRF)). After a delay of e. g.

23 Chapter 1. Introduction

Figure 1.11.: Schematic principle of the LANCEUltra assay. B – biotin, Eu-Ab – europium-labeled antibody, SA – streptavidin. Modiﬁed from Ref. 122.

100 µs, there is still a strong fluorescence signal from the europium complex. However, possible interfering fluorescence from organic compounds will have faded.[106] This means that only exceptionally small amounts of enzyme and peptide are required to obtain good signal-to-noise ratios (cf. Section 3.3.1). On the other hand, the cost of the proprietary assay components can be a limiting factor for the use of such an assay.[106] The development of LANCEUltra assays was published by PerkinElmer for lysine-specific demethylase 1 (LSD1)[123] and described for JumonjiC histone demethylases in a number of technical application notes and posters.[122] The advantages of time-resolved fluorescence (TRF) can also be used in a different format termed dissociation-enhanced lanthanide fluorescent immunoassay (DELFIA), also marketed by PerkinElmer. In this heterogeneous assay involving several washing steps, biotinylated substrates are immobilized on streptavidin-coated microtiter plates. Only if the primary europium- labeled antibody binds the immobilized demethylated peptide after the incubation step with

24 1.3. Assays for Histone Demethylases the enzyme will it remain bound to the plate and not be removed by a successive washing step. Finally, the Eu3+ ions are released with an enhancement solution and their fluorescence can be measured. This principle has been described in the development of a JMJD2C inhibitor.[124] The many washing steps eliminate compound interference in the detection step, but also make such assays tedious and time-consuming and, hence, unsuitable for high-throughput screening. Lastly, a screening method using an amplified luminescent proximity homogeneous assay (AlphaScreen) has been published.[125] Here, the methylation state-specific antibodies are immobilized on beads. Additionally, the biotinylated substrates are also immobilized on a different kind of beads coated with a photosensitizer dye. Upon irradiation, these beads emit 1 singlet oxygen O2, which can only travel the distance to the antibody-presenting acceptor bead if all partners have correctly assembled the complex, i. e. after enzymatic demethylation has taken place. There, it reacts with a thioxene, which decomposes over several steps, yielding a chemiluminescence signal that can be detected.

1.3.3. Mass Spectrometry-Based Assays

Yet another conceptually different assay technology consists in the use of mass spectrometry (MS) to detect whether or not enzymatic demethylation has taken place based on the mass change of multiples of 14 units (one or more CH2 groups) of peptidic substrates. This technique using matrix-assisted laser desorption ionization (MALDI) has been employed in the discovery of the very first JMJC histone demethylase inhibitors (cf. Section 1.4).[113, 126] Using such a mass spectrometry approach systematically on a panel of synthetic peptides, it could be shown that many demethylases seem to be more promiscuous than previously assumed (cf. Table 1.1), with e. g. KDM4A, KDM4C, and KDM4E also accepting H3K27me3 and H3K27me2 as substrates. For KDM4D, no additional targets were found, questioning the presumed redundancy of all KDM4 members.[127] However, it remains to be determined how the results from such a simple in vitro experiment translate to the in vivo biological function of these enzymes. Traditionally, the throughput of mass spectrometric analyses is rather low. In an effort to overcome this, a novel label-free RapidFire MS assay has been developed with a sampling time of only 7 s per well.[128] Only recently, this analysis time was further shortened by the

25 Chapter 1. Introduction development of a multiplexing strategy, in which several mass-tagged samples are pooled and which has allowed for the screening of libraries even on the order of millions of compounds in reasonable time frames.[129] Rather than assaying the enzymatic activity of JMJC histone demethylases, nondenaturing mass spectrometry assays measure the binding of test compounds to the enzyme. If electrospray ionization (ESI-MS) is used, certain interactions can survive the transition from solution to gas phase and result in a detectable mass shift corresponding to the mass of the protein plus that of the bound small molecule.[130] Hit compounds thus discovered should, however, also be tested for their ability to actually inhibit enzymatic activity in order to rule out binding events of the compound in other domains of the protein or on its surface.

1.3.4. Other in vitro Assays

Other recently developed assay systems that use different physical principles and read-outs include a fluorescence polarization (FP) assay. Once a potent inhibitor of JMJC demethylases with proven binding in the active site of the enzyme has been developed, this can be fluores- cently labeled and used as a tracer in such an FP assay. If another test compound binds to the enzyme, it will competitively displace the labeled tracer and its polarization be decreased due to the greater flexibility in solution. This has been demonstrated with a hydroxamate- based inhibitor in an assay for JHDM1A and JMJD2A that can likely be adapted to other subtypes.[131, 132] Furthermore, another binding assay has been developed based on nuclear magnetic resonance 13 (NMR) spectroscopy with [1,2,3,4- C4]-labeled 2-oxoglutarate as a binding probe. Its signal in a 13C-NMR spectrum is attenuated and drastically broadened when bound inside the active site of the demethylase. Upon binding of a 2-OG-competitive inhibitor, it is liberated and gives a clearer spectrum in solution. This assay also has the advantage of ascertaining that positive hits actually bind in the active site of the enzyme and not on other parts.[133] However, the cost of the 13C-labeled probe and the measurement times in NMR likely prohibit the application of this assay in high-throughput screening campaigns.[106] Moreover, a scintillation proximity assay was developed for LSD1, JMJD1A, and JMJD2A using tritium-labeled S-adenosyl methionine (SAM). Once a histone methyltransferase incor-

26 1.3. Assays for Histone Demethylases porates the 3H-labeled methyl groups into a peptidic substrate in positions that have previously been demethylated by a histone demethylase, the scintillation of these peptides can be measured in especially coated microtiter plates. The quantiﬁed intensity of radioactivity can be correlated to the activity of the demethylase.[134]

In conclusion, a great number of different assay formats based on different physical read- out principles has been published. They each have their own advantages and disadvantages inherent in their methodology. It is, thus, wise to always employ several different assays based on orthogonal techniques to screen and, more importantly, validate the screening hits. This can help eliminate false positive results that arise from interferences of the test compound with components of the assay system itself.[106]

27 Chapter 1. Introduction

1.4. Inhibitors of Histone Demethylases

Potent and selective inhibitors of JumonjiC domain-containing histone demethylases are highly sought after both as candidates for pharmacological intervention in a number of diseases and as tool compounds to further elucidate the biological roles of these enzymes. This has led to the initiation of a number of drug discovery and development programs, the outcomes of which will be brieﬂy summarized in this section and have also been amply reviewed in the recent literature.[52, 86, 88, 106, 135–139] The most up-to-date and comprehensive summary is Ref. 139. The vast majority of reported small molecule inhibitors functions by chelating the central ferrous ion and, therefore, competing with the enzyme’s co-substrate 2-oxoglutarate (2-OG). These molecules, thus, all contain a metal-binding group (MBG), often a bi- or tridentate chelate structure. Throughout this section, the MBG of the inhibitors is shown in bold in the respective structures. In addition, potent inhibitors contain further functional groups, which allow additional interactions with active site residues and increase binding potency. One interaction of particular interest is that with a highly conserved lysine residue in the active site (Lys206 in the case of JMJD2A), which normally interacts with the terminal carboxylate group of the co-substrate 2-oxoglutarate. This residue is typically addressed by carboxylate or other charged or polar moieties on the inhibitor molecule. Although quite a number of inhibitors have been published so far, there are often no or only limited reports on their selectivity, both within the large family of JumonjiC domain-containing histone demethylases as well as with regard to other iron(II)-/2-oxoglutarate-dependent enzymes. Given the high degree of conservation of the JMJC domain active site, it is probable that most compounds function as pan-inhibitors with no or only limited selectivity. This may, however, be one of the reasons why so far no JMJC demethylase inhibitors have been advanced into the clinic.[52]

1.4.1. Co-Substrate Analogs as Inhibitors

Shortly after the discovery of JumonjiC domain-containing histone demethylases, the Schoﬁeld group published a large collection of potential inhibitor scaﬀolds[126, 140] based on known inhibitors of other Fe(II)-/2-oxoglutarate-dependent enzymes. Using a combination of the

28 1.4. Inhibitors of Histone Demethylases

Figure 1.12.: Structure of the enzyme’s co-substrate 2-oxoglutarate 1 and ﬁrst co- substrate analog inhibitors. See text for details.

formaldehyde dehydrogenase (FDH)-coupled enzymatic assay and mass spectrometry (MS) approaches, they were able to demonstrate potent inhibition of the model subtype JMJD2E by a number of these compounds.[126] This ﬁrst collection included very simple analogs of the co-substrate 2-OG 1, such as N -oxalyl glycine (NOG) 2 (cf. Fig. 1.12). This amide analog of 2-oxoglutarate can chelate the central ferrous ion, but would not be cleaved like the co-substrate into succinate and CO2. Therefore, [126] it can function as a mild competitive inhibitor (IC50 = 78 µM on JMJD2E). However, such a compound is very un-druglike due to its double negative charge and very polar nature. Furthermore, it is also a very general inhibitor of any 2-OG-dependent enzyme and as such not useful as an actual drug molecule to study the function of JumonjiC domain-containing histone demethylases. However, it can be used as a reference inhibitor for in vitro experiments.[52] Very recently, NOG 2 was also shown to occur as a natural product in rhubarb and spinach leaves.[141] The N -oxalyl glycine (NOG)-template has been used in follow-up studies yielding other oxalyl derivatives of further unnatural amino acids. This includes the NOG derivative 3 by Hamada et al. using a dimethylaminophenyl moiety, which was expected to yield additional hydrogen bonds of the tertiary amine with OH groups of tyrosine and serine residues in the active

29 Chapter 1. Introduction site. This optimized derivative was only potent at millimolar concentrations on the KDM4 subfamily, however.[142] This report also introduced a prodrug concept masking the negative charges of NOG as methyl esters. The uncharged diester would be more likely to pass the cell membrane and then be cleaved by intracellular esterases to release the active diacid molecule NOG within the cell.[142] In a larger study aimed at improving the selectivity of these inhibitors, the d-tyrosine derivative 4 was discovered as an inhibitor of JMJD2E with improved potency. The idea was to exploit another hydrophobic subpocket adjacent to the active site that is significantly larger in JMJC demethylases than in other related enzymes. Binding of 4 was confirmed by nondenaturing mass spectrometry and the predicted effect on the hydrophobic pocket in a co-crystal structure with JMJD2E.[130] As another inhibitor based on analogy to the co-substrate 2-oxoglutarate (2-OG), the reduced metabolite 2-hydroxyglutarate 5 was reported to competitively inhibit a JMJC demethylase from C. elegans as well as human JHDM1A in vitro with millimolar potency. The (S) enantiomer was found to be more potent than the (R) enantiomer. The results are relevant, however, as 5 is a metabolite that commonly accumulates in gliomas and leukemias.[143] The original collection of JMJC demethylase inhibitors[126] also contained the very simple, yet surprisingly potent inhibitor pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6. This molecule binds the central ferrous ion via its pyridine nitrogen atom and a charged oxygen atom of the

2-carboxylate group. 6 was found to inhibit JMJD2E with an IC50 value in the single-digit micromolar range in vitro and to be competitive to 2-oxoglutarate.[126] As revealed by a co- crystal structure, the para-carboxylic acid interacts with the conserved lysine residue in the active site and, thus, drastically increases binding potency through this ionic interaction. 6 was also found to inhibit JARID1B in vitro as well as in vivo.[144, 145] 2,4-PDCA is certainly also very un-druglike, but due to its simplicity and high potency has become one of the most widely used reference inhibitors for in vitro applications.[52, 106] In an eﬀort to improve the selectivity of 2,4-PDCA for JMJC demethylases over related enzymes, a series of 3-substituted derivatives was synthesized culminating in the amine-linked ﬂuorophenyl derivative 7, which nearly retained the potency of the unsubstituted 2,4-PDCA, but was completely selective for JMJD2E over prolyl hydroxylase 2 (PHD2) in vitro.[146]

30 1.4. Inhibitors of Histone Demethylases

In a very recent report, the 2-carboxylate group of 2,4-PDCA 6 was replaced by an ortho- phenol moiety culminating in compound 8 with an outstanding Ki value of 43 nM on JMJD2C. Addition of this extra phenyl ring allowed to grow this molecule and include further potency- inducing groups like an extra terminal OH. A library of 600,000 fragments was docked into the crystal structure of JMDJ2A, revealing 5-aminosalicylates as midmicromolar inhibitors. After investigations of their docking poses, fragments were linked and iteratively optimized to this highly potent inhibitor of the KDM3–6 subfamilies, which exhibited mild selectivity over KDM2 demethylases. Its mode of binding by x-ray crystallography is analogous to that of the co-substrate and of 2,4-PDCA 6, even though the chelating atoms (pyridyl-N and phenol-OH) are in a 1,5-arrangement as opposed to the typical 1,4-structure. Inhibition was competitive to 2-oxoglutarate.[147]

1.4.2. Catechol-Based Inhibitors

When Sakurai et al. redesigned and miniaturized the FDH-coupled assay (cf. Section 1.3.1),[112] they also used it to screen a library of natural products for inhibition of JMJD2E. Hit compounds were largely from the group of flavinoids and catechols, which can act as unspecific metal ion chelators via their ortho-diphenol group. IC50 values were generally in the single- digit micromolar range and inhibition verified in a mass spectrometry assay.[112] A selection 9–11 is summarized in Fig. 1.13. Further catechol-based compounds like caffeic acid 12 were discovered as KDM4C and KDM6B inhibitors in an antibody-based assay.[120] As many of these compounds also have distinct other functions in the organism or are even used as drugs with much higher potency on other targets, it is unlikely that such molecules can be developed further into useful JMJC demethylase inhibitors.[52]

Figure 1.13.: Structures of catechol-based inhibitors. See text for details.

31 Chapter 1. Introduction

1.4.3. Hydroxyquinoline-Based Inhibitors

When the newly optimized FDH-coupled assay was used for a high-throughput screen of a 236,000-member library of diverse compounds, 8-hydroxyquinolines were discovered as novel lead structures (cf. Fig. 1.14). The most potent derivative, 5-carboxy-8-hydroxyquinoline

(8-HQ-5-COOH, also known as IOX1 13a) inhibited JMJD2E with an IC50 value of only 0.2 µM.[113] 13a inhibits JMJD2E by chelation of the central ferrous ion via its quinoline- nitrogen atom and the adjacent hydroxyl group similar to the binding mode of 2,4-PDCA 6. In the same way, the 5-carboxylic acid is involved in an ionic interaction with the active site lysine residue. While IOX1 13a already showed moderate cellular potency, an n-octyl ester derivative 13b was prepared and improved the in vivo potency 30-fold in HeLa cells.[148] Recently, more elaborate derivatives based on the 8-HQ template have been prepared by Betti reactions, which introduced derivatizations in the 7 position of the quinoline ring. The optimized compound CCT1 14 demonstrated reduced in vitro potency, but improved selectivity for the KDM4 subfamily over other JMJC demethylases and also inhibited demethylation in vivo.[149] Substitution of the 8-HQ scaﬀold in position 6 yielded the inhibitor 15, one of the few examples of a JMJC demethylase inhibitor successfully proceeding beyond cell culture experiments into an animal model. It is reported as selective for the KDM4 subfamily and blocked prostate tumor growth in mice by suppression of androgen receptor-regulated genes, validating the hypothesis of KDM4 inhibition for cancer treatment (cf. Section 1.2.2.4). The biological outcomes resembled that of siRNA-mediated knockdown of KDM4B and seemed to involve downregulation of the critical cell-cycle gene PLK1.[150]

Figure 1.14.: Structures of hydroxyquinoline-based inhibitors. See text for details.

32 1.4. Inhibitors of Histone Demethylases

Derivatization of IOX1 13a by substitution in position 2 was also attempted based on docking simulations. However, this yielded only compounds with reduced potency on JMJD2A or complete lack of activity. No clear structure-activity relationship (SAR) could be deduced. Nonetheless, these derivatives exhibited improved selectivity over PHD2.[151] IOX1 13a has also been used as the selectivity-inducing group in a small-molecule probe coupled to a photoreactive cross-linking moiety and biotin as an affinity purification tag. This probe allows for the identification of Fe(II)-dependent oxygenases, which bind IOX1, e. g. from cell extracts or in competitive displacement assays.[152] Moreover, IOX1 13a and 2,4-PDCA 6 have been coupled to the LSD1 inhibitor tranyl- cypromine yielding novel compounds with dual selectivity both for LSD1 as well as JumonjiC domain-containing histone demethylases. These produced global hypermethylation in HeLa cells and exhibited potent antitumor activity in LNCaP and HCT116 cell lines.[153]

1.4.4. Hydroxamate-Based Inhibitors

Yet another metal-binding group (MBG) that was presented in the first collection of putative JMJC demethylase inhibitors[126] was that of hydroxamic acids R−CO−NH−OH. This included the natural product trichostatin A (TSA, 16) and its simplified analog suberoylanilide hydroxamic acid (SAHA, 17) with reported inhibition of JMJD2E (cf. Fig 1.15).[126] These can bind the central ferrous ion via the carbonyl oxygen atom and the terminal OH group of the hydroxamic acid. Hydroxamates have also been reported as inhibitors of zinc-dependent histone deacetylases (HDACs) and some are already in clinical use like SAHA 17,[42] but if the remainder of the molecule can be tuned correctly so as to exploit the features of the very different active sites, selectivity for one or the other class of enzymes could be achieved. Therefore, the hydroxamate scaffold was extensively studied by the Miyata group and led to the development of dimethylamino derivative 18 with an internal hydroxamic acid that is N - alkylated with a propanoate side chain.[154] This can be viewed as a mimic of the enzyme co-substrate 2-oxoglutarate and addresses the conserved lysine residue in the active site in an ionic interaction. Compound 18 was shown to be a potent inhibitor of JMJD2A and JMJD2C, with >100-fold selectivity over PHD2. A methyl ester prodrug of compound 18 showed growth inhibition in cancer cells in synergy with LSD1 inhibitors and, thus, makes it

33 Chapter 1. Introduction

Figure 1.15.: Structures of hydroxamate-based inhibitors. See text for details.

the first cell-permeable inhibitor of JumonjiC histone demethylases based on the hydroxamate binding motif.[154] On the basis of an available crystal structure of demethylase KDM7B, the same group de- signed the cyclopropyl hydroxamate inhibitor 19, which despite its similarity to 18 showed remarkable selectivity for KDM7A and KDM7B with only little inhibition of KDM2A. The demethylases of the KDM4 subfamily, however, were almost not inhibited. This compound also showed antiproliferative in vivo activity on HeLa cancer cells.[155] In similar fashion, the same group was also able to develop KDM5 subfamily-selective inhibitors like compound 20. Based on a docking study, it was reasoned that this compound can benefit from an additional interaction of the tertiary amine embedded in the alkyl chain with a tyrosine residue in the active site (Tyr472 in JARID1A), which is oriented differently in other demethylases. 20 increased trimethylation of H3K4 in vivo and inhibited lung cancer cell growth when used in combination with the HDAC inhibitor SAHA 17.[156] Another example of a small molecule probe derived from rational structure-based design is compound 21a, containing the hydroxamate as metal-binding group, a fumarate moiety as co- substrate mimic that addresses the positively charged lysine residue, and selectivity-inducing

34 1.4. Inhibitors of Histone Demethylases lipophilic portions. Its methyl ester 21b has become known as methylstat and been used as reference inhibitor for in vivo applications. 21a, however, is reported as relatively unselective with regard to histone demethylases and also shows activity on other iron(II)-dependent enzymes.[124] Furthermore, the highly reactive Michael acceptor fumarate is an unusual group in drug molecules as it can be prone to undesired reactions within the cell. Unfortunately, none of the reports on hydroxamate-based inhibitors mentioned above[124, 154–156] comments on the potential oﬀ-target inhibition of zinc-dependent histone deacetylases (HDACs) by these compounds or discusses to which extent this may play a role in the eﬀects observed in in vivo experiments.

1.4.5. N,N’-Biheterocyclic Inhibitors

Another privileged structure motif for the chelation of metal ions is clearly the concept of N,N ’-biheterocyclic compounds as depicted in Fig. 1.16. These can also be understood as extensions of the very simple inhibitor pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6. The simplest representative 2,2’-bipyridine-4,4’-dicarboxylate 22a was already included in the ﬁrst collection of JMJD2E inhibitors by Rose et al.[126] In a follow-up study, this was elaborated on and the structure-activity relationship (SAR) investigated leading to optimized compound

22b with an additional ethylenediamine group and an improved IC50 value of only 180 nM. A co-crystal structure of 22b in complex with JMJD2E revealed its binding to the central ferrous ion via the two pyridyl nitrogen atoms, the ionic interaction with Lys206 via the free carboxylate, and an additional ionic interaction of the terminal primary amine with an aspartate residue in the active site.[157] The best-known example of this structure class is possibly the case of pyrimidylpyridine inhibitor 23a that was published in 2012 by researchers at GlaxoSmithKline and termed GSK- [158] J1. It was described as outstandingly potent with an IC50 value of only 60 nM on JMJD3 (KDM6B) and as selective for the KDM6 subfamily as deduced from counterscreens against a number of different histone demethylases of the KDM3 and KDM4 subfamilies and unrelated enzymes. Its cell-permeable ethyl ester prodrug GSK-J4 23b showed the corresponding in vivo effects, i. e. increase in H3K27 trimethylation and modulation of the proinflammatory response in stimulated macrophages. A co-crystal structure with JMJD3 revealed a novel mode

35 Chapter 1. Introduction

Figure 1.16.: Structures of N,N ’-biheterocyclic inhibitors. See text for details.

of binding with the tetrahydrobenzazepine group extending into a hydrophobic cleft of the enzyme by mimicking a specific proline residue in the histone peptide. Furthermore, GSK-J1 23a induces an unusual metal shift in the active site.[158] Shortly after, however, the reported outstanding selectivity of GSK-J1 23a for the KDM6 subfamily was called into question[159, 160] as it was found to also inhibit members of the KDM5 subfamily with almost equal potency both in vitro and in vivo (for GSK-J4 23b), which were not investigated in the original study. In a recent study, the SAR of GSK-J1 23a was further studied by replacing the pyridine ring with thiazole, pyrazole, or triazole groups and exchanging the tetrahydrobenzazepine unit for other hydrophobic residues. However, no improvement in potency against JMJD3 could be obtained.[161] Nonetheless, the observed effects are useful information for further structure-based development of inhibitors of this enzyme. A different N,N ’-chelating motif based on the pyrazolone substructure was recently used in the development of inhibitors of hypoxia-inducible factor prolyl hydroxylases, which also belong to the class of iron(II)-/2-OG-dependent enzymes. However, the optimized compound 24 also inhibited JMJD3 with single-digit micromolar potency.[162] It is an unusual structure insofar

36 1.4. Inhibitors of Histone Demethylases as the carboxylic acid that is released after cleavage of the tert-butyl ester is in meta position to the pyridyl nitrogen atom as opposed to para in other known similar inhibitors. Furthermore, a triazolopyridine structure has been used in the development of highly potent inhibitors with selectivity for the KDM2 subfamily. The optimized compound 25 exhibited nanomolar inhibition of FBXL11 (KDM2A) with high selectivity over representatives of other KDM subfamilies. Changes in the substitution pattern of the piperidine moiety also gave useful insights to allow for the development of inhibitors of other subfamilies. This could, therefore, be used to ﬁne-tune the selectivity via the position and nature of the substituents attached to this moiety.[163] In a large collaborative eﬀort, the Structural Genomics Consortium (SGC) has very recently published compound 26,[164] the result of an optimization campaign based on a pyridylamino- thiazole high-throughput screening (HTS) hit from a 150,000 compound library. In an Alpha-

Screen assay, it is outstandingly potent with IC50 values of 17 nM on KDM4B and 14 nM on KDM5B, with selectivity over representatives from the KDM2, KDM3, and KDM6 subfamilies. Remarkably, the typical 4-carboxylate group in this type of inhibitor has been isosterically replaced by a pyrimidin-4-one moiety. This reduces the polarity of the compound and masks its negative charge all the while retaining the interaction pattern of the COOH group, i. e. of the carbonyl group with Lys206 and a hydrogen bond of the OH respectively NH with Tyr132. While the COOH group is highly beneficial for ligand affinity, it also contributes to the poor cell permeability. Loss of the charge in 26 made this compound cell permeable as experimentally validated without the need for a prodrug strategy. Aside from confirming the expected binding mode, crystallographic studies in complex with KDM4A revealed the possibility to address the adjacent histone peptide binding site with lipophilic residues and optimization of this part of the molecule led to the conformationally restrained phenylpiperidine systems. Analysis of the effects of the mono-Cl derivative in HeLa cells showed the expected global hypermethylation by immunofluorescence. However, no further in vivo effects were disclosed.[164] The last years have seen a surge of patent literature becoming available with regard to cor- porate efforts in the development of JMJC demethylase inhibitors, especially focusing on the N,N ’-chelators as a central structure. These are amply reviewed elsewhere.[139] As one example, Quanticel Pharmaceuticals has coupled the well-established pyridine-4-carboxylate fragment to imidazole and pyrazole rings to generate structures like 27. This compound

37 Chapter 1. Introduction

was reported to inhibit KDM4C, KDM5A, and KDM5B with IC50 values of less than 0.1 µM. Its methyl ester speciﬁcally increased H3K4 trimethylation in breast cancer cells.[165] Furthermore, EpiTherapeutics has built N,N ’-chelators using the simple aminomethylpyridine scaﬀold represented in example 28. This was also potent at less than 100 nM, but showed only limited selectivity, with strongest inhibition of the KDM5 subfamily. Its esters were potent inhibitors of cell proliferation in a number of cancer cell lines.[166]

1.4.6. Other Mechanism-Based Small Molecule Inhibitors

Apart from these well-established classes of inhibitors with common metal-binding groups (MBGs), other structural scaﬀolds have also been reported. This includes the fragment-like 4- hydroxypyrazole 29 (cf. Fig. 1.17), which was shown to inhibit KDM4C, albeit with relatively weak potency. Given its small size, it may be used as a starting point for the development of more elaborate and, potentially, more portent inhibitors. An SAR study revealed the importance of the 4-hydroxyl and 3-carbonyl groups, identifying this as the MBG.[167] Moreover, the crop protectant and plant growth regulator daminozide 30 was shown to be a potent and 2-OG-competitive inhibitor with selectivity for the KDM2 and KMD7 subfamilies.

Figure 1.17.: Structures of other reported demethylase inhibitors. See text for details.

38 1.4. Inhibitors of Histone Demethylases

Crystallographic and kinetic studies revealed that it chelates the central ferrous ion via its hydrazide function. Daminozide itself is likely of no medicinal use, but given the size of the molecule, its potency is remarkable. In addition, the study contributed a novel MBG and this fragment-like inhibitor may be developed into a more potent one.[168] Using a phenotypic reporter-gene screening assay, Wang et al. identified the structurally unique small molecule JIB-04 31 as an unselective JMJC demethylase inhibitor. It may be considered a tridentate metal ion chelator via the pyridine nitrogen atoms and the hydrazone bridge. Kinetic analyses, however, revealed this compound not to be competitive to the co-substrate 2-oxoglutarate. The (E)-isomer was significantly more potent than the (Z )-form. In cell culture experiments and a breast cancer mouse model, JIB-04 31 showed significant growth inhibition. However, in vivo modulation of methylation levels was not demonstrated.[121] In 2013, Kim et al. identified the natural product tripartin 32 from a Streptomyces sp. strain associated with dung beetle larvae and resolved its chemical structure. In cellular experiments with HeLa cells, it increased H3K9 trimethylation, making it likely a KDM4 subfamily-selective inhibitor of demethylation. However, the mode of action could not be explained.[169] By employing a virtual screening campaign of the National Cancer Institute (NCI) compound database against the crystal structures of KDM4A and KDM4B, NCS636819 33 was identified as a mildly potent inhibitor of these enzymes. It was shown to be competitive to the peptide substrate rather than 2-oxoglutarate and blocked prostate cancer cell viability. Many of the differentially regulated genes upon treatment are controlled by the androgen receptor (AR), highlighting again the strategy of KDM4 subfamily inhibition for treatment of AR-dependent cancer types like prostate cancer (cf. Section 1.2.2.4).[170] Lastly, another study observed that BIX-01294, an inhibitor of the histone methyltransferase (HMT) G9a, and its analog E67 34 also inhibited a human histone demethylase (HDM), namely KDM7A, with a single-digit micromolar IC50 value in a mass-spectrometric assay. Both classes of enzymes bind methyllysine residues either as substrates or as products. An inhibitor mimicking the methyllysine moiety, in this case via the extended aminoalkyl chain, may, therefore, inhibit both enzymes. Cytotoxicity against a number of cancer cell types was observed.[171] However, the question remains to which extent the inhibition of one or the other class of enzyme contributed to the cellular activity.

39 Chapter 1. Introduction

Figure 1.18.: Structures of zinc-ejecting demethylase inhibitors. See text for details.

1.4.7. Inhibitors Based on Zinc Ejection

A last group of small-molecule inhibitors takes advantage of the fact that certain histone demethylases like JMJD2A (KDM4A) contain a structural Zn2+ ion in the active site in proximity to the methyllysine binding site. Zinc-ejecting compounds like the cyclized derivative of disulﬁram 35 or ebselen 36 (cf. Fig. 1.18) disrupt the zinc-binding site and structural inte- grity of the demethylase and also lead to inhibition without being actual competitors of the demethylase’s substrate or co-substrate.[172, 173] These compounds may be considered selective for the KDM4 subfamily containing such a zinc-binding site over other demethylases that do not.

1.4.8. Peptidic Inhibitors

A conceptually very different approach to histone demethylase inhibition is the use of peptidic inhibitors based on the sequence of the enzyme’s substrate, the histone tail. Through investigations of the kinetics and binding affinities of several shortened peptides to different Jumonji enzymes and determining the required lengths that would still be recognized, Lohse et al. developed a truncated histone H3 tail peptide including a trimethyllysine mimic as an unnatural amino acid and coupled it to a bromouracil moiety. By exploiting the binding affinity of the peptide to the active site tunnel and the iron chelating capacities of the bromouracil group, potent inhibition of JMJD2C with four-fold selectivity over JMJD2A could be observed.[174] In a similar approach, another two-component peptidic inhibitor could be obtained that addresses both the substrate binding site as well as the 2-oxoglutarate binding site. From investigations of the co-crystal structure of JMJD2A in complex with the histone substrate, the

40 1.4. Inhibitors of Histone Demethylases correct positioning could be deduced to couple a peptide mimic to the 2-OG competitor N - oxalyl cysteine via a disulﬁde bridge. This resulted in a nanomolar inhibitor both of JMJD2A and JMJD2E with selectivity over members of the KDM6 and KDM7 subfamilies. The desired simultaneous binding to both pockets was conﬁrmed by co-crystal structure analyses.[175] Very recently, another approach was published that used cyclic peptides not as inhibitors that would bind within the active site, but to a remote binding pocket on the surface of the enzyme and, thus, allosterically inhibited the demethylation reaction. By screening a library of ran- domized peptides originating from phage display and subsequent optimization, this yielded a potent JMJD2C inhibitor with a sequence completely unrelated to that of the histone substrate.[176] While these studies revealed potent and somewhat selective inhibitors as well as insights into their binding modes, the peptidic nature of these compounds may hinder further development with regard to their cell permeability or metabolic stability.

1.4.9. Discussion

Since the discovery of JumonjiC domain-containing histone demethylases a decade ago, a substantial number of potential inhibitor scaffolds has been published. However, many of these compounds are derived from known inhibitors of other iron(II)-/2-OG-dependent enzymes or are very general metal ion chelators (e. g. catechols, cf. Section 1.4.2), and are therefore, inherently unselective. Furthermore, for many of the reported structures, the selectivity has not been investigated or only against a small subset of other targets. On the other hand, given the probable redundancy of many demethylases (cf. Section 1.2.2.1), one may wonder how selective an inhibitor really needs to be to obtain meaningful in vivo effects. A central question remains for many of the published inhibitors, that is their function as metal chelators. The different metal-binding groups (MBGs) were discussed in the text. There could be the risk that such compounds might be unselective with regard to other metal-dependent enzymes or generally chelate cytosolic metals. However, in a seminal study, Day and Cohen investigated the selectivity of a broad panel of metal-binding drugs, both clinically approved as well as experimental, against an equally broad panel of metalloproteins. They found that, contrary to popular belief, metal-binding drugs are indeed “quite selective for their intended

41 Chapter 1. Introduction targets” and that “metalloprotein inhibitors are not prone to widespread oﬀ-target enzyme inhibition activity.”[177] The same was observed in a more elaborate study even when possible competing metalloproteins were present.[178] One criticism that can be made regarding many of the published inhibitors is that they do not follow what has become known as Lipinski’s “rule of 5”-criteria for drug-likeness.[179, 180] Because of the metal-binding moieties of the compounds, which require electron pair-donating heteroatoms, many of the compounds are quite polar. Moreover, the carboxylate moiety that is often used to interact with a conserved lysine residue makes these compounds even charged at physiological pH conditions and, therefore, unlikely to cross cell membranes. As histone demethylases are, however, intracellular drug targets, this is a common pitfall. On the other hand, some compounds, especially the hydroxamate-based inhibitors (cf. Section 1.4.4) seem to gain potency and selectivity by the use of extended lipophilic groups and long saturated alkyl chains. This again makes compounds un-druglike because they are too lipophilic and on the limit of solubility in aqueous media, reducing their bioavailability. This trend to increase potency through lipophilicity and at the expense of promiscuity is, unfortunately, a quite commonly observed problem in modern medicinal chemistry. Consequently, the ‘molecular obesity’ arising from this has also been criticized.[181, 182]

Since the initiation of the research project reported in this thesis in 2012, some notable progress has been made by other groups in the development of potent JMJC demethylase inhibitors, some even with subfamily-selectivity (cf. e. g. hydroxamates (Section 1.4.4) or N,N ’-chelators (Section 1.4.5)). Nonetheless, there is still a dire need for potent and selective inhibitors of JumonjiC domain-containing histone demethylases with clear in vivo eﬀects, which can be valuable both as drug candidates as well as biological tool compounds in the study of the function of these enzymes.

42 2. Project Approach

JumonjiC domain-containing histone demethylases (JMJC demethylases) have been implicated in the development of a number of malignant tumor diseases[52, 55, 56] and there is, thus, a high demand for potent and selective inhibitors of these enzymes both as drug candidates as well as biological tool compounds to gain a deeper understanding of their biological roles.[26, 39, 85] Therefore, the main focus of the research project described in this thesis was the discovery and development of novel lead structures suitable for inhibitors of JumonjiC domain-containing histone demethylases, in particular of the subtype JMJD2A (KDM4A). The work can be sub- divided into three diﬀerent subsequent steps of research.

Firstly, in vitro screening assays needed to be developed, optimized, and validated so that they could be used for the screening of compound libraries. While several such assays had already been published (cf. Section 1.3), none were available in the Jung lab at the beginning of the project. Hence, the work consisted in the adaptation of these assays to the subtype JMJD2A as well as the reagents, plate readers, and conditions used in the research group. The formaldehyde dehydrogenase (FDH)-coupled assay was chosen because of its simplicity, widespread use in the literature, and low cost of consumed reagents. As an orthogonal approach suitable for validation and cross-check of the hit compounds, the proprietary LANCEUltra system was chosen (cf. Section 1.3 for descriptions). This assay was also elaborated to be employed with diﬀerent enzymes such as JARID1A (KDM5A) and JMJD3 (KDM6B). For JARID1A, this was a completely novel assay with no prior application of the LANCE system being described in the literature. Furthermore, other assays for validation and characterization purposes needed to be developed such as an FDH-counterscreen assay to rule out compound interference with the coupled enzyme as well as a Ferrozine-displacement assay to study the Fe2+-binding abilities of inhibitors.

43 Chapter 2. Project Approach

Figure 2.1.: Drug discovery workﬂow and strategies for lead structure generation.

Secondly, based on these assays, several compound screening campaigns could be started to generate novel lead structures. With the limitations of scope and capacities of an academic lab, clearly no high-throughput screening (HTS) could be performed. Instead, other more rational approaches to lead generation were employed. Figure 2.1 summarizes a few strategies commonly used in medicinal chemistry.[180] Throughout the different lead structure campaigns (cf. Chapter 4), all these different screening strategies were followed. This included virtual screening, i. e. the computational screening of large libraries of commercially available compounds to identify potential structures worth acquiring and testing. Furthermore, fragment-based drug discovery (FBDD) was employed identifying small metal- binding fragments and elaborating them into larger scaffolds. Moreover, rational drug design was performed based on the known mechanism of action of JMJC histone demethylases, such as screening iron chelators. Analogy-based screening was performed based on other reported lead structures. Upon identification of useful fragments and scaffolds, ‘focused library’ screens were also carried out to generate further hits based on a known lead structure.

44 Thirdly, once the novel lead structures were identiﬁed in screening campaings, the various diﬀerent techniques of lead optimization could be used to develop the hit compounds into more potent and selective derivatives. This involved iterative optimizations based on chemical synthesis of analogs, the establishment and testing of structure-activity relationships, as well as helpful insights into the mode of action of the hit compounds from x-ray crystallography as well as in silico modeling and docking studies. The rounds of optimization were carried out by synthesis in feedback with thorough characterization of the compounds in the previously established in vitro assays.

Lastly, the novel inhibitors thus developed needed to be tested for their biological outcomes e. g. in cell culture studies to verify their potency and on-target eﬀects. This was, however, beyond the scope of the thesis project reported herein, but carried out separately by Dr. Inga Hoﬀmann in her thesis.[78]

3. Assay Development for JMJC Histone Demethylases

3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

In order to generate novel lead structures for inhibitors of JumonjiC domain-containing histone demethylases (JMJC demethylases), an in vitro screening assay platform needed to be established. As outlined previously (cf. Section 1.3), a number of different assay technologies can be envisaged to determine JMJC demethylase activity. As a first screening assay, the formaldehyde dehydrogenase (FDH) enzyme-coupled in vitro assay was chosen because of its reported use in the literature,[72, 106, 112, 126] low cost of reagents, and simplicity in performance. A detailed description of the assay principle can be found in the Introduction (cf. Section 1.3.1 and Figure 1.10(a)). Shortly, the coupled product formaldehyde that is formed in stoichiometric amounts during the demethylation is further oxidized by a second enzyme, formaldehyde dehydrogenase (FDH). This enzyme uses NAD+ as co-factor reducing it to NADH, a fluorescent species that can be quantified via measurements of its fluorescence intensity.

3.1.1. Assay Development and Optimization for JMJD2A

3.1.1.1. Calibration of Fluorescence Intensity

As a ﬁrst step, a calibration curve needed to be set up to verify that ﬂuorescence intensity indeed correlates with the amount of NADH present in a sample and to determine the detection limit. Based on previous literature reports using this assay,[112, 113, 126] it was expected that concentrations of the peptide substrate in the range of 20–50 µM would be used in the demethylation assays. Assuming a conversion rate of 100% and assuming that only mono-demethylation takes place (i. e. the reaction from me3 → me2), this would also give concentrations of formaldehyde and, consequently, of NADH in the same range.

47 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.1.: Calibration curve for ﬂuorescence intensity of NADH at λex = 330 nm and

λem = 460 nm. Shown are mean ± s. d. of triplicates.

Such a calibration curve is shown in Figure 3.1. To this end, commercially available NADH was dissolved and pipetted onto microtiter plates in different concentrations. Fluorescence intensity (FI) was measured at an excitation wavelength of λex = 330 nm and emission wavelength of λem = 460 nm. This calibration curve shows that FI increases strictly linearly with regard to NADH concentration and that it can, therefore, be used to quantify the amount present in the sample. This is true for the entire expected concentration range. Only at very small concentrations (< 5 µM) does the standard deviation become unacceptably large and prohibits a meaningful measurement. Notably, even if no NADH is present in the sample, a small FI can be measured. This means that values measured during an activity assay will need to be corrected for this effect using a blank measurement. In order to test whether components of an assay buffer (like the HEPES buffer, cf. Section 7.1.1.4) would have an influence on fluorescence, NADH was also diluted in assay buffer and measured likewise. As can be seen from Fig. 3.1, this is not the case and FI values for measurements in water as well as in buffer overlay perfectly. Lastly, the stability of fluorescence intensity (FI) over time was evaluated by repeating the measurement of the same plate after incubation at 37◦C for one hour. As depicted in Fig. 3.1,

48 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

Figure 3.2.: Example of a kinetic FDH-coupled enzyme activity assay using diﬀerent batches of recombinant JMJD2A. Shown are mean ± s. d. of triplicates.

there is a slight decrease in FI values when comparing the measurement at t = 0 and t = 1 h. This is likely due to degradation of NADH in solution. However, this needed to be taken into account in the development of the FDH-coupled assay as it limits the possible incubation times.

3.1.1.2. Enzyme Activity Measurements

Once it was established that fluorescence intensity correlates with the amount of NADH and, thus, also with that of formaldehyde, enzyme activity measurements could be started. Figure 3.2 shows a representative example of a kinetic demethylase activity assay (cf. Section 7.1.2.2). For this, different batches of enzyme were diluted to the same nominal working concentration of 0.10 mg/mL and incubated at 37◦C with a substrate solution containing all required components, that is ferrous sulfate, ascorbic acid, formaldehyde dehydrogenase (FDH), NAD+, 2-oxoglutarate (2-OG), and a peptide substrate. Fluorescence was recorded in short intervals at λex = 330 nm and λem = 460 nm, the fluorescence wavelengths of NADH.

The enzyme working concentration of 0.10 mg/mL represents the optimal compromise between a possibly strong increase in ﬂuorescence and an enzyme consumption as low as possible as determined from a number of preliminary kinetic experiments.

49 Chapter 3. Assay Development for JMJC Histone Demethylases

It becomes evident that fluorescence increases over time corresponding to the ever increasing amount of formaldehyde HCHO that is released and transformed into formic acid, concomi- tantly releasing NADH. After roughly one hour (3600 s) this increase reaches a plateau. A possible reason may be that the enzyme loses activity or degrades under the assay conditions. This could apply to both the demethylase JMJD2A as well as formaldehyde dehydrogenase (FDH). Furthermore, as was observed in Fig. 3.1, the fluorescent product NADH also degrades over time. As such, the incubation time for all future experiments and, in particular the inhibition curves, was set to be 60 minutes for this assay. In principle, the observed plateau may also be due to the fact that all peptide substrate has already been transformed into the demethylated species. However, when one compares the observed fluorescence values in Fig. 3.2 to the calibration curve (cf. Fig. 3.1), one obtains an increase ∆ FI of roughly 20,000 units over one hour corresponding to a concentration of NADH of approximately 25 µM, translating to 25 µM of formaldehyde. In these experiments, the peptide concentration was 35 µM. Assuming mono-demethylation, this corresponds to a typical conversion rate observed in all experiments of 60–70%. Importantly, different batches of enzyme all showed very comparable activity when diluted to the same nominal concentration (cf. Fig. 3.2). Such an experiment was, thus, repeated with every new batch that was obtained throughout this research project in order to verify that their activities were comparable (cf. Section 7.1.2). In only two instances, batches had to be rejected because they did not meet these criteria.

3.1.1.3. Substrate Solution Components

When Sakurai et al. optimized and miniaturized the FDH-coupled assay,[112] they had already noticed that the observed enzyme activity can vary dramatically depending on which assay components were added in what order. In particular, they found that the demethylase JMJD2E did not tolerate any iron(II) species or ascorbic acid already present in the buﬀer. When these were stored together, enzyme activity dropped to zero over only a few hours due to reagent degradation in the presence of enzyme. Consequently, they suggested that the enzyme dilution be stored by itself in buﬀer alone while all other assay components (ferrous sulfate, ascorbate, 2-oxoglutarate, peptide, FDH, and NAD+) be mixed in a multi-component ‘substrate solu-

50 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays tion’ to be added to the enzyme only upon initiation of the demethylase reaction.[112] In the following section, the optimization of such a substrate solution is described with regard to the concentrations of its components. Firstly, the required concentration of the peptide substrate was tested. Figure 3.3(a) shows the result of a kinetic experiment analogous to that represented in Fig. 3.2, but with varying concentrations of the peptide substrate H3K9me3 (residues 7–14). The observed fluorescence over the incubation time of one hour was transformed into a reaction velocity as ∆ FI v = (3.1) ∆ t and plotted against the concentration of the peptide. Using the standard Michaelis-Menten equation for enzyme kinetics[11] v · [peptide] v = max , (3.2) KM + [peptide] one can obtain the typical saturation curve as shown in Figure 3.3(a). This yields a KM value of 13.0 ± 6.3 µM for the peptide and a maximum velocity of 8.1 ± 1.2 RFU/s. As a compromise between a quick enzymatic transformation and low substrate consumption, a concentration of 35 µM was chosen for all subsequent experiments. As becomes clear from Figure 3.3(a), no significant increase in reaction velocity would be possible if more substrate were used. Furthermore, it is known that some JMJC KDMs can catalyze oxidation of 2-OG to succinate in the absence of a prime peptide substrate.[183] In order to ascertain that the measured increase in fluorescence was indeed due to peptide demethylation and not to other background reactions, a rather large concentration of peptide was chosen. Recently, Lohse et al. have reported that the enzymatic activity of some JMJC demethylases like JMJD2A and JMJD2C varies greatly if other modifications than the substrate are present on the same peptide.[68] In that spirit, it was attempted to establish the assay with a longer peptide containing trimethylation both at the K4 as well as the K9 position. This is summarized in Figure 3.3(b). In contrast to the literature report, this gave virtually identical results with a KM value of 13.0 ± 5.3 µM for the peptide and a maximum velocity of 9.2 ± 1.1 RFU/s. This may possibly be due to the fact that the improvement in enzymatic activity is mainly observed for full-length proteins containing further reader domains, while in our experiments, only the catalytic domain of JMJD2A was used. As no improvement could be obtained, experiments with the double modified peptide were discontinued.

51 Chapter 3. Assay Development for JMJC Histone Demethylases

(a) H3K9me3 peptide residues 7–14. Sequence: ARK(me3)STGGK-NH2

(b) H3K4me3K9me3 peptide residues 1–24. Sequence: ARTK(me3)QTARK(me3)-

STGGKAPRKQLATKA-NH2

Figure 3.3.: Michaelis-Menten kinetic curves for two peptide substrates in the FDH- coupled assay with JMJD2A. Shown are mean ± s. d. of triplicates.

52 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

Similarly, the concentration of the co-substrate 2-oxoglutarate in the substrate solution was varied and a Michaelis-Menten curve established as depicted in Figure 3.4(a). This gave a

KM value, i. e. the concentration at which half the maximum reaction velocity is observed, of 29 ± 11 µM for 2-oxoglutarate. Consequently, a final assay concentration of 50 µM was used for all other assays as no significant increase in reaction velocity can be obtained even if more co-substrate is used. In the same way, it was attempted to optimize the concentration of ferrous sulfate added to the mixture. However, as the results shown in Figure 3.4(b) indicate, enzyme velocity was independent of the concentration of exogenous ferrous ions. The reason may be that the recombinant catalytic domain of JMJD2A indeed already contained a ferrous ion bound in the active site or at least enough iron(II) dissolved in the enzyme storage buffer. This is surprising as enzymes purified via the nickel/His-tag system like the JMJD2A used in this study are commonly expected to lose their incorporated metal ions during the purification procedure. In order to maintain a constant level of ferrous ions in every assay independent of the batch of enzyme used and the amount of residual iron it might still contain, it was decided to sup- plement the substrate solution with 10 µM ferrous sulfate for all experiments. This is well

(a) 2-oxoglutarate (2-OG) (b) FeSO4 Figure 3.4.: Determination of optimal assay concentrations for 2-oxoglutarate (2-OG)

and ferrous sulfate (FeSO4) in the FDH-coupled assay with JMJD2A. Shown are mean ± s. d. of triplicates.

53 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.5.: Comparison of kinetic FDH-coupled activity assays with FeSO4 and

(NH4)2Fe(SO4)2 using JMJD2A. Shown are mean ± s. d. of triplicates.

in excess of the concentration of enzyme used (ﬁnal: 1.75 µM), ensuring that the apparent activity of every enzyme preparation would remain constant. One concern with the use of ferrous sulfate in the buﬀer is its limited stability in aqueous stock solutions in contact with air. Over time, iron(II) readily oxidizes to iron(III), which can be seen as a formation of brown turbid solutions or even a precipitate (typically Fe(OH)3). It was reported that solutions of ferrous ammonium sulfate (NH4)2Fe(SO4)2 would be more stable in aqueous solution.[112] A kinetic experiment employing substrate solutions containing either

FeSO4 or (NH4)2Fe(SO4)2 in parallel is depicted in Figure 3.5. No noticeable diﬀerence can be observed with regard to enzyme activity. On the other hand, the stability of such solutions until oxidation could be observed was also only very slightly improved. Therefore, all activity assays were continued with regular ferrous sulfate solutions. However, it was decided that these aqueous stock solutions would be stored for a maximum of one week and then have to be replaced with fresh solutions from newly weighed-in solid FeSO4 · 7 H2O. Furthermore, all substrate solutions were supplemented with a large excess of ascorbic acid as reducing agent in order to ascertain that all iron ions in solution were indeed in the +2 oxidation state. The amount of formaldehyde dehydrogenase (FDH) in the substrate solution was also varied. Considering the assay principle, it follows that all formaldehyde that is formed needs

54 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

Table 3.1.: Optimized composition of the substrate solution for the FDH-coupled JMJD2A assay. See text for details.

ﬁnal assay concentration substrate solution concentration ascorbic acid 100 µM 400 µM

FeSO4 10 µM 40 µM FDH 0.001 Unit/µL 0.004 Unit/µL NAD+ 500 µM 2000 µM 2-oxoglutarate 50 µM 200 µM

peptide H3K9me3 7–14 35 µM 140 µM

to be rapidly oxidized so that the corresponding amount of NADH is formed. This coupled enzymatic reaction must not be the rate-limiting step. This, together with the consideration that FDH is commercially available at relatively little cost led to the decision to use a rather large concentration of 0.001 Unit/µL. Greater concentrations led to no noticeable improvement, while smaller amounts actually changed the observed kinetic parameters, indicating that the transformation of HCHO may be slowing down the entire process (data not shown). The same consideration goes for the co-substrate of formaldehyde dehydrogenase (FDH), i. e. NAD+. In order to ensure rapid conversion, a large excess of 500 µM was used, following a literature recommendation.[112] The optimized composition of the substrate solution resulting from these experiments is summarized in Table 3.1. Upon initiation of the assay, the substrate solution would be added to the enzyme preparation in a ration of 1:3 (6 µL into 18 µL), requiring the substrate solution to be 4X concentrated compared to the ﬁnal assay concentrations.

3.1.1.4. DMSO Tolerance

One important consideration to be made in the development of this assay was the possible eﬀect of organic solvents like dimethyl sulfoxide (DMSO) on enzyme activity. Commonly, test compounds like organic small molecules are not soluble in water alone, but rather are stored in DMSO stock solutions. When these are added to the assay mixture, a residual concentration

55 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.6.: Eﬀect of DMSO content on performance of the FDH-coupled activity assay with JMJD2A. Shown are mean ± s. d. of triplicates.

of DMSO remains in the system, which can have an influence on the stability of the enzyme and, possibly, lead to denaturation and, therefore, loss of activity. To this end, kinetic experiments were performed, in which different amounts of DMSO were added to the substrate solution. When comparing the results of each test run to that of a control containing no DMSO (cf. Fig. 3.6), only a small reduction in activity could be observed even for quite elevated concentrations of DMSO. In the optimized version of the FDH-coupled assay used for inhibition experiments (cf. Section 7.1.2.1), wells usually contained a final concentration of 2% DMSO.

3.1.1.5. Microtiter Plates

Another factor to consider in the development of such an in vitro screening assay is the question of which kind of microtiter plates should be used. Considering the required consumption of enzyme and other reagents, it was clear that only 384-well plates with a working volume of 20 µL per well could be used rather than 96-well plates (working volume 100–200 µL). In the initial optimization, three different kinds of plates were used, including white and black OptiPlates-384 and black ProxiPlates-384 F (all from PerkinElmer). Typically, fluorescence measurements are carried out on black plates in order to minimize noise from light reflection

56 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

Table 3.2.: Comparison of diﬀerent microtiter plates for the FDH-coupled JMJD2A assay. ‘+’ denotes a positive property, while ‘–’ denotes a property that makes these plates unsuitable for the assay.

signal intensity standard deviation OptiPlate-384 white + + + + OptiPlate-384 black – – – – ProxiPlate-384 F black + –

and/or scattering. Surprisingly, the best results were obtained on white OptiPlates, both with regard to the signal intensity as well as signal variation from well to well. These results are summarized qualitatively in Table 3.2. On black ProxiPlates, ﬂuorescence intensity could be measured. However, the signal was always decreased compared to white OptiPlates and exhibited a much greater variability. When black OptiPlates were used for an activity assay, only noise was observed and no meaningful ﬂuorescence signal could be obtained. All data from experiments described herein were obtained on white OptiPlates.

3.1.1.6. Assay Buﬀer

All literature references to the FDH-coupled assay and, in particular, those describing its development,[72, 112] use a 2-[4-(2-hydroxyethyl)piperazin-1-yl] ethanesulfonic acid (HEPES)-based assay buffer for all dilution steps at a pH value of 7.50. In order to determine the optimal pH for the FDH-coupled assay, kinetic measurements like in Section 3.1.1.2 were run using a variety of different buffers, including the HEPES buffer adjusted to different pH values, an acetate/acetic acid-based buffer, phosphate buffers at different pH values, and a bicarbonate/carbonate-based buffer. Surprisingly, good results with comparable enzyme activity could be obtained for any of the HEPES-based buffers, in a range from pH 5.50 to 8.50, demonstrating that both JMJD2A and FDH are quite robust enzymes tolerating a range of different pH conditions. Only the very extreme conditions of acidity

(NaOAc/AcOH, pH 4.00) and basicity (NaHCO3/Na2CO3, pH 11.00) led to complete loss of activity. However, these buffers were also based on different buffer substances than HEPES,

57 Chapter 3. Assay Development for JMJC Histone Demethylases which may also have had an eﬀect on the stability or activity of JMJD2A and FDH or of another assay component such as the biomolecules NAD+ and NADH. In conclusion, the literature-recommended buﬀer system of 50 mM HEPES at pH 7.50 was kept and used for all assays described herein (cf. Section 7.1.1.4).

3.1.2. Assay Validation – Statistical Robustness

One very important factor in assay development is its statistical robustness, i. e. the com- parability of data when the assay is run many times in a row. In order to use this assay for inhibition experiments with small-molecule inhibitors, it would be necessary to create a large enough assay window between an uninhibited positive control (100%) and a fully inhibited or inactive enzyme (0%), in between which the activity in the presence of a compound should lie. This requires the difference between 100% and 0% to be substantially large and the variation, with which they are measured, to be quite small. In 1999, Zhang et al. have introduced a statistical parameter to describe the suitability of an assay for such purposes and termed it the Z0 factor.[184] It is defined as 3 · σ + 3 · σ Z0 = 1 − + − , (3.3) |µ+ − µ−| where σ+ and σ− denote the standard deviations of the positive and negative control, respectively, and µ+ and µ− their mean values. The denominator, thus, represents the separation band. In a hypothetical perfect assay, the separation band would be infinitely large and the measurement variation infinitely small, making Z0 equal to 1. In their paper, Zhang et al. defined different quality categories of assays based on the Z0 value.[184] To test the robustness and, hence, suitability of the FDH-coupled assay developed so far, an experiment was performed where 30 wells were filled with identical positive controls (i. e. enzyme and substrate solution as optimized above) and another 30 wells were filled with negative controls (i. e. enzyme alone without any substrate), incubated at 37◦C for one hour, measured under the optimized conditions, and the difference in fluorescence intensity between t = 0 and t = 1 h taken as indicator of enzyme activity. The outcome is depicted in Fig. 3.7. As can be seen from the plotted values, the separation band between positive and negative control is quite large, allowing for the determination of activity values that lie between 100% and 0%. Furthermore, the variation within measurements of the negative control is reasonably

58 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

Figure 3.7.: Determination of statistical robustness of the FDH-coupled assay for JMJD2A. Wells 1–30: positive controls, wells 31–60: negative controls.

small, while for the positive control a certain divergence is observed. Using equation 3.3, one obtains 3 · 2,468 + 3 · 768 Z0 = 1 − = 0.61. (3.4) 23,799 − (−786) A value of Z0 = 0.61 falls into the category of 1.0 > Z0 > 0.5, which, according to Zhang et al., makes this assay set-up an ‘excellent assay’.[184]

3.1.3. Assay Validation – Reference Inhibitors

As the next step in assay validation, it was tested whether the inhibition of JMJD2A by reference compounds could be reproduced in this FDH-coupled assay. As two examples, the previously published reference inhibitors N -oxalyl glycine (NOG) 2 and pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6 were used.[126] Kinetic experiments were performed using JMJD2A and the optimized substrate solution in the presence or not of compound in diﬀerent concentrations. The results are summarized in Figure 3.8.

59 Chapter 3. Assay Development for JMJC Histone Demethylases

(a) NOG 2 (b) 2,4-PDCA 6

Figure 3.8.: Validation of the FDH-coupled activity assay for JMJD2A using reference inhibitors N -oxalyl glycine (NOG) 2 and pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6 in kinetic experiments (a,b) and representative inhibition curve from end-point measurements for 2,4-PDCA 6 (c).

Figure 3.8(a) shows the results for reference inhibitor N -oxalyl glycine (NOG) 2, reportedly a relatively weak in vitro inhibitor.[126] The positive control shows the expected increase in ﬂuorescence intensity (FI) over time, while the negative control without substrate does not change markedly. As the concentration of NOG in solution increases, the increase in FI over time slows down and the enzyme is, apparently, inhibited. At a concentration of 400 µM of

60 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

NOG, the rate of increase in FI is only half that of the positive control, meaning that the enzyme is approximately 50% inhibited. Likewise, the same picture can be observed using reference compound pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6, except that here, much smaller concentrations are sufficient to reach a comparable effect of enzyme inhibition. This coincides with literature reports describing 2,4-PDCA 6 as a much more potent in vitro inhibitor.[126] If such experiments are not run in kinetic mode continuously taking measurements of FI, but using only two measurements at t = 0 and t = 1 h, one can perform end-point inhibition experiments (see Section 7.1.2.1 for details). Then, the difference in FI between the two time points can be taken as a measure of enzyme activity and expressed relative to that of an uninhibited positive control (100%) and a negative control, where no substrate is present (0%). When these relative activities are plotted against the concentration of test compound, one obtains inhibition curves like Fig. 3.8(c). As outlined in detail in Section 7.1.2.1, these values can be fitted against a 4-parameter logistic curve, where one plateau is ideally 100% and the other

0%. From the ﬁt parameters, a so called IC50 value can be calculated, i. e. the concentration, at which 50% inhibition is observed. This value can be used to compare diﬀerent compounds tested against the same target in the same assay. For 2,4-PDCA 6, an IC50 value of 0.83 µM is obtained, which is well in agreement with literature reports for the closely related subtype JMJD2E.[126]

3.1.4. Possible Adaptation to other Enzymes and Substrates

Once the FDH-coupled enzyme activity assay was successfully optimized for the demethylase

JMJD2A and its substrate, the H3K9me3 peptide (residues 7–14), attempts were made to test the scope and limitations of this assay. Initial experiments showed that this assay system could easily be transferred to two commercially available demethylases, namely the catalytic domains of JMJD2A and JMJD2C, both accepting the same H3K9me3 substrate. The kinetic parameters were comparable and inhibition studies with test compounds gave similar results. While it would have been interesting to develop a JMJD2C demethylase assay for selectivity screening purposes, the high enzyme consumption in this assay together with the cost of commercial JMJD2C quickly prohibited

61 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.9.: Example of a kinetic FDH-coupled enzyme activity assay using full-length histones as substrate for JMJD2A. Shown are mean ± s. d. of triplicates.

further use of this system. For selectivity investigations, the LANCEUltra assay was later developed, employing far less enzyme, thus making it feasible to be used also with commercial demethylases (cf. Section 3.3). In terms of substrates, it was already shown that the FDH-coupled enzymatic assay for

JMJD2A works with both the H3K9me3 peptide (residues 7–14) as well as an H3K4me3K9me3 peptide (residues 1–24) as discussed in Section 3.1.1.3 and Figure 3.3. Furthermore, the assay could also be run with a longer substrate peptide of the H3 sequence, that is H3K9me3 (residues 1–15), which was obtained from a commercial source. Although not fully characterized, its kinetic properties were, as expected, comparable to that of the H3K9me3 (residues 7–14) peptide. As it showed no improvement and the shorter peptide was more affordable, however, development of an assay with such a substrate was discontinued. Lastly, an effort was undertaken to test this activity assay not with synthetic peptides based on the histone sequence, but with full-length native histones as substrates. This was based on the previously reported successful development of an assay using native histones as substrates for LSD1.[111] Full-length histones are commercially available as a purified protein extract from calf thymus. A representative result is summarized in Fig. 3.9.

62 3.1. Formaldehyde Dehydrogenase (FDH)-Coupled Assays

Surprisingly, a certain activity was observed for different concentrations of histones (approx. 40% of peptide control), although it never reached the levels of the demethylation reaction observed for JMJD2A with a synthetic peptide. Furthermore, this apparent activity was completely independent of histone concentration in a wide range from 1.25 µg/mL to 1.50 mg/mL, as judged from several kinetic experiments. This makes the observed data likely an artifact. The fact that such a histone preparation from a biological source constitutes a mixture of different histone proteins, with a variety of different modifications on several methylation sites where the degree of methylation is unknown, complicates the understanding of the observed effects. It is possible that some methylation marks that are substrates for JMJD2A are present, but it would be rather complicated to judge to which extent they are. As such, the development of a histone-based activity assay was terminated and experiments were continued with the synthetic peptides described above, for which sequence and degree of methylation are unequivocally known.

63 Chapter 3. Assay Development for JMJC Histone Demethylases

3.2. FDH-Counterscreening Assay

The formaldehyde dehydrogenase (FDH)-coupled assay for JMJD2A is based on the increase in fluorescence intensity (FI) of NADH, which is formed by oxidation of the coupled demethylation product formaldehyde by the secondary enzyme FDH. An inhibitor of JMJD2A would, thus, be recognized by reduction of measured fluorescence intensity. However, this assay principle has an inherent drawback as test compounds, which inhibit the detection enzyme FDH rather than the actual test enzyme JMJD2A, would also lead to a reduction in FI and would, therefore, be recognized as positive hits. This requires a method to distinguish whether a test compound is indeed a true JMJD2A inhibitor or an FDH inhibitor. To this end, an FDH-counterscreening assay was developed. Here, the secondary enzyme formaldehyde dehydrogenase is incubated with exogenous formaldehyde and its co-factor NAD+ to yield formic acid and fluorescent NADH. This reaction must still procede uninhibited even in the presence of test compounds if they are to be considered veritable demethylase inhibitors. In order to establish such an assay, calibration experiments were performed, in which formaldehyde dehydrogenase was incubated with different amounts of formaldehyde at 37◦C and the FI measured at λex = 330 nm and λem = 460 nm. A range of formaldehyde concentrations was chosen that corresponds to that expected to be formed in a typical FDH-coupled enzymatic demethylase assay (cf. Section 3.1.1.2). The results are summarized in Figure 3.10. Clearly, FDH was able to transform exogenous formaldehyde into the fluorescent product and the fluorescence intensities correlate with those previously observed in demethylase activity assays. The obtained signals increase in a linear fashion with increasing concentrations of HCHO. Quasi-instantaneous measurements (t = 0) led to higher values, but also greater variability. However, if the plate was incubated for a time of at least 15 min on a shaker at 37◦C, very stable results were obtained, irrespective of incubation time. Based on this encouraging result, a formaldehyde concentration of 40 µM and an incubation time of 30 min was chosen for the counterscreening assays. In these experiments, a sample containing enzyme buffer would be supplemented with the corresponding amount of formaldehyde and pre-incubated with DMSO or test compounds. Then, a substrate solution containing FDH and its co-factor NAD+ would be added, initiating the oxidation reaction. The substrate solution was based on that used in the demethylase activity assays. In order to make

64 3.2. FDH-Counterscreening Assay

Figure 3.10.: Calibration experiments incubating formaldehyde dehydrogenase (FDH) with exogenous HCHO and NAD+. Shown are mean ± s. d. of triplicates.

the results comparable, it also contained FeSO4, ascorbic acid, and 2-oxoglutarate, although they are clearly not required for the formaldehyde oxidation reaction. In an effort to reduce consumption of costly assay components like the peptide substrate, this was left out. The composition of the final substrate solution for this assay is summarized in Table 3.3. All test wells were also supplemented with 2% DMSO. Similar to the procedure for the demethylase activity assay, one part of substrate solution was added to three parts of buffer/HCHO mixture and it, thus, needed to be 4X concentrated. Further details for the FDH-counterscreening assay are summarized in Section 7.1.3. Unfortunately, no specific chemical inhibitors of formaldehyde dehydrogenase (FDH) are reported in the literature that could have been used to validate the assay. Instead, as a negative control, wells were used that contained only formaldehyde, but not the substrate solution. As no FDH and no NAD+ is, thus, present in solution, the FI in these values is negligible compared to the blank wells as expected. One example of how this FDH-counterscreening assay could be used to verify JMJC demethylase inhibitors is depicted in Figure 3.11 using reference compound pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6, a definitive JMJC inhibitor, in parallel with a hydroxamate-based test

65 Chapter 3. Assay Development for JMJC Histone Demethylases

Table 3.3.: Composition of the substrate solution for the FDH-counterscreening assay. See text for details.

ﬁnal assay concentration substrate solution concentration ascorbic acid 100 µM 400 µM

FeSO4 10 µM 40 µM FDH 0.001 Unit/µL 0.004 Unit/µL NAD+ 500 µM 2000 µM 2-oxoglutarate 50 µM 200 µM

Figure 3.11.: Examples of test compounds in the FDH-counterscreening assay. PDCA – reference compound pyridine-2,4-dicarboxylic acid 6, Cmpd – test compound. Shown are mean ± s. d. of duplicates.

66 3.2. FDH-Counterscreening Assay compound as a putative confounder. The observed FI for the positive control wells containing FDH, NAD+, and HCHO was taken as a measure of activity and all values are expressed relative to this control. As can be seen from the experiment summarized in Fig. 3.11, these test compounds have only a very small effect on the oxidative activity of formaldehyde dehydrogenase. In the case of 2,4-PDCA, relative activity was always around 100% and it seemed to even increase with increasing concentrations of 2,4-PDCA. However, this is likely an artifact from the measurement uncertainty. This result is not surprising as 2,4-PDCA is known to be a veritable inhibitor of Fe(II)-/2-OG-dependent enzymes like JMJC demethylases.[126] As for the other test compound, a slight drop in FDH activity was observed at very high concentrations. However, this is well beyond the concentration range, in which a test compound would be considered a positive ‘hit’ in the demethylase inhibition assay. Insofar, the slight modulation of FDH activity is irrelevant. Therefore, both test compounds, which strongly attenuated fluorescence intensity in the FDH- coupled demethylase assay, were shown to be bona fide demethylase inhibitors rather than confounding FDH inhibitors. This observation holds true for all other tested compounds discussed in this project as lead structures for JMJC inhibitors. Deviations are discussed in the context relevant to a given structure (cf. Section 4.2).

In summary, through simple modiﬁcations of the FDH-coupled assay, an easy and quick veriﬁ- cation system has been established that allows for double-checking of hit compounds from the FDH-coupled assay for their true demethylase inhibitory potential.

67 Chapter 3. Assay Development for JMJC Histone Demethylases

3.3. LANCEUltra Assays

The formaldehyde dehydrogenase (FDH)-coupled enzyme activity assay developed so far has several inherent drawbacks as already discussed in the Introduction (cf. Section 1.3). The problem of false positive hits due to FDH inhibition can be overcome using the FDH-counterscreening assay developed in Section 3.2. However, the problem of interference with the measurement, particularly by auto-fluorescent test compounds and fluorescence quenchers, remains. It was, therefore, desirable to develop a second orthogonal assay system that is based on a different physical principle in order to validate hits discovered in the FDH-coupled assay and to test compounds that could not be tested in this assay. To this end, the proprietary antibody-based LANCEUltra technology (developed by Perkin- Elmer)[122, 123] was chosen because of the advantages allowed by measurements of time-resolved fluorescence resonance energy transfer effects (TR-FRET) and because of the different detection principle. While the FDH-coupled assay quantifies demethylase activity by detecting a coupled product (HCHO), the antibody-based system allows to measure demethylation directly on the peptide substrate. This very different measurement principle, thus, allows for independent verification of assay hits. A detailed description of the assay principle and its advantages is given in Section 1.3.2 and Figure 1.11.

3.3.1. Assay Development and Optimization for JMJD2A

By its very nature, such an assay requires a highly specific antibody that can recognize a histone peptide of a given sequence and distinguish between the degree of methylation on a given lysine residue. In this example, it would have to bind the product of demethylation, the H3K9me2 peptide, but not the substrate, the H3K9me3 peptide. Such a high degree of specificity is rarely achieved[119] and represents one of the caveats of antibody-based assay systems. As one of the components of this proprietary assay system, PerkinElmer provides a europium-labeled antibody with the described specificity. How it was developed and how this specificity could be accomplished is not disclosed, however. To determine its suitability for the JMJD2A demethylase assay, its specificity was verified using biotinylated peptides carrying either the trimethyl or dimethyl mark. The result is summarized in Figure 3.12.

68 3.3. LANCEUltra Assays

Figure 3.12.: Speciﬁcity of the α-H3K9me2 antibody used in LANCEUltra assays when incubated with diﬀerent peptides and the detection reagent ULight. Shown are mean ± s. d. of duplicates.

Pleasingly, this antibody showed outstanding selectivity and was able to fully discriminate between the trimethylated peptide, which is not bound at all, and the dimethylated peptide, for which a concentration-dependent increase in signal intensity was observed. Notably, the concentrations of peptide that can be detected are in the upper nanomolar range, drastically reducing reagent consumption. Up until a concentration of approximately 500 nM, the signal increases linearly (R2 = 0.9927) with peptide concentration, allowing for the use of this experiment as a calibration curve. Only at much larger concentrations does the signal increase reach a plateau, likely due to saturation of the antibody. Furthermore, in spite of the extremely small amounts of reagent used in this experiment, the precision, with which they can be determined, is remarkable as demonstrated by the very small standard deviations. For the linear part of the calibration curve ([peptide] < 500 nM), the linearity allows for the establishment of the calibration equation (cf. Section 7.1.4.4)

LANCE − 200 [H3K9me ] = · 1 nM. (3.5) 2 275

69 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.13.: Kinetic experiments using the LANCEUltra assay with 400 nM of

H3K9me3 substrate to determine optimal enzyme concentration and incubation time. Shown are mean ± s. d. of triplicates.

Taking into account the result from the FDH-coupled activity assay that a typical conversion rate by JMJD2A lies between 60% and 70% (cf. Section 3.1.1.2), it was decided to use a concentration of 400 nM of biotinylated H3K9me3 peptide as substrate. This would result in

240–280 nM of H3K9me2 and, thus, LANCE signals, which would allow for a large enough assay window that could precisely be measured.

Next, the activity of JMJD2A in this assay was tested in order to determine whether H3K9me2 formed by enzymatic demethylation could indeed be recognized by this assay system. To this end, several diﬀerent concentrations of enzyme were incubated for diﬀerent times with 400 nM of H3K9me3 peptide and then stopped by addition of the detection mix. This detection mix containing the antibody and ULight also contains a large excess of the general metal ion chelator ethylenediaminetetraacetic acid (EDTA) in order to complex all iron ions in solution and stop further demethylation reactions while antibody binding takes place. The result of such a kinetic experiment is summarized in Fig. 3.13. It should be noted that unlike in the FDH-coupled assay, continuous measurements of reaction progress (on-line mea-

70 3.3. LANCEUltra Assays surements) are not possible in the LANCEUltra system as it requires another incubation step after the end of the reaction when the antibody solution is added. This means that every data point in Figure 3.13 represents an individual experiment, later summarized in the same graph, explaining some of the larger deviations observed here. Pleasingly, the results depicted in Figure 3.13 were as expected. With increasing incubation time, the observed LANCE signal increased for all enzyme concentrations due to the progression of the demethylation reaction and the formation of more H3K9me2 peptide to be bound by the antibody. In addition, greater concentrations of enzyme also increased the rate, at which H3K9me2 was formed leading to larger LANCE signals in shorter times. Remarkably, concentrations as low as 1 nM of enzyme led to a detectable change in methylation levels after one hour. Expectedly, control experiments, where no enzyme or no peptide was added, led to no increase in signal intensity above blank values regardless of incubation time. For the ﬁnal JMJD2A assays, an enzyme concentration of 60 nM was chosen with an incubation time of 45 min. This represented the best compromise between limited reagent consumption (in particular enzyme and peptide substrate) and a large assay window. It should be noted, though, that the required enzyme concentrations in this assay are far below that used in the FDH-coupled system (optimized assay: ﬁnal concentration = 1.75 µM).

3.3.2. Assay Validation – Statistical Robustness

In an eﬀort to evaluate the statistical robustness of the LANCEUltra assay developed with the conditions described so far, an experiment was performed analogous to that described in Section 3.1.2 and Figure 3.7. Here, 26 wells were ﬁlled with positive controls containing JMJD2A and

400 nM of the H3K9me3 substrate and compared to 30 wells with negative controls. As this assay always requires the presence of a biotinylated peptide, to which the ULight-streptavidin complex can bind, negative controls could not be constructed like in the FDH-coupled assay by leaving out the substrate solution. Instead, negative controls in LANCEUltra contain the full substrate solution, but have the enzyme replaced by buﬀer. The result of this statistical test is summarized in Figure 3.14. Pleasingly, the data showed a large separation band between the positive and negative controls that allows for precise measurements of enzyme inhibition by test compounds. Moreover, the

71 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.14.: Determination of statistical robustness of the LANCEUltra assay for JMJD2A. Wells 1–30: positive controls, wells 31–60: negative controls.

data variability in both controls is relatively small. Using equation 3.3,[184] one obtains 3 · 3,280 + 3 · 430 Z0 = 1 − = 0.64. (3.6) 42,063 − 10,969 for this assay. Again, a Z0 value of 0.64 falls into the category of an ‘excellent assay’ as judged by the criteria established by Zhang et al.[184]

3.3.3. Other Assay Parameters

Encouraged by these remarkable results and the outstanding reproducibility of the assay data, several other aspects of this in vitro assay were not optimized, partly also due to the very high cost of the proprietary assay components, in particular the labeled antibody. Instead, these parameters were adapted from technical application notes published by PerkinElmer.[122] Parameters that remained unoptimized include the use of white 384-well OptiPlates with a total assay volume of only 10 µL.[122] Furthermore, the use of a HEPES buﬀer at pH 7.50 is recommended supplemented with 0.01% of bovine serum albumin (BSA).[122] This ubiquitous

72 3.3. LANCEUltra Assays

Table 3.4.: Composition of the substrate solution for LANCEUltra JMJD2A assays.

ﬁnal assay concentration substrate solution concentration ascorbic acid 100 µM 400 µM

FeSO4 5 µM 20 µM 2-oxoglutarate 1 µM 4 µM

peptide biot-H3K9me3 1–21 400 nM 1600 nM

and relatively cheap hydrophobic protein is a common additive in antibody-based assays that is used to block the surfaces of hydrophobic plastic components like pipette tips, reaction tubes, and microtiter plates in order to minimize unspecific interactions that would lead to greater background signals if, for example, the antibody were to bind to the walls of the wells rather than the peptide. Moreover, the concentrations of the remaining components of the substrate solution were not optimized and adapted from the values that were recommended for the very closely related subtype JMJD2C.[122] Likely, the affinity of JMJD2A to its co-substrate 2-oxoglutarate (2-OG) is comparable. The composition of the final substrate solution is given in Table 3.4. Just like in the FDH-coupled assay (cf. Section 3.1.1.3), the substrate solution was added in a ratio of 1:3 to the enzyme solution to initiate the reaction, requiring it to be 4-fold concentrated. Another factor that remained unoptimized is the composition of the detection mix and the incubation time required for antibody binding after the demethylation reaction was stopped. It was used as described in a technical note[122] and Table 3.5. The detection mix was added

Table 3.5.: Composition of the detection mix for LANCEUltra JMJD2A assays.

ﬁnal assay concentration detection mix concentration

Eu-labeled α-H3K9me2 antibody 2 nM 4 nM ULight-streptavidin adduct 50 nM 100 nM EDTA 1 mM 2 mM

73 Chapter 3. Assay Development for JMJC Histone Demethylases in an equal volume to the assay mixture in every well, requiring it to be two-fold concentrated. After termination of the reaction by addition of the detection mix, plates were incubated for another 60 min at room temperature as recommended and then measured. Further details about the LANCEUltra assay are given in the experimental section 7.1.4.

3.3.4. Assay Validation – Reference Inhibitors

As was the case for the FDH-coupled assay (cf. Section 3.1.3), an experiment was performed in order to test whether inhibition of demethylation by a reference inhibitor could be reproduced in this assay. For this purpose, the previously reported inhibitor pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6[126] was used. A representative result of an inhibition experiment in the LANCEUltra assay is summarized in Figure 3.15. In this assay system, a total volume of only 10 µL per well is used, resulting in a DMSO content of 5%. In order to compensate for a possible loss in activity because of this high DMSO concentration, the same amount of DMSO was also added to all control wells (positive and negative) as well as the blank.

Figure 3.15.: Validation of the LANCEUltra activity assay for JMJD2A using reference inhibitor 2,4-PDCA 6. Shown are mean ± s. d. of duplicates.

74 3.3. LANCEUltra Assays

Given that the LANCE signal correlates linearly with the concentration of the resulting demethylated H3K9me2 peptide (cf. Figure 3.12), the blank-corrected LANCE signals could be taken as a simple measurement of enzyme activity. The corrected signal for wells in the presence of diﬀerent concentrations of inhibitor was expressed relative to that of a positive control (100%) and a negative control containing no enzyme (0%) and plotted as shown in Figure 3.15. Data analysis was performed as previously described (Section 3.1.3) and led to an IC50 value of 34 ± 7 nM for 2,4-PDCA 6.

This is notably less than the IC50 value obtained in the FDH-coupled assay, underlining again that only IC50 values obtained in the same assay may be compared. The reason for the lower value is that the LANCEUltra assay also uses dramatically less enzyme, meaning that the amount of a compound required to inhibit it is also smaller. Furthermore, the concentration of the co-substrate 2-oxoglutarate (2-OG), to which most inhibitors are competitive, is also smaller, meaning that less compound is required to outcompete it. However, this value is in [122] good agreement with an IC50 value published for JMJD2C, i. e. 43 nM.

3.3.5. Comparison of JMJD2A Activity Assays

So far, two in vitro assay systems to determine activity of the histone demethylase JMJD2A and its inhibition by test compounds have been established, i. e. the FDH-coupled assay (cf. Section 3.1) and the LANCEUltra system (cf. Section 3.3). Their principle characteristics are summarized in Table 3.6 in order to highlight common features and differences. The most notable difference is the drastically reduced consumption of reagents, in particular of the enzyme JMJD2A in the LANCEUltra assay. Here, one uses only about 1/30 of the enzyme concentration required in the FDH-coupled assay in order to obtain a detectable signal. In addition, the concentration of the peptide substrate required is also drastically smaller. Based on the principle of the LANCEUltra assay, the peptide needs to be biotinylated, so that a different (and more expensive) substrate is required. Remarkably, the concentration of the enzyme co-substrate 2-oxoglutarate (2-OG) is also much lower than in the FDH-coupled assay. This, together with the smaller concentration of enzyme, means that IC50 values generated in the LANCEUltra assay are generally smaller than those observed in the FDH-coupled assay.

75 Chapter 3. Assay Development for JMJC Histone Demethylases

Table 3.6.: Comparison of the FDH-coupled and LANCEUltra JMJD2A in vitro assays.

FDH-coupled assay LANCEUltra assay JMJD2A 1.75 µM 60 nM H3(7–14)K9me biot-H3(1–21)K9me peptide substrate 3 3 35 µM 400 nM 50 mM HEPES, pH 7.50, 50 mM HEPES, pH 7.50, buﬀer 0.01% Tween-20 0.01% Tween-20, 0.01% BSA

FeSO4 10 µM 5 µM ascorbic acid 100 µM 100 µM 2-oxoglutarate (2-OG) 50 µM 1 µM DMSO content 2% 5% 500 µM NAD+ 2 nM Eu α-H3K9me ab detection system 2 0.001 Unit/µL FDH 50 nM ULight-streptavidin microtiter plates white OptiPlate-384 white OptiPlate-384 BMG POLARstar PerkinElmer EnVision plate reader and method ﬂuorescence intensity (FI) TR-FRET assay volume 20 µL 10 µL total assay time approx. 1.5 hours approx. 2.5 hours

The dramatically reduced consumption of costly assay reagents would make the LANCEUltra assay an ideal primary screening method more suitable than the FDH-coupled assay. However, the problem with this assay is the very expensive europium-labeled highly specific antibody that is required for the detection system as well as the costly proprietary dye ULight, which prohibit larger screening campaigns based on this assay system. Moreover, the FDH-coupled assay can, in principle, readily be adapted to other demethylase subtypes for selectivity investigations (cf. Section 3.1.4) as they all produce formaldehyde, which is detected in this assay. On the other hand, the LANCEUltra assay would require a different labeled antibody for every new histone modification that is to be studied, further increasing the cost for this screening platform. In addition, the LANCEUltra assay is also

76 3.3. LANCEUltra Assays somewhat more laborious and more time-consuming as it requires a second incubation step after the demethylation reaction (cf. Table 3.6). Based on these considerations, the FDH-coupled assay was chosen as a primary screening platform for JMJD2A inhibitors as it allows for a ‘quick and dirty’ screening method. If test compounds showed promising activity in this assay, they were validated using the LANCEUltra assay. Only if they showed inhibitory potency in both assays, which are based on very different physical read-out principles, were they considered true hits. An exception to this workflow was made for compounds, which could not be tested in the FDH-coupled assay due to their interference with the measurement principle, usually because of auto-fluorescence. These were then directly tested and characterized using the LANCEUltra assay (cf. e. g. Section 4.5).

3.3.6. Adaptation to other Enzymes and Substrates

The very small enzyme consumption in the LANCEUltra assay system has made it possible for this platform to be used with other enzymes, which were only commercially available, together with their respective substrates. This has allowed for investigations into the selectivity of certain test compounds with regard to their inhibitory potency on diﬀerent targets. For such selectivity investigations, the H3K27me3 demethylase JMJD3 (KDM6B) was chosen because of its reported involvement in oncogenesis[104, 185] and because it had been shown that the development of selective inhibitors for this subfamily is possible.[158] In addition, an assay for the H3K4me3 demethylase JARID1A (KDM5A) was developed because representatives of the KDM5 subfamily were also convincingly demonstrated to be involved in cancer.[87]

3.3.6.1. JMJD3 (KDM6B)

First, the development of a LANCEUltra-based assay for JMJD3 (KDM6B) and its substrate

H3K27me3 was attempted. The different steps in the assay development are summarized in Figure 3.16. Other assay parameters were adapted from the JMJD2A assay set-up. PerkinElmer also provides a proprietary europium-labeled antibody with declared specificity for the demethylation product H3K27me2. Like in the development of the LANCEUltra assay for JMJD2A (cf. Section 3.3.1), this was verified using different biotinylated peptides. The result in Figure 3.16(a) clearly shows that the antibody has the desired specificity for

77 Chapter 3. Assay Development for JMJC Histone Demethylases

(a) Speciﬁcity of the α-H3K27me2 antibody (b) Kinetics of H3K27me3 demethylation by JMJD3

Figure 3.16.: Steps in the development of a LANCEUltra-based assay for JMJD3 and

its substrate H3K27me3. See text for explanations.

the product binding only the dimethylated peptide with outstandingly large signal intensities, while discriminating it from the trimethyl peptide, which would be used as the substrate. As a control, the biotinylated H3K9me3 peptide was included in this experiment, which was, as expected, not bound at all due to the diﬀerent amino acid sequence not recognized by the antibody. At very high concentrations of the H3K27me2 peptide, a saturation of the antibody

78 3.3. LANCEUltra Assays can be observed, while the signal increase is otherwise linear. Based on this result, a peptide concentration of 400 nM was chosen for subsequent activity assays. The result of a kinetic assay varying the amount of JMJD3 and the incubation time is summarized in Figure 3.16(b). Surprisingly, the demethylase JMJD3 seemed to be much less active than its related subtype JMJD2A. Signiﬁcant turnover was only observed at longer incubation times of 120 min and with greater enzyme concentrations. This may be related to the preparation procedure of this commercially available enzyme or its purity. Nonetheless, using 50 nM of JMJD3 with an incubation time of 120 min allowed for the development of an assay with reasonable signal intensities. Whether the chosen assay conditions ([peptide] = 400 nM, [enzyme] = 50 nM, incubation time = 2 hours) allowed for a robust assay, was again tested in a statistical Z0 test as previously outlined (cf. Section 3.3.2). The result for JMJD3 is depicted in Figure 3.16(c). Even though the observed signal intensities are relatively small compared to the peptide calibration curve due to the low turnover activity of JMJD3, it can still be used for a robust assay because the values can be measured quite precisely. Applying equation 3.3 to these values gives

3 · 1,892 + 3 · 494 Z0 = 1 − = 0.61, (3.7) 23,893 − 5,567 which again falls into the category of 1.0 > Z0 > 0.5, making this set-up an ‘excellent assay’.[184] Lastly, this assay was validated by determining whether inhibition by a reference compound was possible. For JMJD3, a highly potent and speciﬁc inhibitor was reported in the literature, i. e. GSK-J1 23a (cf. Section 1.4.5).[158] The result for inhibition of JMJD3 by GSK-J1 23a is depicted in Figure 3.16(d). With very high precision for this experiment in duplicate, an inhibition curve could be obtained like for the previously discussed inhibition experiments with

JMJD2A (cf. Section 3.3.4). Here, an IC50 value of 128 nM is obtained for GSK-J1 23a, which is in a comparable range to the reported value of 60 nM in the literature.[158] However, this was also determined using a very diﬀerent assay technology (AlphaScreen). In conclusion, a very robust in vitro screening assay has been developed for JMJD3 using its substrate H3K27me3 based on the LANCEUltra principle, which can be used for inhibitor screening and characterization experiments.

79 Chapter 3. Assay Development for JMJC Histone Demethylases

3.3.6.2. JARID1A (KDM5A)

Encouraged by the good results for the LANCEUltra assay for demethylases JMJD2A and JMJD3, another assay was developed for the KDM5 subfamily representative JARID1A. Its main substrate is H3K4me3 (cf. Table 1.1). The diﬀerent steps in the assay development are summarized in Figure 3.17. Other assay parameters were adapted from the JMJD2A and JMJD3 assay set-ups. PerkinElmer also provides a proprietary europium-labeled antibody with a declared dual speci-

ficity for the peptides H3K4me2 and H3K4me1. Like in the development of the LANCEUltra assays for JMJD2A and JMJD3 (cf. Section 3.3.1 and 3.3.6.1), this was verified using different biotinylated peptides. The result in Figure 3.17(a) clearly shows that the antibody has the desired specificity for the product binding only the dimethylated peptide with exquisite selectivity over the substrate peptide H3K4me3 and several other peptides with modifications at different lysine positions, which were added to this experiment for comparison. Surprisingly, though, the antibody was apparently already saturated at a concentration of approximately 100 nM, after which no further signal increase was observed. This is likely due to the very high binding affinity of this antibody to its epitope. The experiment was, thus, repeated with further diluted peptide as represented in Figure 3.17(b). This shows the expected linear increase in signal with increasing peptide concentration. Based on these data, a peptide concentration of

100 nM of H3K4me3 was chosen for subsequent activity assays assuming that partial turnover by JARID1A would yield concentrations of H3K4me2 in the linear range. The result of a kinetic assay varying the amount of JARID1A and the incubation time is summarized in Figure 3.17(c). Already at a concentration as little as 10 nM of enzyme could a time-dependent signal increase be observed, which was even stronger for greater amounts of JARID1A, reaching a plateau at about 25–50 nM, after which no further increase could be obtained. For the development of an assay with reasonable signal intensities and limited enzyme consumption, a concentration of 25 nM with an incubation time of 45 min was, therefore, chosen. Whether the chosen assay conditions ([peptide] = 100 nM, [enzyme] = 25 nM, incubation time = 45 minutes) allowed for a robust assay, was again assessed in a statistical Z0 test as previously outlined (cf. Section 3.3.2). The result for JARID1A is depicted in Figure 3.17(d).

80 3.3. LANCEUltra Assays

(a) Speciﬁcity of the α-H3K4me2/1 antibody (b) Speciﬁcity of the α-H3K4me2/1 antibody

0 (c) Kinetics of H3K4me3 demethylation by JARID1A (d) Z test for statistical robustness

Figure 3.17.: Steps in the development of a LANCEUltra-based assay for JARID1A

and its substrate H3K4me3. See text for explanations.

Here, the robustness of the assay is reduced compared to the other assays previously developed because of the greater variability of the positive control values. However, when inserting the values from this experiment into equation 3.3, one obtains

3 · 4,771 + 3 · 311 Z0 = 1 − = 0.54, (3.8) 39,210 − 6,411

81 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.18.: Validation of the LANCEUltra activity assay for JARID1A using reference inhibitor 2,4-PDCA 6. Shown are mean ± s. d. of duplicates.

which is arguably smaller than what was obtained for JMJD2A and JMJD3, but is still in the range of 1.0 > Z0 > 0.5, allowing for this set-up to be considered an ‘excellent assay’.[184] As the last step of the assay development workflow, this assay was also validated using a reference inhibitor, in this case pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6, which had been reported to be a potent inhibitor of demethylases of the KDM5 subfamily before.[144] The result of this inhibition experiment is depicted in Figure 3.18. This yielded an IC50 value of 100 nM for 2,4-PDCA 6. The somewhat increased variability in controls and measured data is also reflected by the larger error bars and standard deviation compared to the precision that is typically achieved in LANCEUltra assays (cf. Section 3.3.4). Nonetheless, one can still obtain interpretable inhibition data suitable for comparison of the inhibitory potency of different compounds when tested against JARID1A. Overall, an excellent in vitro screening assay could be developed for JARID1A using its substrate H3K4me3 based on the LANCEUltra principle, which can be used for inhibitor screening and characterization experiments. Notably, this is the first such assay for JARID1A with no previous publication of the use of the LANCEUltra principle for this demethylase or the H3K4 methylation site.

82 3.4. DELFIA-Based Assay for JMJD2A

3.4. DELFIA-Based Assay for JMJD2A

As another in vitro screening assay using antibodies to detect peptide modifications, it was attempted to develop a DELFIA-based assay for JMJD2A. Just like the LANCEUltra assays (cf. Section 3.3), it relies on antibody binding of a demethylated peptide, but differs greatly in how this binding is transformed into a measurable signal and eventually quantified. The theoretical principle is outlined in Section 1.3.2. The development of the broad-spectrum JMJC demethylase inhibitor methylstat 21b (cf. Section 1.4.4) is the only literature reference, where a DELFIA-type assay has previously been developed and used for compound characterization.[124] This was, therefore, used to guide the development of the DELFIA protocol for JMJD2A. While many assay parameters were adapted from the literature,[124] several optimizations were attempted both with regard to enzyme, peptide, and antibody concentrations and incubation times. The anti-H3K9me2 detection antibody was from a different vendor for availability reasons, while the secondary europium-labeled anti-mouse antibody was the one marketed by PerkinElmer and commonly used in these assays. A typical result of optimized conditions is represented in Figure 3.19. This shows that the detection antibody used can in fact discriminate the dimethylated product peptide from the trimethylated substrate. However, the signal increase is only quite small (less than 4-fold) when compared to that observed in LANCEUltra assays (for example cf. Fig. 3.12). Consequently, the amount of H3K9me2 formed by enzymatic demethylation is almost not detectable when compared to the H3K9me3 substrate. This problem could be somewhat alleviated when larger concentrations of enzyme (up to

0.50 mg/mL) were used, but still the separation band between positive and negative controls never yielded a useful assay window. Furthermore, the reproducibility of these data points was greatly diminished. In addition, the point of the development of an antibody-based in vitro assay was to decrease enzyme consumption compared to the FDH-coupled assay ([enzyme] =

0.10 mg/mL) and not to increase it even further.

Moreover, in order to obtain such values, a relatively high concentration (0.90 µg/mL or more) of the costly antibody needed to be used. There is also a substantial batch-to-batch variation in data obtained from diﬀerent streptavidin-coated strip plates. The age and storage conditions of these plates can lead to degradation of the coated protein and, therefore, dramatically

83 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.19.: Representative example of data obtained in DELFIA measurements. Pep-

tide concentrations were 500 nM and 0.01 mg/mL JMJD2A enzyme was used. Shown are mean ± s. d. of duplicates.

inﬂuence the peptide binding capacity of the plates or, even worse, of individual wells on the same plate, resulting in incomparable values. Even by variation and optimization of all assay parameters, it was never possible to develop a statistically robust and reproducible in vitro assay to monitor JMJD2A activity based on the

DELFIA system with this anti-H3K9me2 antibody from abcam. This was due in large parts to the fact that the antibody binds both the H3K9me3 substrate and the H3K9me2 product peptides with only little difference in signal. This again raises the central question of antibody specificity for epigenetic modifications and points to the problem that many do not have the declared selectivity and, therefore, have substantial problems in their usability as revealed in a published evaluation study.[119] It may be possible to develop such an assay with a different batch of this antibody or with a different antibody altogether, but this problem may also explain why Ref. 124 is the only reported use of the DELFIA system for JMJC demethylases. Since a different antibody-based assay system with drastically reduced enzyme consumption amenable even to several different demethylases was already available (LANCEUltra, cf. Section 3.3), the development of a DELFIA-based platform was consequently terminated.

84 3.5. Iron Chelation Assay Based on Ferrozine Displacement

3.5. Iron Chelation Assay Based on Ferrozine Displacement

Lastly, an assay was developed that can measure the relative binding capacities of test compounds to iron as this is assumed to be the central mode of action for most inhibitors. This assay was to be based not on measurements of enzymatic activity, which can be inﬂuenced by many factors including protein degradation or reactive compounds covalently binding to the enzyme, but on a direct measurement of the formation of compound–iron complexes. This can aid in the investigation of a central problem for some inhibitors, i. e. to delineate whether they bind iron(II) ions within the active site of the enzyme or act by ejection of the metal ions and formation of complexes in solution. This assay was, therefore, based on Ferrozine, a spectrophotometric dye developed in the 1970s to selectively detect iron ions in analytical samples like potable water.[186, 187] Its structure and the basic principle is outlined in Figure 3.20(a). While Ferrozine (5,6-diphenyl-3-(2-pyridyl)- 1,2,4-triazine-4,4”-disulfonic acid monosodium salt) itself only gives a colorless to pale yellow solution in water, it forms an intensely colored purple 3:1 complex in the presence of iron,‡ which can be used to detect very small concentrations of ferrous and ferric ions even in the presence of many competing ions.[186] This is also corroborated by UV/vis spectra recorded in the recommended acetate/acetic acid buﬀer at pH = 4.50 of the Ferrozine reagent alone (cf. Figure 3.20(b)) and in complex with iron (cf. Figure 3.20(c)). The reagent alone gives no detectable absorbance in the visible range even at very elevated concentrations, while the complex results in an absorption band with a maximum wavelength of λmax = 563 nm. Absorption in the green band corresponds to a purple color of the complex. When analyzing the observed absorbance and using the standard Beer-Lambert law A = ε · c · d, (3.9)

‡The commercial Ferrozine reagent was obtained as the monosodium salt as depicted in the structure (cf. Figure 3.20(a)). In this form, the ligand is neutral and the complex with a divalent iron ion would carry a 2+ +2 charge, i. e. [Fe(Ferrozine)3] as drawn and as used in the following text. In aqueous solution and in the buﬀer used, however, both sulfonic acid groups can be expected to be deprotonated and the sodium ions to be disscoiated, which would change the overall charge to −4.

85 Chapter 3. Assay Development for JMJC Histone Demethylases

(a) Structure of Ferrozine and basic principle for the photometric detection of iron(II)

II 2+ (b) UV/vis spectrum of Ferrozine (c) UV/vis spectrum of [Fe (Ferrozine)3] complex

Figure 3.20.: Principle of the Ferrozine-based detection of iron(II) (a) and UV/vis spectra of the reagent alone (b) and in complex with iron(II) (c). Spectra were recorded in pH 4.50 buﬀer at [Ferrozine] = 5 mM and [Fe2+] = 25 µM in a d = 1 cm cuvette. See text for explanations.

one obtains a molar extinction coeﬃcient ε of A 0.6611 L ε = = = 26444 , (3.10) c · d µmol mol · cm 25 L · 1 cm which is in agreement with a previously reported literature value[186] and which explains the intense color of the complex. Unbound Ferrozine does not interfere with the measurement (cf. Figure 3.20(b)). Therefore, a large excess of the reagent was always used to ensure that the concentration of iron added in solution was equal to the concentration of the colored complex.

86 3.5. Iron Chelation Assay Based on Ferrozine Displacement

In order to minimize reagent consumption and to drastically increase throughput of these experiments, the measurement of absorbance was transferred from cuvettes onto microtiter plates in 96-well format (transparent SpectraPlate-96 MB from PerkinElmer) using a working volume of only 100 µL as opposed to 3 mL in cuvettes. Measurements were performed on a plate reader using a filter with a center wavelength of λ = 570 nm and a bandwidth of ±10 nm, thus encompassing the maximum absorption wavelength of the complex (cf. Figure 3.20(c)). Figure 3.21 depicts a simple calibration experiment demonstrating that absorbance at λ = 570 nm indeed correlates with the concentration of iron ions present in solution even at very small concentrations and how precisely it can be measured. The absorbance values here are always smaller than those observed for measurements in cuvettes. This is because a working volume of 100 µL results in a filling height of only roughly 0.3 cm in every well. As this is the optical layer, through which light passes, the absorbance is also smaller than in 1 cm cuvettes (cf. equation 3.9). Nonetheless, iron content can easily be determined as shown in Figure 3.21. This is also true with a reduced molar excess of Ferrozine ligand (c = 500 µM) as shown. Subsequent optimizations revealed that in order to obtain stable values for the absorbance of the complex, at least a molar ratio of 6:1 of Ferrozine to Fe2+ should be employed, meaning double the amount that would stoichiometrically be required for complex formation. Similar to the enzyme activity assays (cf. Section 3.1.1.4), the influence of the organic solvent dimethyl sulfoxide (DMSO) on data was investigated as most test compounds to be studied would likely be soluble only in DMSO, but not in water. The different characteristics of the solvent such as the decreased polarity and different optical properties of DMSO may influence absorbance measurements. However, it was found that even samples containing up to 20% DMSO showed no significant difference in their absorbance values and the final assay was, thus, performed with 10% DMSO. Furthermore, variations of the assay buffer from the acetate/acetic acid-based buffer at pH = 4.50 as in the literature[186] to neutral or even basic HEPES-based buffers revealed that the best data were indeed obtained in the recommended buffer at pH = 4.50. This is in contrast to the original report claiming the usability of this reagent over a broad pH range of 4–11[186] or even as low as pH 2.[187] Once this iron(II) detection system was established, the actual competition experiments using test compounds could be carried out. The general principle here was that test compounds,

87 Chapter 3. Assay Development for JMJC Histone Demethylases

Figure 3.21.: Calibration curve for iron(II) in the Ferrozine-based photometric assay at λ = 570 nm at a concentration of 500 µM Ferrozine measured in microtiter plates. Shown are mean ± s. d. of triplicates.

which bind ferrous ions such as in the active site of JMJD2A, would displace Ferrozine out of II 2+ the [Fe (Ferrozine)3] complex and, therefore, lead to decolorization. By variations of the ratio of Ferrozine to test compounds, information about the relative binding strength of the compound could be obtained. To this end, dilution series of compounds were usually added to samples on a microtiter plate containing 50 µM of iron(II), pre-incubated, and then detected with 300 µM of Ferrozine. After a short equilibration period, absorbance was measured at λ = 570 nm and blank-corrected. Full details for these assays are given in Section 7.1.5. This assay was validated using two typical general metal ion chelators, namely deferoxamine 37 and ethylenediaminetetraacetic acid (EDTA) 38. Their chemical structures are represented in Figures 3.22(a) and 3.22(b), respectively. A typical result of the displacement assay is depicted in Figure 3.22(c). For this, positive controls containing only Fe2+ and Ferrozine were measured (defined as 100% absorbance) as well as negative controls lacking Ferrozine (defined as 0%) and samples in the presence of compounds referenced accordingly. Curve fitting allowed for the determination of an IC50 value, which gives qualitative information about the binding strengths

88 3.5. Iron Chelation Assay Based on Ferrozine Displacement

(a) Deferoxamine mesylate 37 (b) EDTA 38

II 2+ (c) Displacement of Ferrozine from the [Fe (Ferrozine)3] complex by test compounds

Figure 3.22.: Structures of metal chelators deferoxamine 37 (a) and EDTA 38 (b) and displacement of Ferrozine in the iron(II)-binding assay by these test compounds (c). Shown are mean ± s. d. of triplicates.

of test compounds. In this case, deferoxamine 37 can be seen to displace Ferrozine from the complex as it is a potent metal ion chelator. Moreover, it binds ferrous ions more strongly than EDTA 38 as a 4-fold smaller concentration was required for the same decolorization eﬀect. Unfortunately, it was diﬃcult to reproduce this kind of result for other test compounds and inhibitors of the JMJC demethylases. Even when the co-substrate 2-oxoglutarate (2-OG) was

89 Chapter 3. Assay Development for JMJC Histone Demethylases used, which clearly binds ferrous ions, no such displacement could be observed. One reason may II 2+ [187] be that the [Fe (Ferrozine)3] complex itself is very stable (pKD = 15.5 ) and requires an even more strongly binding ligand to displace Ferrozine from it. Deferoxamine 37 and EDTA 38, which both form hexadentate complexes, may be capable of displacing Ferrozine due to the entropic gain upon complex formation, whereas bidentate ligands like 2-OG and the developed demethylase inhibitors of other structure classes are not. Another diﬃculty that was observed for some test compounds is their limited solubility at the high concentration range required for the assay, which is typically much higher than that, in which potent inhibition is observed in enzymatic activity assays. Moreover, some test compounds form colored complexes with iron(II) ions themselves and, thus, interfere with the measurement of absorbance at a visible wavelength. While these factors limit the applicability of this assay to a larger range of test compounds, for some substance classes, a simple and quick, inexpensive, and robust test system has been established to determine iron-binding properties and to compare the relative binding strengths of diﬀerent compounds.

90 4. Novel Inhibitors of JMJC Histone Demethylases

4.1. Co-Substrate Analogs as Inhibitors

In a first attempt to develop novel inhibitors of JumonjiC domain-containing histone demethylases (JMJC demethylases), a small library of very simple structural analogs of the enzyme co-substrate 2-oxoglutarate (2-OG) was tested against JMJD2A. It was reasoned that they may also be capable of binding to the central ferrous ion in these enzymes and, thus, competitively displacing 2-OG and thereby inhibiting the demethylation reaction. A variety of structurally related compounds was acquired and tested for their in vitro JMJD2A inhibition in the FDH-coupled and LANCEUltra assays. Their structures and the assay results are summarized in Table 4.1. As discussed in the development of the screening platform (cf. Section 3.3.5), compounds were generally tested in the simpler FDH-coupled assay first and only hit compounds of promising inhibitory potency or structural features also evaluated in the LANCEUltra assay. The latter uses far less enzyme and co-substrate, explaining why the observed IC50 values for test compounds are generally smaller in that assay. This is also the reason why some compounds appear to have no inhibitory potency at the highest tested concentration (200 µM or 400 µM) in the FDH-coupled assay, but can show mild inhibition in the LANCEUltra assay. The simple amide analog of 2-OG, N -oxalyl glycine (NOG) 2, was previously shown to be a very weak JMJD2E inhibitor.[126] As such, it was also used as reference inhibitor in the development of the FDH-coupled assay (cf. Figure 3.8(a), Section 3.1.3). The only mild inhibition is reflected by the very high IC50 value for this compound, which is close to the highest concentration tested in FDH-coupled assays so that no full inhibition curve could be obtained. As this compound was only such a weak inhibitor, it was not pursued any further. Dr. Ludovica Morera (Jung group) attempted to synthesize a fluorinated analog of 2-oxoglutaric acid 39, assuming that this may displace the unlabeled version of the co-substrate and thereby

91 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.1.: Inhibition data of co-substrate analogs as potential inhibitors of JMJD2A.

IC50 values are mean ± s. d. of duplicate experiments. n. i. – no inhibition. n. t. – not tested.

FDH-coupled assay LANCEUltra assay Compound Structure IC50 / µM IC50 / µM

N -oxalyl glycine 2 ∼400 n. t.

39 n. i. n. i.

(R)-5 n. i. 32.9 ± 6.5

glyphosate 40 n. i. n. t.

41 n. i. n. i.

zileuton 42 n. i. n. t.

FR-900098 43a n. i. 14.4 ± 0.3

act as an inhibitor. While the purity of this substance could not be ascertained, preliminary data suggested no such eﬀect (cf. Table 4.1). Notably, kinetic experiments revealed that 39 was also not recognized as a substrate by JMJD2A. This may be due to the high electronega- tivity of the ﬂuorine atom, which facilitates formation of the carbonyl hydrate species, which would lead to a structure that cannot be cleaved as required by the enzymatic mechanism. The reduced forms of the co-substrate, that is (S)- or (R)-hydroxyglutarate 5, have also been reported to be mildly potent inhibitors of some 2-OG-dependent enzymes.[143] For availability reasons, only the (R) enantiomer could be obtained and tested. Compound (R)-5, however, exhibited only weak inhibition in the LANCEUltra assay and no inhibition at concentrations up to 400 µM in the FDH-coupled assay. In studies using a JMJC demethylase from C. elegans, it was already determined that the (S) enantiomer was more potent than the (R) enantiomer.[143]

92 4.1. Co-Substrate Analogs as Inhibitors

The phosphonate-based herbicide glyphosate 40 exhibits some structural similarity to the co- substrate 2-OG considering the phosphonic acid as a bioisosteric replacement of the terminal carboxylate in 2-OG. However, this compound also proved to be inactive in inhibiting JMJD2A. The reason may be that it is missing the α-keto group present in 2-OG that is required for complexation of the metal ion. Moreover, very simple metal chelators based on the hydroxamate moiety were tested for their potential to inhibit JMJD2A. This includes the fragment hydroxyurea 41 as well as the licensed asthma drug zileuton 42, an inhibitor of the enzyme arachidonate 5-lipoxygenase, which is likewise an Fe2+-dependent oxygenase. However, both compounds were again devoid of any inhibitory potency. This underlines the notion that the active sites of JumonjiC domain- containing histone demethylases are highly conserved and can speciﬁcally recognize their substrate 2-oxoglutarate. This is recognized and bound with very high aﬃnity so that displacement by only loosely related analogs, which do not undergo any further interactions with the active site, is not possible. One exception to this rule is compound 43a, known as FR-900098, which had been reported as an inhibitor of a reductoisomerase required for isoprenoid biosynthesis in many bacteria and the malaria pathogen P. falciparum and, therefore, as a potential antimalarial drug.[188] This molecule can also be considered a structural analog of 2-OG as it is able to complex metal ions through its internal hydroxamic acid and to interact with a basic active site residue through its negatively charged phosphonate group. However, the distance between the metal-binding group and the negative charge is increased by one CH2 spacer. This compound potently inhibited JMJD2A in the LANCEUltra assay with an IC50 value of 14 µM. While this fragment is still a more potent inhibitor of its original target, it seemed promising to incorporate this as an oxoglutarate-mimic in the development of other inhibitors (cf. Section 4.4).

The screening of very simple structural analogs of the co-substrate 2-oxoglutarate was surprisingly unsuccessful in unveiling new leads for potent JMJD2A inhibitors. However, it has led to the discovery of fragment 43a, which could be incorporated as a novel co-substrate mimic into more elaborate lead structures. To date, no phosphonate-based inhibitors of JumonjiC domain-containing histone demethylases have been reported.

93 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

4.2. Tetrazolyl Hydrazide-Based Inhibitors

4.2.1. Discovery

The inhibitor structures discussed in this section were discovered by a combination of analogy- based screening of compounds based on a previously reported inhibitor and fragment-based drug discovery (FBDD). The results of the screening and subsequent investigations have been published in large parts in Ref. 189. Recently, it was discovered that the plant growth regulator daminozide 30 is a selective and potent inhibitor of JumonjiC domain-containing histone demethylases of the KDM2 and KDM7 subfamilies and the N,N -dimethylhydrazide function in this inhibitor[168] was introduced as a metal-binding group (MBG) (cf. Section 1.4.6). Inspired by this, a library of small fragment- like hydrazides was screened for potential inhibition of JMJD2A (KDM4A). A diverse collection of small molecules for FBDD was available from the Link group (Insti- tute of Pharmacy, Ernst-Moritz-Arndt-University Greifswald). These were manually inspected for compounds containing a hydrazide function and subjected to in vitro screening against JMJD2A in the FDH-coupled assay. The selection of the thirteen hydrazides screened is depicted in Figure 4.1. Unlike the previously reported daminozide 30, these compounds are unmethylated, but the CO−NH−NH2 motif likely retains the metal-binding capacities. Surprisingly, compounds 44a–j were all devoid of any inhibitory potency (less than 50% inhibition at the highest screening concentration, i. e. 200 µM or 400 µM) in spite of their metal-binding group and the diversity of the substituted aryl functions. Only the compounds containing a tetrazole ring, i. e. 44k, 44l, and 44m showed any notable activity. These compounds were then further characterized using the FDH-coupled as well as LANCEUltra assays and the results of inhibition experiments are summarized in Table 4.2. Notably, these three compounds were relatively potent inhibitors in the FDH-coupled assay and even more so in the LANCEUltra assay. The diﬀerence can be attributed to the drastically smaller concentration of enzyme required for the LANCE assay technology and, more importantly, to the smaller concentration of the competitive co-substrate 2-oxoglutarate (2-OG). The compounds with an extended spacer between the MBG and the tetrazole moiety, i. e. 44l and 44m, exhibited somewhat reduced potency, which may be the result of an entropic penalty

94 4.2. Tetrazolyl Hydrazide-Based Inhibitors

Figure 4.1.: Overview over the 2-substituted acetohydrazides screened for JMJD2A inhibition based on the Link IV compound screening plates.

due to the greater flexibility of the alkyl chain spacer or of suboptimal positioning of the two parts of the molecule. Considering the small size and low molecular complexity of 44k, its potency is remarkable. This compound and a number of analogs were re-synthesized by Nicole Rüger(Link group) and this authentic sample confirmed the data from the screening plates. As the syntheses of these compounds were performed elsewhere and were not part of this thesis project, they are not discussed here. For details regarding the syntheses, the reader is directed to Ref. 189.

Table 4.2.: IC50 values of tetrazolyl hydrazide-based inhibitors against JMJD2A. Data are mean ± s. d. of at least two independent experiments.

FDH-coupled assay LANCEUltra assay Compound Structure IC50 / µM IC50 / µM

44k 46.6 ± 0.9 2.38 ± 0.37

44l 64.1 ± 12.8 8.32 ± 1.08

44m 69.8 ± 15.5 7.68 ± 1.54

95 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Figure 4.2.: Control compounds to verify the structure-activity relationship (SAR) for tetrazolyl hydrazide inhibitors of JMJD2A.

Based on the results from the screening collection, it seemed likely that the acidic tetrazole moiety was essential to the potency of these compounds. In order to test this hypothesis, a control compound was synthesized, in which the tetrazole ring was N -methylated, masking its acidic properties (compound 45, cf. Figure 4.2). Indeed, this compound was again devoid of any inhibitory activity. Likewise, a control compound, which contained the tetrazole ring, but had the metal-binding hydrazide replaced by a simple ethyl ester (compound 46), also exhibited no notable inhibition. Taking this structure-activity relationship (SAR) together revealed that only those compounds, which contain both the acidic tetrazole ring as well as the metal-binding hydrazide function, were potent inhibitors of JMJD2A, with the shortened derivative 44k being the most potent one.

4.2.2. Competitivity of Inhibition Towards 2-Oxoglutarate

In order to elucidate the mode of action of these inhibitors, competitivity investigations were undertaken by performing the LANCEUltra in vitro assay with varying concentrations of the enzyme co-substrate 2-oxoglutarate (2-OG). Details for these types of assays are outlined in Section 7.1.4.4 and the results of these experiments using 44k are summarized in Figure 4.3 and Table 4.3. As becomes evident from Figure 4.3(a), the activity of JMJD2A in the absence of an inhibitor ([44k] = 0), follows a typical Michaelis-Menten saturation curve, which can be ﬁtted to the equation v · [2-oxoglutarate] v = max , (4.1) KM + [2-oxoglutarate] which, in this case, gives a maximum velocity for the enzymatic demethylation vmax of 13.2 nM/min and a KM value of 1.312 µM for binding of 2-OG to the enzyme.

96 4.2. Tetrazolyl Hydrazide-Based Inhibitors

(a) Inhibition of JMJD2A by 44k (b) Lineweaver-Burk representation of data

Figure 4.3.: Inhibition of JMJD2A by the tetrazolyl hydrazide inhibitor 44k is competitive to the co-substrate 2-oxoglutarate. See text for explanations. Shown are mean ± s. d. of duplicates. Modiﬁed under license from Ref. 189.

Table 4.3.: Kinetic data of inhibition experiments with JMJD2A and the tetrazolyl hydrazide inhibitor 44k (cf. Figure 4.3(a)).

[44k]/ µM 0.00 0.50 1.00 2.00 2.50 −1 vmax / nM · min 13.2 12.7 11.6 10.2 13.2 app KM / µM 1.312 1.590 1.707 1.997 3.608

97 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

With increasing concentrations of inhibitor, however, the curve shifts to the right, i. e. to app higher apparent Michaelis-Menten constants KM . Notably, the maximum velocity of the enzyme vmax remains unchanged within a margin of error (cf. Figure 4.3(a) and Table 4.3). This shows that inhibition by 44k is indeed competitive to 2-oxoglutarate (2-OG) and can be overcome by increasing concentrations of the co-substrate. Another representation of the same data set is the linearized double reciprocal plot in Figure 4.3(b), the so-called Lineweaver-Burk representation. This plot shows, as expected, linear behavior of the kinetic data for every inhibitor concentration and, more importantly, that all lines intersect in one point and that this point lies on the 1/v axis, again illustrating the competitive nature of inhibition by test compound 44k. app Moreover, the increase in the apparent Michaelis-Menten constants KM is linear with regard to the inhibitor concentration as depicted in Figure 4.3(c). This is expected for a competitive inhibitor. When these values are ﬁtted to the equation app 0 [inhibitor] KM = KM · 1 + , (4.2) Ki an assay-independent inhibition constant Ki is obtained for 44k, which is as low as 1.97 µM.

4.2.3. Structural Binding Model

Based on the structure-activity relationship observed for inhibitors of the tetrazolyl hydrazide series and the kinetic data from competition experiments with the enzyme co-substrate 2-oxoglutarate (2-OG), a structural binding model can be envisaged, which is schematically depicted in Figure 4.4. As such, the inhibitor 44k would bind the central ferrous ion in the active site of JMJD2A via its hydrazide function as its metal-binding group (MBG) and, thereby, act as a bidentate chelate ligand, which competitively displaces the co-substrate 2-OG. At the same time, the acidic tetrazole moiety is oriented towards two highly conserved amino acids in the active site, lysine K206 and tyrosine Y132. Under physiological conditions, the tetrazole ring can be expected to carry a negative charge, forming an additional ionic interaction with the positively charged ε-ammonium group of the lysine residue. This explains why only derivatives with this tetrazole moiety were potent inhibitors, but not the N -methylated control compound or any other nitrogen heterocycle (cf. Figures 4.1 and

98 4.2. Tetrazolyl Hydrazide-Based Inhibitors

Figure 4.4.: Cartoon representation of the proposed binding mode of tetrazolyl hydrazide inhibitor 44k in the active site of JMJD2A. See text for details. Adapted under license from Ref. 189.

4.2). In this sense, the tetrazole ring can be thought of as a bioisosteric replacement of the terminal carboxylate group of the co-substrate 2-oxoglutarate. Similarly, the metal-binding hydrazide can be considered a mimic of the α-ketocarboxylate encountered in 2-OG, making the inhibitor 44k an uncleavable co-substrate analog.

4.2.4. Selectivity

As the last step in the investigation of the lead compound 44k, its selectivity towards other JumonjiC domain-containing histone demethylases was investigated using the newly established LANCEUltra assays for JARID1A and JMJD3 (cf. Section 3.3.6). The IC50 values obtained in these assays can be used as an approximate measure of the compound’s selectivity and are summarized in Table 4.4. As controls, the previously reported JMJD3-selective inhibitor GSK-J1 23a[158] and the relatively unselective inhibitor pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6[126] were included for comparison. The tetrazolyl hydrazide inhibitor 44k, thus, exhibited a remarkable selectivity for JMJD2A with a four-fold greater IC50 value on JARID1A and a ca. 41-fold selectivity over JMJD3.

99 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.4.: Selectivity data of the most potent tetrazolyl hydrazide-based inhibitor 44k against JMJD2A, JMJD3, and JARID1A in LANCEUltra assays. Data are mean ± s. d. of duplicates. Reference compounds pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6 and GSK-J1 23a are included for comparison.

IC50 / µM Compound JMJD2A JMJD3 JARID1A 44k 2.38 ± 0.37 98.0 ± 7.2 10.4 ± 1.8 2,4-PDCA 6 0.0337 ± 0.0070 36.0 ± 4.7 0.100 ± 0.028 GSK-J1 23a 54.4 ± 1.6 0.128 ± 0.018 0.418 ± 0.153

Given the small size of this fragment-like inhibitor and its simple mode of action, this selectivity is outstanding. Likewise, the broad-spectrum inhibitor 2,4-PDCA 6 is most potent on JMJD2A, with a similar trend regarding JARID1A and the KDM6 subfamily demethylase JMJD3. Notably, the published inhibitor GSK-J1 23a does indeed exhibit its greatest potency on JMJD3, however with only a small selectivity factor around three-fold over JARID1A, which coincides with recent reports questioning the selectivity of GSK-J1 23a.[159, 160]

4.2.5. Elaboration of the Lead Structure

Encouraged by the remarkable potency and outstanding selectivity of the fragment-like screening hit 44k, it was attempted to elaborate this scaﬀold into larger, more potent, and potentially more selective inhibitors. This was attempted by two strategies. Firstly, the hydrazide 44k was transformed into acylhydrazones by reaction with diverse carbonyl components and secondly, by modiﬁcations to the spacer between the two relevant parts of the molecule, the metal-binding group and the tetrazole ring.

4.2.5.1. Acylhydrazone Derivatives

As a ﬁrst attempt towards the elaboration of the tetrazolyl hydrazide lead structure, the hydrazide motif was transformed into acylhydrazones with a variety of alkyl and aryl substituents

100 4.2. Tetrazolyl Hydrazide-Based Inhibitors as summarized in Table 4.5. These were synthesized and characterized by Nicole Rüger(Link group) as outlined in Ref. 189. The novel acylhydrazones 47a–u thus obtained were subjected to pre-tests performed in duplicate against JMJD2A at a limited number of concentrations using the FDH-coupled as well as LANCEUltra assays to determine their inhibitory potency. If compounds exhibited less than 30% inhibition at the highest concentration of 400 µM in the FDH-coupled assay or less than 50% at 100 µM in LANCEUltra, they were rejected as inactives (denoted as ‘no inhibition’ in the data table). If a compound passed either cut-off criterion, it was further evaluated. The data for all acylhydrazones is summarized in Table 4.5. Surprisingly, most compounds in this series were completely devoid of any inhibitory potency and even those compounds, which exhibited some inhibition, were much less potent compared to the original lead structure 44k. This may mean that the substituents introduced by hydrazone formation were generally too large and led to steric clashes in the active site near the iron-binding center. In addition, the introduction of such residues may have impeded the metal-binding capacities of the hydrazone function even though the electron density at the distal nitrogen atom required for a dative bond towards the metal center should rather be increased in these structures. The only test compounds, which showed somewhat acceptable inhibition, were those with very small alkyl substituents like the dimethyl derivative 47f or the diethyl derivative 47g. When this was increased to the rigidified cyclopentyl derivative 47h, a dramatic drop in potency was already observed. This reinforces the notion that the steric demand of the residues in most derivatives was simply too great to accommodate these compounds in the active site. For the nitrophenyl derivatives 47o and 47p and the 2-methoxyphenyl derivative 47u, the inhibitory potency in the FDH-coupled assay was on the limit of detectability and they were, consequently, not tested in the LANCEUltra assay. For the cinnamyl derivative 47s, an unex- pected reversal in potencies was observed. This compound was apparently more potent in the FDH-coupled assay than in the LANCEUltra assay, even though the latter uses dramatically less enzyme. As this compound contains a reactive Michael acceptor substructure, this is likely an artifact caused by interference with assay components. As the acylhydrazone modification to the lead structure led to no notable improvement in potency, this strategy was not pursued any further.

101 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.5.: Potency of tetrazolyl acylhydrazone-based test compounds against JMJD2A. Data are mean ± s. d. of duplicates. n. t. – not tested. n. i. – no inhibition (see text for deﬁnition).

FDH-coupled assay LANCEUltra assay R1 R2 IC50 / µM IC50 / µM

47a CH3 4-pyridyl n. i. n. i.

47b CH3 4-methoxyphenyl n. i. n. i. 47c H phenyl n. i. n. i.

47d CH3 phenyl n. i. n. i. 47e H 4-methoxyphenyl n. i. n. i.

47f CH3 CH3 391 ± 61 11.8 ± 0.4

47g CH2CH3 CH2CH3 154 ± 42 6.19 ± 1.51

47h −(CH2)4− 400 81.6 ± 1.3 47i H 4-(dimethylamino)phenyl n. i. n. i. 47j H 4-(triﬂuoromethyl)phenyl n. i. n. i. 47k H 4-chlorophenyl n. i. n. i. 47l H 4-cyanophenyl n. i. n. i. 47m H 4-bromophenyl n. i. n. i.

47n CH3 4-methylphenyl n. i. n. i. 47o H 4-nitrophenyl 310 n. t.

47p CH3 4-nitrophenyl 285 ± 78 n. t. 47q H 4-methylphenyl n. i. n. i. 47r H 2-methylphenyl n. i. n. i. 47s H cinnamyl 159 ± 7 283 ± 51 47t H 4-ethylphenyl n. i. n. i. 47u H 2-methoxyphenyl 219 ± 21 n. t.

102 4.2. Tetrazolyl Hydrazide-Based Inhibitors

4.2.5.2. Rigidiﬁcation of the Spacer

As the next method to elaborate on the lead structure of 44k, the spacer joining the two parts of the molecule, i. e. the metal-binding hydrazide and the acidic tetrazole ring, was exchanged by more rigid phenyl moieties as depicted in Table 4.6. These compounds, synthesized by Nicole R¨ugerand a student under her supervision (Link group), were again subjected to testing in both JMJD2A in vitro assays.

Notably, for these compounds, again a reversal in potencies was observed, i. e. that the IC50 values in the FDH-coupled assay were, unexpectedly, smaller than in the LANCEUltra assay (cf. Table 4.6). When compounds 48 and 49 were subjected to the FDH-counterscreening assay (cf. Section 3.2), it was revealed that they indeed inhibited oxidation of formaldehyde by formaldehyde dehydrogenase (FDH). The observed apparent potency for these compounds was, thus, only an artifact from interferences with the assay system and only the LANCEUltra assay could be used to evaluate the actual JMJD2A inhibitory potential. Importantly, such a behavior was not observed for the lead compound 44k. As can be seen from the results of the LANCEUltra assays (cf. Table 4.6), the rigidiﬁcation of the spacer led to a marked loss in potency when compared to the original lead structure 44k.

Table 4.6.: IC50 values of rigidiﬁed tetrazolyl hydrazide-based inhibitors against JMJD2A. Data for LANCEUltra are mean ± s. d. of two independent experiments. n. t. – not tested.

FDH-coupled assay LANCEUltra assay Compound Structure IC50 / µM IC50 / µM

48 21.8 216 ± 33

49 28.3 195 ± 39

50 n. t. 206 ± 42

103 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

This may be related to the suboptimal geometry between the two molecule parts enforced by the rigid spacer substituted in a meta or para pattern. Moreover, the distance between the hydrazide and tetrazole moieties is increased compared to 44k and, for the meta derivative coincides with that in 44m, which is also less potent. As such, it may be desirable to test the ortho derivative, which was, however, synthetically not accessible. Consequently, the strategy of rigidifying the spacer between the hydrazide and tetrazole groups was abandoned.

4.2.5.3. Branched Derivatives

Lastly, the lead structure scaffold of 44k was expanded by the synthesis of branched derivatives. The fragment-like inhibitor 44k can be seen as an anchor, which tightly binds in the active site of JumonjiC domain-containing histone demethylases and connects the metal-binding site to the conserved lysine residue stabilizing co-substrate binding (cf. Figure 4.4). If the free methylene bridge in this molecule is substituted, this anchor can be expanded into larger inhibitors exploring the remainder of the active site binding pocket. As such, four branched derivatives 51a–d of 44k carrying different substitutions were synthesized by Nicole Rügerand students under her supervision (Link group) and tested against JMJD2A. Their structures and the results are summarized in Table 4.7. It should be noted that all these compounds are chiral and were provided as racemates, complicating a detailed analysis of their structure-activity relationship. Unexpectedly, all these branched derivatives 51a–d were also dramatically less potent JMJD2A inhibitors than the original lead structure 44k by a factor of at least 20-fold when comparing data from the LANCEUltra assays. This may be the case because unsuitable or excessively large substituents were chosen, which can not be accommodated in the active site or because the substitution is too close to the relevant parts of the molecule binding the central ferrous ion or the conserved lysine residue. This may apply to one or both of the corresponding enantiomers. In order to draw a clearer conclusion on the effect of substitutions on these branched derivatives, the enantiomers would need to be separated and tested individually. The lack of detailed structural knowledge about the binding situation of lead compound 44k in the active site of JMJD2A further impedes a more rational approach towards the design of

104 4.2. Tetrazolyl Hydrazide-Based Inhibitors

Table 4.7.: IC50 values of branched tetrazolyl hydrazide-based inhibitors against JMJD2A. Data shown are mean ± s. d. of two independent experiments.

FDH-coupled assay LANCEUltra assay Compound R = IC50 / µM IC50 / µM

51a 123 ± 9 183 ± 8

51b 120 ± 4 94.9 ± 6.1

51c 96.9 ± 13.7 69.2 ± 3.2

51d 400 42.8 ± 7.6

more elaborate inhibitor structures. This could be overcome by the use of molecular docking techniques or, if possible, by co-crystallization of 44k with JMJD2A and resolution of the x-ray structure.

While the attempts at elaborating on the tetrazolyl hydrazide lead scaﬀold were largely unsuccessful and led to no improvement in potency, the original screening hit 44k remains as a strikingly potent and remarkably selective JMJD2A inhibitor, particularly considering its small size and low molecular complexity, which can potentially be useful as a simple and inexpensive in vitro reference inhibitor. The replacement of carboxylic acids by isosteric tetrazoles is a common method in medicinal chemistry.[190] However, this is the ﬁrst reported use of such a structure in JMJC demethylase inhibitors.

105 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

4.3. Tropolone-Based Inhibitors

4.3.1. Discovery

The structural class of tropolones (2-hydroxytropones) has recently attracted attention by epigenetic drug discovery programs since simple substituted derivatives had been reported to be outstandingly potent inhibitors of zinc-dependent histone deacetylases (HDACs). For example, the β-phenyl substituted 2-tropolone was reported to inhibit HDAC2 with a Ki value of only 60 nM and, remarkably, an exquisite selectivity for only this subtype over all other HDACs.[191] The Jung group was not able to reproduce these results with similarly substituted derivatives like the natural product β-thujaplicin 52 for HDACs. However, the suggested mode of action of these compounds, i. e. by active-site chelation of Zn2+ ions, spurred an interest into investigations, whether they might also be inhibitors of Fe2+-dependent JumonjiC domain- containing histone demethylases (JMJC demethylases). In fact, a number of natural products and synthetic compounds based on the tropolone structure are reported as potent inhibitors of a variety of metal-dependent enzymes.[192] To this end, a series of tropolone-based test compounds was acquired and tested on JMJD2A both in the FDH-coupled and LANCEUltra in vitro assays. The natural product β-thujaplicin 52 is commercially available, while benzotropolones 53a–d were synthesized by Deniz Arican (Br¨uckner group, Institute of Organic Chemistry, Albert-Ludwigs-University Freiburg). Their structures and the results of inhibition tests are summarized in Table 4.8. Indeed, all tropolone-based test compounds 52 and 53a–d inhibited JMJD2A with remarkable potency given the relatively small fragment-like nature of these molecules. Both the β-iso- propyl substituted tropolone β-thujaplicin 52 as well as the more lipophilic benzotropolones exhibit comparable potency around 20 µM in the FDH-coupled and around 5–7 µM in the LANCEUltra assay. With regard to the substitution pattern on the benzene ring, no clear structure-activity relationship (SAR) can be deduced as all compounds are similarly potent. Substitution by a methyl group on the tropolone ring itself led to a slight decrease in potency in the LANCEUltra assay, possibly because of the steric demand of this residue.

106 4.3. Tropolone-Based Inhibitors

Table 4.8.: IC50 values of tropolone-based inhibitors against JMJD2A. Data are mean ± s. d. of at least two independent experiments.

FDH-coupled assay LANCEUltra assay Compound Structure IC50 / µM IC50 / µM

β-thujaplicin 52 14.1 ± 0.3 7.27 ± 0.83

53a 23.3 ± 3.5 6.39 ± 1.17

53b 30.1 ± 0.1 7.58 ± 0.83

53c 14.7 ± 1.9 5.33 ± 0.42

53d 26.2 ± 1.8 14.34 ± 3.77

53e 136.6 ± 7.3 26.60 ± 1.75

4.3.2. Competitivity of Inhibition Towards 2-Oxoglutarate

The mode of action of these compounds is very likely by active site complexation of the ferrous ion and, thus, competitive displacement of the enzyme co-substrate 2-oxoglutarate (2-OG). The hydroxyl groups of tropolones are reported to be acidic,[192] while the resulting corresponding anion is stabilized by a mesomeric structure with the adjacent carbonyl group, dividing the negative charge onto the two oxygen atoms. This makes tropolones ideal bidentate metal chelators.

107 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Figure 4.5.: Docking pose of compounds 53c (dark green) and 53e (light green) in the active site of JMJD2A. The Fe2+ ion is shown as a cyan sphere, oxygen atoms in red, and nitrogen atoms in blue. Interaction distances in A˚ with relevant amino acids Lys206 and Tyr132 are highlighted.

Such a binding mode is also suggested by a docking pose of compound 53c in the active site of JMJD2A as depicted in Figure 4.5. This image was generated by the Sippl group (Institute of Pharmacy, Martin-Luther-University Halle-Wittenberg). It shows benzotropolone 53c, here depicted in dark green, complexing the central ferrous ion via its carbonyl and hydroxyl oxygen atoms in a bidentate manner. The planar benzotropolone scaffold fits nicely into the active site shaped by the surrounding amino acid residues (cf. surface representation in Fig. 4.5). Both methoxy groups of 53c can be accommodated in the active site, but form no specific interactions. Based on this docking pose, a new compound was devised in silico bearing an additional carboxylic acid on the tropolone ring, which would be able to engage in an additional ionic interaction with the highly conserved active site residue lysine K206 and a hydrogen bond with tyrosine Y132. In order to allow for these interactions, the planar benzotropolone scaf-

108 4.3. Tropolone-Based Inhibitors fold would be shifted slightly out of its preferred binding geometry (cf. Fig. 4.5, optimized carboxylate derivative in light green). This compound 53e was, subsequently, synthesized by Deniz Arican (Brückner group) and tested in both in vitro assays. Surprisingly, the novel carboxylate derivative 53e was approximately 5-fold less potent on JMJD2A when compared to its closest analog 53d in the FDH-coupled assay or two-fold in the LANCEUltra assay (cf. Table 4.8). The reason may be that the steric demand of the additional residue is too great and forces the planar ring system out of its favorable Fe2+-binding position as already observed for 53d. Furthermore, the methoxy groups of 53e protrude into a water-filled subpocket of JMJD2A as shown in Fig 4.5 (water molecule HOH630). Depending on how strongly this water network is bound, perturbations may lead to an enthalpic penalty, making binding in this position unfavorable. However, this cannot be predicted using the docking procedure. In order to investigate the iron-binding capacities of these test compounds and to determine whether there are intrinsic differences between 53e and its related analogs 53a–d, the Ferrozine-displacement assay was employed (cf. Section 3.5). However, these compounds could not be tested in this assay as they form brown complexes with iron themselves. UV/vis- spectral analyses revealed that absorption of these complexes overlapped with that of the II 2+ [Fe (Ferrozine)3] complex. Nonetheless, the competitivity of inhibition with regard to 2-oxoglutarate (2-OG) could be demonstrated by kinetic experiments in the LANCEUltra assay set-up as discussed previously (cf. Section 4.2.2). The results for exemplary compounds 53b and β-thujaplicin 52 are depicted in Figure 4.6. These results clearly indicate that by increasing the concentrations of the enzyme co-substrate 2-OG, the inhibition by test compounds can be overcome, meaning that they are mutually competitive. In the kinetic experiments (cf. Figures 4.6(a) and 4.6(c)), the curves all reach the same maximum velocity, independent of inhibitor concentration. How- ever, they are gradually shifted to the right, meaning to higher apparent Michaelis-Menten constants. The linearized double reciprocal representations of the same data sets (Lineweaver-Burk representations, cf. Figures 4.6(b) and 4.6(d)) clearly indicate that the data follow the linear trend required for competitive inhibition and, moreover, all lines for the different inhibitor concentrations intersect the 1/v-axis in the same point. Taking these observations together

109 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

(a) Inhibition of JMJD2A by 53b (b) Lineweaver-Burk representation of data

Figure 4.6.: Inhibition of JMJD2A by tropolone-based inhibitors 53b and β-thujaplicin 52 is competitive to co-substrate 2-oxoglutarate. See text for explanations. Shown are mean ± s. d. of duplicates.

proves that these compounds inhibit JMJD2A in a 2-OG-competitive manner, which supports the predicted docking pose for these compounds as active site metal chelators and, therefore, co-substrate competitors (cf. Figure 4.5).

From the analysis of the apparent KM values in these experiments, an inhibition constant Ki of 8.40 µM for 53b and 2.52 µM for β-thujaplicin 52 could be obtained, respectively.

110 4.3. Tropolone-Based Inhibitors

Table 4.9.: Selectivity data of tropolone-based inhibitors against JMJD2A, JMJD3, and JARID1A in LANCEUltra assays. Data are mean ± s. d. of duplicates.

IC50 / µM Compound JMJD2A JMJD3 JARID1A β-thujaplicin 52 7.27 ± 0.83 4.77 ± 0.003 5.86 ± 1.37 53b 7.58 ± 0.83 6.89 ± 0.96 7.38 ± 0.42

4.3.3. Selectivity

As the next step, the selectivity of tropolone-based inhibitors among a panel of JumonjiC domain-containing histone demethylases (JMJC demethylases) was evaluated using the established LANCEUltra assays against JMJD2A, JMJD3, and JARID1A (cf. Section 3.3). The results for exemplary compounds 53b and the natural product β-thujaplicin 52 are summarized in Table 4.9. As becomes evident from the in vitro inhibitory data, these simple (benzo)tropolone-based inhibitors are not selective among the tested JMJC demethylases. This was to be expected as these compounds function as relatively general active site metal chelators. Given their small molecular size and complexity, they can rather be considered fragment-like and likely not selective. However, optimized derivatives can be envisaged that elaborate this fragment scaﬀold into larger structures containing selectivity-inducing functionalities. This may be accomplished based on the predicted binding pose (cf. Figure 4.5).

4.3.4. Cellular Activity

Lastly, the biological eﬀects of tropolone-based inhibitors were evaluated in cell culture experiments using the esophageal cancer cell line KYSE-150.[193] The results of cell proliferation studies for exemplary compounds 53b and β-thujaplicin 52 carried out by Karin Schmidtkunz (Jung group) are summarized in Table 4.10. Remarkably, these compounds, despite their low molecular complexity and only mild in vitro potency already showed notable inhibition of cell proliferation in the lower micromolar range. As these compounds are likely unselective pan-JMJC demethylase inhibitors (cf. Table 4.9),

111 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.10.: Cellular potency of tropolone-based inhibitors on KYSE-150 cells.[193] Data are mean ± s. d. of at least triplicates.

KYSE-150 cells Compound Structure GI50 / µM

β-thujaplicin 52 23.3 ± 3.3

53b 6.6 ± 2.3

it is not possible to distinguish between the different biological effects of inhibition of one of the many JMJC demethylases and to determine, to which extent they contributed to the cytotoxicity. Moreover, similar structures were reported as potent HDAC inhibitors,[191] which, together with other putative off-targets of such a small fragment-like inhibitor, may also contribute to the cellular effects. It should be noted that the benzotropolone derivative 53b was ∼4-fold more potent on cells than the natural product 52, which may have to do with its reduced polarity due to the more lipophilic side chains, which facilitates cell permeation.

In summary, the (benzo)tropolone scaﬀold represents a novel metal-binding group (MBG) as demonstrated in these fragment-like inhibitors with remarkable cellular potency, which can be used as an anchor to elaborate it into larger structures with improved potency and selectivity.

112 4.4. Hydroxamic Acid-Based Inhibitors

4.4. Hydroxamic Acid-Based Inhibitors

4.4.1. Discovery

This section describes the development of novel inhibitors of JumonjiC domain-containing histone demethylases based on the hydroxamic acid CO−NH−OH warhead. The lead structure for this project was discovered by screening a focused library of hydroxamates as known histone deacetylase (HDAC) inhibitors. Parts of this project have been summarized in a manuscript recently submitted for publication.[194] Hydroxamic acid-based structures have previously been reported as inhibitors of iron-dependent JMJC demethylases, in part even exhibiting good subfamily selectivity as outlined in the Intro- duction (cf. Section 1.4.4).[126, 154–156] However, this warhead is also a very common moiety in inhibitors of zinc-dependent HDACs. Several such structures have already reached the clinic and have received FDA approval. As there was a great expertise in the development of hydroxamic acid-based HDAC inhibitors in the Jung group, it seemed reasonable to screen a small focused library of these molecules for their oﬀ-target inhibition on JMJC demethylases. Given their known metal-binding properties, it was assumed that they could also chelate the central ferrous ion present in these enzymes and competitively displace the co-substrate 2-oxoglutarate (2-OG). If a potent inhibitor could be found, this could serve as a lead structure initiating a medicinal chemistry-driven project to alter its selectivity to become less potent of an inhibitor on HDACs and more potent on JMJC demethylases. To this end, a small library of known HDAC inhibitors was screened for their JMJD2A inhibition using the FDH-coupled assay. Their chemical structures and inhibition data are summarized in Figure 4.7. Importantly, many of these compounds did not exhibit any noteworthy inhibition at concentrations of up to 200 µM despite the metal-binding warhead. The fact that some but not all hydroxamic acids inhibited JMJD2A indicates that inhibition by these compounds is not an artifact caused by metal chelation in solution, but that there is a speciﬁc interaction between some inhibitors and the enzyme. While 54d exhibited very mild inhibition of JMJD2A (IC50 = 111 µM), none of the other substituted thiazoles seemed promising so that this structure was abandoned.

113 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Figure 4.7.: Focused library screening of hydroxamic acid-based HDAC inhibitors

against JMJD2A. IC50 values are given for inhibition in the FDH-coupled assay. If no value is given, inhibition was less than 50% at 200 µM.

The clinically used HDAC inhibitor SAHA 17 was found to be a very weak inhibitor of JMJD2A

(IC50 = 182 µM), which coincides with literature reports using the closely related enzyme JMJD2E.[126] Related structures 55–58 could all be shown to inhibit JMJD2A, which led to the conclusion that the extended alkyl chain linker between the metal-binding group (MBG) and a lipophilic cap group was required for potent inhibition of JMJC demethylases. Of particular note is compound 57a, which in addition to the metal-binding hydroxamate and a long alkyl linker possesses a large lipophilic biphenyl residue. This is synthetically derived

114 4.4. Hydroxamic Acid-Based Inhibitors from the unnatural amino acid biphenylalanine, which explains its stereochemistry. This very lipophilic side chain made compound 57a the most potent inhibitor in this series with an IC50 value of only 25.4 µM so that 57a can be considered the hit compound from this focused library screening. It should be noted, however, that this compound is a much more potent [195] inhibitor of HDACs with IC50 values in the nanomolar range. Moreover, it is noteworthy that alkylation at the nitrogen atom of the hydroxamic acid as in compound 58 was well tolerated or even led to an increased potency compared to its closest analog SAHA 17. This is interesting as such a modiﬁcation is generally considered to be detrimental to HDAC inhibition. This could, thus, be a potential starting point for the development of JMJC demethylase-selective inhibitors. In order to validate that the hydroxamic acid in 57a was indeed a required structural motif for potent JMJD2A inhibition, several control compounds were tested, in which this structural feature was masked as outlined in Figure 4.8. As expected, neither the benzyl-protected amide 57b nor the methyl ester of the hydroxamic or carboxylic acid 57c or 57d, respectively, exhibited any inhibition of JMJD2A in the FDH-coupled assay. This conﬁrms that the hydroxamate moiety in 57a is essential to its inhibitory potency.

Figure 4.8.: Control compounds to validate the relevance of the hydroxamic acid in 57a as an inhibitor of JMJD2A in the FDH-coupled assay. If no value is given, inhibition was less than 50% at 200 µM.

115 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

4.4.2. Elaboration of the Lead Structure – SAR

Based on the data from the focused library screening, which led to the discovery of 57a as a potent JMJD2A inhibitor, a first structural model could be envisaged for relevant parts of this structure. This means that such compounds contain three important features: the hydroxamic acid as a metal-binding group, an alkyl spacer, and a lipophilic cap group, which can possibly undergo hydrophobic interactions elsewhere in the active site. In 57a, this hydrophobic residue is the biphenyl group. This is the same three-component set-up commonly encountered in HDAC inhibitor design.[42, 195] In a modular approach, modifications were made to all three parts of the structure in order to establish a structure-activity relationship (SAR) and to, firstly, increase potency on JMJC demethylases and, secondly, also selectivity over other targets.

4.4.2.1. Modiﬁcations to the Cap Group

In a first attempt at elucidating the SAR of hydroxamic acid-based inhibitors of JMJD2A based on 57a, a panel of structural derivatives was tested with modifications to the lipophilic cap group. Their structures and inhibitory potencies against JMJD2A as determined in the FDH- coupled assay are summarized in Table 4.11. While most of these compounds have previously been published as HDAC inhibitors,[195, 196] compounds 57e, 57g, and 57n were synthesized for this study by Dr. Ludovica Morera (Jung group). Upon analysis of this in vitro inhibitory data, it becomes clear that the type of ester modification to the structure (residue R3) did not change the potency of these compounds dramatically (compare 57a to 57e and 57f to 57g). The larger and more lipophilic tert-butyl ester only slightly increased potency compared to its methyl analogs. However, a dramatic change in potency could be observed for different substitutions on the phenylalanine ring (residue R2). Importantly, shortened derivatives with only one phenyl ring like 57f and 57g were almost devoid of any inhibitory potency. The same is true, albeit to a lesser extent, for the nitrophenyl derivative 57h, which was still 5-fold less potent than the biphenyl analog 57a. This very steep drop in potency reinforces the notion that large residues based on the biphenylalanine scaffold are vital to the observed in vitro potency of 57a. Analogs, in which such a lipophilic biphenylalanine-type scaffold is present with substitutions

(57i–m), were all equally potent with IC50 values around 14–30 µM like screening hit 57a.

116 4.4. Hydroxamic Acid-Based Inhibitors

Table 4.11.: Potency of substituted phenylalanine-based hydroxamic acids as inhibitors of JMJD2A. Data are mean ± s. d. of duplicates.

FDH-coupled assay R2 = R3 = IC50 / µM

57a phenyl OCH3 25.4 ± 2.9

57e phenyl OC(CH3)3 14.3 ± 2.5

57f H OCH3 312 ± 11

57g H OC(CH3)3 201 ± 49

57h NO2 OCH3 122 ± 11

57i 4-chlorophenyl OCH3 18.3 ± 3.3

57j 3,4-dichlorophenyl OCH3 25.0 ± 2.2

57k 4-methylphenyl OCH3 22.9 ± 1.0

57l 4-methoxyphenyl OCH3 14.0 ± 1.8

57m 2-naphthyl OCH3 29.7 57n phenyl NHOH 8.55 ± 0.79

Curiously, replacement of the terminal ester group by another hydroxamic acid group as in 57n led to a slight increase in observed in vitro potency. However, based on the structural binding model envisaged for these compounds, this moiety cannot be involved in metal ion binding in the active site. In a second series of analogs of 57a with modifications to the lipophilic cap group, such derivatives were derived not from phenylalanine, but from tyrosine. Such analogs with 3-aryl substitutions could be synthesized employing a radical arylation strategy using aryl diazonium salts. The use of 3-substituted tyrosines as building block allowed for deviations from the otherwise linear substitution pattern found in 57 towards branched derivatives that could potentially better fill the active site. Compounds 59a–h were synthesized by Michael Fürst

117 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.12.: Potency of 3-substituted tyrosine-based hydroxamic acids as inhibitors of JMJD2A. Data are mean ± s. d. of duplicates.

FDH-coupled assay LANCEUltra assay R = IC50 / µM IC50 / µM 59a phenyl 93.9 ± 2.0 17.1 ± 3.3 59b 4-chlorophenyl 48.3 ± 2.0 15.3 ± 2.4 59c 4-ﬂuorophenyl 52.2 ± 3.3 10.8 ± 0.7 59d 4-cyanophenyl 54.8 ± 7.0 8.05 ± 0.20 59e 3-chlorophenyl 27.6 ± 2.9 6.35 ± 0.12 59f 3-ﬂuorophenyl 29.0 ± 0.4 10.8 ± 0.0 59g 3-cyanophenyl 47.6 ± 3.1 12.0 ± 0.5 59h 2-chlorophenyl 56.3 ± 15.5 12.9 ± 1.5

(Heinrich group, Department of Chemistry and Pharmacy, Friedrich-Alexander-University Erlangen-Nuremberg). The in vitro potencies against JMJD2A for compounds of this type are summarized in Table 4.12. As becomes evident from the in vitro data in Table 4.12, the change from a linear biphenylalanine substitution pattern to 3-substituted tyrosines generally led to a loss in potency by almost 4-fold comparing 59a to 57a. However, this could be used to explore the substitution pattern around the meta-phenyl ring. In the absence of structural data of how such compounds bind in the active site of JMJD2A (e. g. crystallography, molecular docking), the SAR can only be based on in vitro data and the explanations of such eﬀects remain speculative. Substitutions with lipophilic residues on the meta-phenyl ring led to an increase in potency (compare 59b–h to 59a). This was particularly true for substitutions in the meta position of this ring with halogen atoms like ﬂuorine or chlorine. These compounds 59e and 59f fall into the same category of potency as the biphenylalanine-based 57a. Possibly, the extended

118 4.4. Hydroxamic Acid-Based Inhibitors ring system can adopt a conformation around the single bonds, where the halo-substituted phenyl ring can be brought into proximity of the pocket, into which the distal phenyl ring of the 4-biphenylalanine derivatives would otherwise protrude. As none of the compounds of the 3-phenyltyrosine series surpassed the 4-biphenylalanine-based inhibitors in in vitro potency, they were not pursued further. However, based on the best compound of this series 59e, further studies could be undertaken into the relevance of the ester moiety and the required distance of the lipophilic cap group from the amide linker. Such compounds were also synthesized by Michael F¨urst(Heinrich group) and the results for inhibition tests against JMJD2A are summarized in Table 4.13. As becomes clear from these data, the ester group found in compounds like the biphenylalanine series 57 and the 3-phenyltyrosine series 59 is dispensable to the potency as a compound based on the decarboxylation product tyramine 60a exhibited virtually equal potency compared to the similarly substituted analog 59e. This is in line with the observations for diﬀerent esters in lead series 57 (cf. Table 4.11). Protection of the phenolic OH groups to more lipophilic methoxy residues even increased the potency further like in compound 60b. Importantly, the distance between the amide coupling site and the lipophilic cap group is not critical to the potency as compounds

60b and 60c exhibited nearly equal IC50 values. This information is valuable since the ester side chain in these compounds was not introduced rationally, but merely for synthetic reasons as amino acids were readily available starting ma-

Table 4.13.: Potency of 3-substituted tyrosine-based hydroxamic acids lacking the ester group as inhibitors of JMJD2A. Data are mean ± s. d. of duplicates.

FDH-coupled assay LANCEUltra assay n = R = IC50 / µM IC50 / µM 60a 2 OH 18.8 ± 0.5 9.13 ± 0.77

60b 2 OCH3 16.6 ± 1.3 4.88 ± 0.79

60c 1 OCH3 14.5 ± 4.1 4.44 ± 0.03

119 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases terials. The ﬁnding that it can be omitted entirely and the establishment of a biaryl coupling protocol suitable for these substrates will greatly facilitate synthesis of further analogs of this lead structure.

4.4.2.2. Modiﬁcations to the Spacer

As the next step in the elucidation of a structure-activity relationship for hydroxamic acid- based inhibitors, the spacer between the metal-binding hydroxamate and the lipophilic cap group was modiﬁed based on the unsubstituted biphenylalanine hit compound 57a. The results of inhibition tests against JMJD2A are summarized in Table 4.14. Compounds 61 and 62 were synthesized by Dr. Ludovica Morera (Jung group). Rigidiﬁcation of the alkyl chain spacer in 57a to an ethylphenyl spacer as in compound 61 was well tolerated and did not lead to any notable change in potency compared to 57a. However, shortening of the spacer by one methylene unit like in compound 62 dramatically reduced potency by nearly 3-fold. This suggests that the distance between the central ferrous ion in the active site of JMJD2A and the pocket, in which the hydrophobic interactions with the cap group occur, is too great to be bridged by this shortened compound.

Table 4.14.: Potency of hydroxamic acid-based inhibitors of JMJD2A with modiﬁca- tions to the spacer. Data are mean ± s. d. of duplicates.

FDH-coupled assay Structure IC50 / µM

61 25.8 ± 3.0

62 70.6 ± 12.2

120 4.4. Hydroxamic Acid-Based Inhibitors

While it would have been interesting to initiate a full investigation into diﬀerent chain lengths with shortened as well as elongated spacers, the (CH2)6 spacer in 57a was retained and attempts at improving the potency of these compounds were made through modiﬁcations to the hydroxamate warhead.

4.4.2.3. Modiﬁcations to the Hydroxamate Warhead – Propanoates

Lastly, modifications to the hydroxamate warhead in 57a itself were introduced. The data from the focused library screening (cf. Figure 4.7) already suggested that alkylation at the hydroxamate nitrogen atom was well tolerated or even increased potency. This was used to introduce an additional carboxylate side chain to these molecules as shown in Table 4.15. The rationale was that the hydroxamate should anchor these molecules at the central ferrous ion in the active site of JMJD2A. Then, the additional carboxylate should be positioned directly opposite the conserved active site residue lysine K206, which normally undergoes an ionic interaction with the terminal carboxylate group of the co-substrate 2-OG. Mimicking this interaction should further increase potency of 57a-based inhibitors. This concept was independently also developed by Suzuki et al. and published in 2013.[155] Moreover, alkylation at the hydroxamate nitrogen atom is generally not well tolerated for inhibition of Zn2+-dependent histone deacetylases (HDACs). This modification may, therefore, also be used to induce selectivity between these different enzyme families (cf. Section 4.4.4). Derivatives of 57a bearing a propanoate side chain were synthesized by Dr. Ludovica Morera (Jung group) and then tested for their JMJD2A inhibition in both the FDH-coupled as well as LANCEUltra assays. Their structures and inhibition data are summarized in Table 4.15. Surprisingly, propanoate-bearing derivatives 63c and 63d were not significantly more potent than the lead structure 57a in spite of the possible interaction of this side chain with active site residue K206. In the absence of structural data, it is difficult to rationalize this. However, one reason may be that binding of the central ferrous ion in JMJD2A required a particular conformation, which made it unfavorable for the propanoate side chain to extend in a correctly oriented manner towards this basic active site amino acid. Curiously, for most examples, it seems as if the methyl esters were more potent than their corresponding carboxylic acids (compare 63b to 63a and 63d to 63c). This suggests that no

121 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.15.: Potency of biphenylalanine-based hydroxamic acids with propanoate side chains as inhibitors of JMJD2A. Data are mean ± s. d. of duplicates.

FDH-coupled assay LANCEUltra assay n = R = IC50 / µM IC50 / µM 63a 4 OH 60.6 ± 13.6 26.4 ± 1.4

63b 4 OCH3 56.0 ± 7.1 11.0 ± 0.3 63c 5 OH 37.1 ± 4.6 8.75 ± 2.38

63d 5 OCH3 17.1 ± 2.3 12.1 ± 0.5 57a 25.4 ± 2.9 12.4 ± 0.8

ionic interaction was formed between the carboxylate and lysine K206 as this would certainly have changed potency dramatically once the negative charge was masked. Based on these data, this interaction rather seems to be only a dipolar interaction or hydrogen bond from the ammonium group of lysine K206 towards this side chain, which is possible both for the acid as well as the ester. On the other hand, these data confirmed the trend already observed for modifications to the alkyl chain linker (cf. Table 4.14) that shortening the distance between metal-binding hydroxamate and lipophilic biphenylalanine cap group dramatically reduced potency. While the addition of a 2-OG-mimicking side chain to these inhibitors led to no notable improvement in potency, it did, however, significantly increase selectivity of these compounds over zinc-dependent HDACs (cf. Section 4.4.4).

4.4.2.4. Modiﬁcations to the Hydroxamate Warhead – Phosphonates

The concept of alkylating the hydroxamic acid with an alkyl spacer and terminal negative charge as in the propanoate series was further tested with phosphonate-bearing side chains. In a screening of small molecule analogs of 2-oxoglutarate (cf. Section 4.1), it was found that the

122 4.4. Hydroxamic Acid-Based Inhibitors anti-malaria drug FR-900098 43a was a surprisingly potent inhibitor of JMJD2A in vitro. It likely chelates the central ferrous ion with its hydroxamic acid function and could engage in an ionic interaction with conserved lysine residue K206 via the negatively charged phosphonate. It should be noted, however, that in this series the distance between these two structural motifs is increased by one CH2 unit compared to the propanoate series. In order to improve on the potency of 43a, derivatives were tested with extended lipophilic side chains attached to the hydroxamate as in the previously discussed inhibitors of lead structure series 57 and 59. These were synthesized by Georg Rennar (Schlitzer group, Institute of Phar- maceutical Chemistry, Philipps-University Marburg). Their structures and in vitro potencies are summarized in Table 4.16. Compared to the original screening hit FR-900098 43a, the pentyl derivative 43b was nearly equally potent. Extending this alkyl chain even further to 43c signiﬁcantly increased potency 4-fold. However, this may also be an artifact of the assay as this compound is structurally very similar to a detergent or other surface-active substances, which may interfere with the assay. A structural modiﬁcation resulting in a hydroxyurea-type structure like 43d was detrimental to inhibition, which may be related to the altered metal-binding properties of this compound. The unsubstituted molecule hydroxyurea 41 was also devoid of any potency (cf. Table 4.1). Substitution of only a phenyl ring to the hydroxamic acid as in 43e was not successful in reproducing the remarkable potency of the alkyl-substituted derivatives. However, sterically more demanding naphthyl derivatives 43f and 43g as well as 4-biphenyl derivative 43i were again remarkably potent compounds with IC50 values around 4–5 µM. Likely, the larger lipophilic side chains could engage in hydrophobic interactions in a pocket farther away from the metal center, which could not be reached by the shorter phenyl ring. When the distance between the metal-binding hydroxamate and the phenyl ring was increased by introduction of a C1 or C2 spacer like in 43j and 43k, potency was subsequently increased with increasing distance compared to 43e, reinforcing the notion that the site of these hydrophobic interactions is farther away from the metal center. However, steric limitations to the size and orientation of larger lipophilic residues could be observed as the 1-naphthylmethyl derivative 43l again exhibited decreased potency, while the 2-naphthylmethyl analog 43m was nearly twice as potent. This could be due to the better orientation of the rigid ring system towards a hydrophobic pocket.

123 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.16.: Potency of hydroxamic acids with phosphonate side chains as inhibitors of JMJD2A. Data are mean ± s. d. of duplicates.

LANCEUltra assay R = IC50 / µM

FR-900098 43a CH3 14.4 ± 0.3 43b 18.5 ± 0.3

43c 4.46 ± 0.40

43d > 250

43e 89.4 ± 17.9

43f 4.36 ± 0.59

43g 4.67 ± 0.46

43h 8.91 ± 0.99

43i 5.17 ± 1.10

43j 21.7 ± 3.1

43k 12.4 ± 0.3

43l 29.6 ± 1.8

43m 16.3 ± 0.2

124 4.4. Hydroxamic Acid-Based Inhibitors

In summary, introduction of phosphonate side chains into hydroxamic acid-based inhibitors has allowed for the development of remarkably potent in vitro inhibitors of JMJD2A. Intriguingly, no phosphonate-based inhibitors of JMJD2A are known in the literature to date. The reason may be their suspected decreased cell permeability due to the negative charge. However, this could be overcome by using acyloxyalkyl ester prodrugs as was demonstrated in analogs of FR-900098 43a for anti-malarial therapy.[197]

4.4.2.5. Other Hydroxamic Acid-Based Inhibitors

Encouraged by the successful development of hydroxamic acid-based JMJD2A inhibitors, in which the hydroxamate group is linked to a lipophilic aryl moiety by an alkyl chain spacer, another compound attracted attention, which was reported as a potent HDAC inhibitor, namely crebinostat 64. It was reported in 2013 as an inhibitor of class I and class IIb histone deacetylases in mouse neurons and to aﬀect gene transcription in these cells to beneﬁcially alter the long term memory formation in mice.[198] Based on its structural similarity to the discovered JMJC demethylase inhibitor 57a, it was also tested for its JMJD2A inhibitory potency. Its structure and in vitro data are depicted in Table 4.17. Crebinostat 64 was synthesized by Dr. Diana Stolfa (Jung group) according to the literature procedure.[198] Remarkably, crebinostat 64 was also found to be a quite potent inhibitor of JMJD2A with an

IC50 value in a range comparable to that of the previously discovered biphenylalanine derivative 57a. It should be noted that the alkyl linker between the hydroxamate and the amide respectively acylhydrazone linkage in 64 is shortened by one CH2 unit compared to 57a, which should have led to a loss in potency based on the previously established structure-activity relationship. However, the additional nitrogen atom in the acylhydrazone moiety increases the overall distance to the metal-binding group and places the biphenyl system in the same position as it was in the lead structure 57a. The only exception is a different bond geometry due to the sp2 hybridization of the hydrazone nitrogen atom. While compound 64 is planar around this part of the molecule, 57a is more flexible. However, the inherent flexibility of the alkyl linker apparently allowed both compounds to adopt a suitable conformation required for nearly equal potency.

125 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.17.: Potency of crebinostat-based hydroxamic acids as inhibitors of JMJD2A. Data are mean ± s. d. of duplicates. n. t. – not tested.

FDH-coupled assay LANCEUltra assay Structure IC50 / µM IC50 / µM

64 23.3 ± 0.8 8.15 ± 1.18

65 64.2 ± 2.3 n. t.

66 200 n. t.

67a 38.5 ± 4.2 16.1 ± 2.1

67b 44.9 ± 2.0 8.85 ± 1.29

Based on this surprising ﬁnding and the ease of synthesis of such acylhydrazone-based compounds, analogs of crebinostat 64 were synthesized by Eva-Maria Herrlinger (65 and 66)[199] and Nicole Steck (67a and 67b) under supervision of Dr. Ludovica Morera (Jung group). The structures and in vitro data of these analogs are also given in Table 4.17. Interestingly, small modiﬁcations to the 4-biphenylalanine backbone in crebinostat 64 like changing the biaryl system to a meta substitution as in 65 already decreased potency of these compounds by a factor of nearly 3. This is analogous to the observation made for the change from the 4-biphenyl compound 57a to 3-phenyltyrosine derivative 59a (cf. Table 4.12). Branched derivative 66 was completely devoid of any inhibitory potency. This can either be the result of a steric clash by this large substituent or it can be seen in analogy to derivatives of 57a with only one phenyl group such as 57f, which was also dramatically less potent.

126 4.4. Hydroxamic Acid-Based Inhibitors

The concept of propanoate side chains as 2-OG-mimics in hydroxamate-based inhibitors was also extended to this lead structure yielding ester 67a and its free carboxylate 67b. The same trend was observed as for propanoate derivatives of 57a that incorporation of this side chain did not signiﬁcantly improve in vitro potency. Importantly, though, the free carboxylate 67b was found to be nearly twice as potent as the methyl ester based on data from the LANCEUltra assay, indicating that the presumed interaction of this side chain with active site residue lysine K206 may be taking place in this case. While the reported HDAC inhibitor crebinostat 64 was not more potent than the original screening hit 57a, it is important to note that some of the biological eﬀects observed for this compound and attributed to HDAC inhibition may also be related to JMJC demethylase inhibition. Moreover, the synthetic simplicity of hydrazone formation may speed up the development of more hydroxamate-based JMJD2A inhibitors for future studies.

4.4.3. Selectivity among JMJC Demethylases

Using the newly established LANCEUltra assays for other subtypes of JMJC demethylases like JARID1A and JMJD3 (cf. Section 3.3.6), a representative set of hydroxamic acid-based JMJD2A inhibitors was also tested against these subtypes in order to evaluate their selectivity. The data from these in vitro inhibition experiments are summarized in Table 4.18. Based on these data, the original screening hit 57a was unselective and inhibited all three demethylases with nearly equal potency. It can likely be considered a pan-JMJC demethylase inhibitor. However, incorporation of the propanoate side chain into this lead structure yielding compounds 63a–c resulted in an increase in selectivity for JARID1A (KDM5A). Importantly, this is only true for the free carboxylate derivatives 63a and 63c, where the selectivity over JMJD2A (KDM4A) reached up to 10-fold. This indicates that the presumed interaction with an active site lysine residue was indeed possible for this demethylase subtype likely due to the slightly diﬀerent size and geometry of the active site. Importantly, methyl ester 63b, which cannot engage in this type of ionic interaction, was found to be unselective. Moreover, the structure-activity relationship regarding the length of the alkyl spacer established for JMJD2A was also validated for the other subtypes JARID1A and JMJD3. This means that derivatives with a shortened spacer like 63a and 63b were less potent than the

127 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.18.: Selectivity of hydroxamic acid-based inhibitors against JMJD2A, JMJD3, and JARID1A in LANCEUltra assays. Data are mean ± s. d. of duplicates. n. t. – not tested.

IC50 / µM Compound JMJD2A JMJD3 JARID1A 57a 12.4 ± 0.8 11.6 ± 0.2 14.5 ± 7.7 63a 26.4 ± 1.4 17.3 ± 1.8 2.57 ± 0.52 63b 11.0 ± 0.3 n. t. 9.02 ± 0.83 63c 8.75 ± 2.38 n. t. 1.65 ± 0.13 64 8.15 ± 1.18 10.8 ± 1.0 14.6 ± 4.2

screening hit 57a or compound 63c. Importantly, the most potent of the JARID1A inhibitors was indeed 63c with the longer (CH2)6 spacer as expected. Crebinostat 64 was also found to be unselective and to inhibit all three tested JMJC demethylases with equal potency. This is expected since this structure does not contain the additional 2-OG-mimicking side chain required for improved potency on JARID1A.

4.4.4. Selectivity over HDACs

The hydroxamic acid-based inhibitors of JMJD2A and other JMJC demethylase subtypes discussed in this section were discovered in a focused library screening of compounds known to be inhibitors of zinc-dependent histone deacetylases (HDACs). Therefore, a relevant question was whether these compounds were selective for JMJC demethylases or also inhibited HDACs, which would confuse the analysis of biological eﬀects. All compounds developed in this program were, hence, also tested for their inhibitory potency on representative human enzymes HDAC1 and HDAC6 in a well-established in vitro screening assay. These counterscreens were performed by Johanna Senger (Jung group). Results for the most potent compounds are taken from Ref. 194 and summarized in Table 4.19. Not surprisingly, the original lead compound 57a is a much more potent inhibitor of histone deacetylases with IC50 values in the nanomolar range, which coincides with literature reports

128 4.4. Hydroxamic Acid-Based Inhibitors

Table 4.19.: Selectivity of hydroxamic acid-based inhibitors over histone deacetylases.

For HDAC1 and HDAC6, their IC50 value or % inhibition at 10 µM is given. JMJD2A data are mean ± s. d. of duplicates. n. i. – no inhibition.

JMJD2A Compound LANCEUltra assay HDAC1 HDAC6 IC50 / µM 57a 12.4 ± 0.8 52.7 nM 107 nM 62 16.7 ± 10.8 86.2 nM 58.1 nM 63a 26.4 ± 1.4 n. i. n. i. 63b 11.0 ± 0.3 n. i. 10% @ 10 µM 63c 8.75 ± 2.38 9% @ 10 µM 11% @ 10 µM 63d 12.1 ± 0.5 16% @ 10 µM 8% @ 10 µM

for this compound tested on a mixture of rat HDACs.[195] The derivative with a shortened alkyl spacer 62 interestingly reversed the selectivity between the two tested HDACs, with a drop in potency on HDAC1, but an increase on HDAC6. More importantly, though, all N -alkylated derivatives with propanoate side chains 63a–d were virtually devoid of any inhibitory potency on HDACs. This is valuable as these compounds are, hence, selective for JMJC demethylases, either as pan-JMJC demethylase inhibitors like 63b or even with a certain subfamily selectivity for JARID1A like 63a and 63c. Although incorporation of this residue has failed to increase potency of these compounds on JMJD2A, it established a strategy for the development of JMJC demethylase inhibitors with no eﬀect on HDACs. Starting from the very potent HDAC inhibitor 57a, this constitutes a remarkable reversal in enzyme class speciﬁcity.

4.4.5. Mode of Action of JMJD2A Inhibition

As the last step of the in vitro characterization of hydroxamic acid-based inhibitors of JMJC demethylases, their mode of action was elucidated. As these compounds were derived from lead structures for potent inhibitors of zinc-dependent histone deacetylases (HDACs) with the well-established Zn2+-chelating hydroxamate warhead, a number of possible modes of action

129 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases could be envisaged to explain the observed JMJD2A inhibition, which may be confounding data analysis. Firstly, the coupled enzyme formaldehyde dehydrogenase (FDH) employed in the JMJD2A activity assay is also a zinc-dependent enzyme,[200] which might be aﬀected by test compounds. However, this can be ruled out because all potent hydroxamate-based JMJD2A inhibitors also exhibited their corresponding eﬀects in the LANCEUltra assay, in which no FDH is used. Moreover, some examples of this structural class like 57a, 57e, and 57j were also subjected to the FDH-counterscreening assay (cf. Section 3.2) and did not notably inhibit FDH. Secondly, JMJC demethylases of the KDM4 subfamily like JMJD2A contain a structural Zn2+ ion in their active site and it has been shown that ejection of this zinc(II) ion can also lead to potent inhibition (cf. Section 1.4.7).[172] However, this can also be ruled out as the mode of action of the hydroxamate-based compounds discussed in this section as they were shown to also inhibit representatives of the KDM5 and KDM6 subfamilies with equal or even greater potency. These enzymes do not contain such a structural Zn2+ ion. Lastly, the competitivity of inhibition towards the enzyme co-substrate 2-oxoglutarate (2-OG) was also investigated for representative examples 57a and 63c. The experimental details of such kinetic analyses have been discussed previously (cf. Section 4.2.2) and the results are summarized in Figure 4.9. As expected, for both of these compounds, the kinetic analysis revealed competitive behavior with regard to the co-substrate 2-oxoglutarate. This is illustrated by the fact that the maximum enzyme velocity vmax remained unchanged with increasing concentrations of inhibitor, while the apparent KM values increased. Similarly, the double reciprocal plot of these data (Lineweaver-Burk representations in Figures 4.9(b) and 4.9(d)) revealed that all lines inter- sected in one point on the 1/v-axis as required for competitive inhibition.

Further analysis of these data led to an assay-independent inhibition constant Ki of 15.3 µM for 57a and 8.36 µM for 63c, respectively, consistent with the improved in vitro potency of this compound as estimated by the inhibition data obtained in the LANCEUltra assay. Taking these data together suggests that inhibition of JMJD2A by hydroxamate-based compounds is indeed competitive to 2-oxoglutarate and that this competitive displacement of the co-substrate is the true mechanism, by which these inhibitors exert their potency.

130 4.4. Hydroxamic Acid-Based Inhibitors

(a) Inhibition of JMJD2A by 57a (b) Lineweaver-Burk representation of data

Figure 4.9.: Inhibition of JMJD2A by hydroxamic acid-based inhibitors 57a and 63c is competitive to co-substrate 2-oxoglutarate (2-OG). See text for explanations. Shown are mean ± s. d. of duplicates.

4.4.6. Cellular Activity

Encouraged by the good results from the in vitro testing and characterization of the newly discovered hydroxamic acid-based inhibitors, their biological eﬀects were also assessed in cell proliferation assays on two cancer cell lines. The esophageal cancer cell line KYSE-150[193] was tested because it had been reported to be sensitive to JMJC demethylase inhibition.[124]

131 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.20.: Cellular potency of hydroxamic acid-based inhibitors on KYSE-150 esophageal cancer and HL-60 leukemia cells. Adapted from Ref. 194.

KYSE-150 cells[193] HL-60 cells % growth inhibition GI50 / µM 63a 12% @ 50 µM 13% @ 50 µM 63b 51% @ 50 µM 14 ± 1 63c 14% @ 100 µM 116 ± 14 63d 82% @ 100 µM 18 ± 9

For comparison, the promyelocytic leukemia cell line HL-60 was also tested. These cell proliferation assays were performed by Dr. Inga Hoffmann assisted by Karin Schmidtkunz (Jung group) and the data in detail have been discussed in her thesis.[78] Table 4.20 summarizes the observed in vivo effects on cell proliferation of compounds 63a–d from the propanoate side chain series. KYSE-150 cells were found to be surprisingly insensitive to treatment with these compounds as judged by pre-tests at relatively high concentrations. However, proliferation of the leukemia cell line HL-60 could be inhibited by these JMJC demethylase inhibitors. Importantly, N - alkylated inhibitors 63a–d were found to be devoid of HDAC inhibitory activity (cf. Table 4.19) so that this can be ruled out as a potential cellular off-target. Growth inhibition of these cells is, therefore, a JMJC demethylase inhibition-dependent effect. Moreover, methyl esters 63b and 63d showed the highest in vivo potency with comparable

GI50 values, while the free carboxylic acids 63a and 63c were dramatically less potent. This is expected as the negative charge on the carboxylate under physiological conditions hampers cell permeability. The esters can be expected to penetrate the cells much more eﬃciently and may then be cleaved intracellularly by unspeciﬁc esterases. It that sense, 63b and 63d can be considered prodrugs of 63a and 63c, respectively.

In summary, screening of a focused library of hydroxamic acid-based inhibitors of zinc-dependent histone deacetylases (HDACs) has led to the discovery that 57a was also a relatively potent inhibitor of JMJD2A. Its structure-activity relationship has been studied extensively and the

132 4.4. Hydroxamic Acid-Based Inhibitors mode of action of inhibition has been elucidated. Importantly, a synthetic strategy has been devised that allows for the creation of structural analogs, which are similarly potent JMJC demethylase inhibitors, but devoid of HDAC inhibitory potency. It were precisely these HDAC- inactive inhibitors 63a–d that exhibited remarkable cellular potencies in a leukemia cell line, which makes them a valuable starting point for the development of biological tool compounds or candidate molecules for therapy.

133 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

4.5. Aminopyrimidylpyridine-Based Inhibitors

4.5.1. Discovery

The lead structures discussed in this section were discovered by a virtual screening campaign in collaboration with the Sippl group (Institute of Pharmacy, Martin-Luther-University Halle- Wittenberg). The methods employed for screening, the results of the iterative lead structure optimization, and all experimental details have been published in Ref. 201. The methodology for virtual screening of JMJD2A inhibitors was developed by Martin Pippel (Sippl group) and is described elsewhere.[201] Shortly, a selection of known metal-binding motifs found in previously reported JMJC demethylase inhibitors was gathered and the ZINC drug-like database of commercially available compounds was screened for structures containing these motifs. The metal-binding motifs that were selected are depicted in Fig. 4.10(a) and their proposed metal-coordinating atoms are highlighted. The step-by-step workflow for virtual screening is outlined schematically in Figure 4.10(b). By applying the filter for putative metal binders, the database of ∼18 million compounds was reduced to ∼530,000 structures. These were docked into an ensemble of available crystal structures of JMJD2A with previously published inhibitors and the results used to create a list of promising hit compounds. Based on the docking score, the structural diversity, and the purchasability of these structures, a first list of virtual screening hits was generated and tested in vitro for inhibition of JMJD2A. The results for the 20 selected virtual screening hits are summarized in Table 4.21. As discussed in the development of the screening platform (cf. Section 3.3.5), screening was generally performed using the FDH-coupled assay first and hits then validated in the LANCEUltra system. Several compounds exhibited substantial auto-fluorescence at the wavelengths used for the FDH-coupled assay. This was noticeable when the fluorescence intensity (FI) values at t = 0 were already markedly different from well to well and increased in a concentration- dependent manner. At the initiation of the demethylation reaction, no formaldehyde can have formed yet and FI should be equal in all wells. These compounds, which were, therefore, not testable in the FDH-coupled assay, could be evaluated in the LANCEUltra system due to the use of time-resolved fluorescence made possible by lanthanide chelate fluorescence and the

134 4.5. Aminopyrimidylpyridine-Based Inhibitors

(a) Metal-binding motifs selected for ﬁltering the ZINC drug-like database. Note that ‘Ar’ denotes any nitrogen-containing aromatic ring.

(b) Virtual screening workﬂow (c) Substructure search for analogs of 86

Figure 4.10.: Virtual screening campaign towards novel JMJC demethylase inhibitors. Adapted under license from Ref. 201.

much larger Stokes shift induced by the FRET eﬀect, which together enable measurements without interference from ﬂuorescing organic compounds. Surprisingly, most of these compounds were devoid of any inhibitory activity (i. e. less than 50% inhibition at the highest screening concentration of 400 µM in the FDH-coupled assay) in spite of their positive outcome in the docking procedure. The only compounds, which showed any noteworthy inhibitory potency, were 76, albeit very weakly, 85, 86a, and 86b. Compound 76 is essentially an analog of the well-known reference inhibitor pyridine-2,4-dicarboxylic acid (2,4-PDCA).[126] Enlargement of the pyridine to the quinoline system apparently dramatically reduced its potency. The 2,2’-bipyridyl compound 85 was also, unsurprisingly, a relatively potent inhibitor as such 2,2’ systems are well-known metal ion chelators. Closely related derivatives with esters and amides of one carboxylic acid have already been reported as inhibitors of JMJD2E (cf. Section 1.4.5).[157]

135 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.21.: Results of in vitro testing of the hits from virtual screening against JMJD2A. Data shown are mean ± s. d. of two independent experiments. ﬂuo – ﬂuorescence artifacts. n. i. – no inhibition. n. t. – not tested. FDH-coupled assay LANCEUltra assay Compound Structure IC50 / µM IC50 / µM

68 n. i. n. t.

69 ﬂuo n. t.

70 n. i. n. t.

71 n. i. n. t.

72 ﬂuo n. t.

73 ﬂuo n. t.

74 n. i. n. t.

75 n. i. n. t.

76 360 ± 36 n. t.

136 4.5. Aminopyrimidylpyridine-Based Inhibitors

Table 4.21.: (continued) FDH-coupled assay LANCEUltra assay Compound Structure IC50 / µM IC50 / µM

77 n. i. n. t.

78 n. i. n. t.

79 ﬂuo n. t.

80 n. i. n. t.

81 n. i. n. t.

82 ﬂuo n. i.

83 ﬂuo n. i.

84 n. i. n. t.

85 5.99 ± 0.88 0.467 ± 0.044

137 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.21.: (continued) FDH-coupled assay LANCEUltra assay Compound Structure IC50 / µM IC50 / µM

86a ﬂuo 3.98 ± 1.26

86b ﬂuo 0.941 ± 0.052

The most interesting compounds in this series were the aminopyrimidylpyridines 86a and 86b as they were remarkably potent in the LANCEUltra assay and also provided a novel, heretofore unreported, lead structure scaffold. As such, these were considered the hit compounds from the virtual screening campaign that were, consequently, followed up on in larger structure-activity relationship (SAR) investigations. These compounds also exhibited auto-fluorescence when tested in the FDH-coupled assay, but could readily be characterized using the LANCEUltra technology. Remarkably, compounds 86a and 86b showed a dramatic 4-fold difference in potency even though only the substitution pattern on the distal pyridine ring was changed. This difference could be confirmed with new batches of both compounds.

4.5.2. Structure-Activity Relationship (SAR)

Encouraged by the discovery of a novel lead scaffold and its remarkable in vitro potency in the single-digit micromolar range in the LANCEUltra assay, another virtual screening of the ZINC drug-like database was initiated, this time filtering specifically for structures containing the newly discovered 2-(2-aminopyrimidin-4-yl)pyridine-4-carboxylate motif. The workflow for this structure-based screening, which was again performed by Martin Pippel (Sippl group), is summarized in Figure 4.10(c). Based on the docking poses of the retrieved 381 hits, a total of 41 additional compounds was acquired in several iterative optimization steps covering a broad range of structural diversity with regard to the substitution pattern at the distal aromatic ring system.

138 4.5. Aminopyrimidylpyridine-Based Inhibitors

The structures of all compounds following the general lead structure 86 and their in vitro potencies are summarized in Table 4.22. The predicted binding pose of compound 86a in the active site of JMJD2A as obtained from the docking set-up suggests chelation of the central ferrous ion via the pyridyl and pyrimidyl nitrogen atoms, while the free carboxylate group is engaged in interactions with the conserved active site residues lysine K206 and tyrosine Y132 and a water molecule. This is in close analogy to the previously reported bipyridyl-carboxylate compounds 22.[157] Additionally, the secondary amine forms a hydrogen bond with glutamate E190, the amino acid, which also complexes the Fe2+ center. This is the reason why only the distal substituent was exchanged in a first attempt at elucidating the SAR of these compounds. Based on their docking score, interaction pattern, and a structural diversity as large as possible, an initial subset of 18 compounds was acquired and tested (cf. compounds 86a–r, Table 4.22). Virtually all compounds showed promisingly potent inhibition of JMJD2A in the single-digit micromolar range between 0.9 and 8.1 µM. Particularly potent were those derivatives with small aromatic substituents like 86b (pyridine, 0.9 µM), 86j (N -imidazolylethyl, 1.8 µM), and 86l (imidazole, 1.4 µM), suggesting that such substituents may be involved in an additional interaction in the active site, like e. g. hydrogen bonds with another active site residue or with the water-filled subpocket, into which these residues protrude. However, interactions with or disruptions of the water network could not be predicted by the docking procedure. Only two compounds of this original subset showed notably poorer inhibition, i. e. 86d with a particularly large aryl substituent (19 µM) and 86q (29 µM). This compound has a branched substituent adjacent to the methylene bridge, which makes it a chiral molecule. As it was commercially available only as racemate, this complicates a detailed analysis of the SAR for this structure. In both cases, the distal substituents were likely too large to be accommodated easily in the active site. Following these encouraging results of the first inhibitor subset, more analogs were searched for computationally, acquired, and evaluated in vitro. The results for this second inhibitor subset (86s–ab) are included in Table 4.22.

This set, again, contained numerous potent inhibitors, with IC50 values ranging from 1.6 to 10.2 µM. The substituents, by and large, conﬁrmed the already established SAR. Notably, the most potent derivatives were again those with small basic side chains like in 86u (imidazolyl-

139 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases methyl, 1.6 µM) and 86w (N -ethylimidazole, 2.0 µM). Aliphatic side chains like piperidine in 86ab were not well tolerated (9.5 µM). Interestingly, the pyridyl derivative 86v was found to be several times less potent (IC50 = 7.4 µM) than the original hits 86a and 86b, possibly due to an inappropriate orientation of the nitrogen atom or because of the increased distance by an extra CH2 spacer. In these ﬁrst two inhibitor sets (86a–r and 86s–ab), most compounds carried quite polar substituents. This, together with the also relatively polar scaﬀold and the negatively charged carboxylic acid, led to the assumption that these compounds would likely not be cell-permeable. Therefore, another set of inhibitors was screened for in the ZINC drug-like database with a particular emphasis on the physicochemical properties of the compounds and a set of derivatives with more lipophilic residues was acquired (86ad–al). The benzyl derivative 86ac was synthesized using the synthetic platform for prodrugs (see below, Section 4.5.6). The structures and in vitro potencies from LANCEUltra assays for this set are included in Table 4.22. As expected, all compounds were quite potent JMJD2A inhibitors. Notably, this set with

IC50 values of 0.37 to 4.1 µM was approximately two-fold more potent than the previous sets.

Remarkably potent compounds were 86ag and 86aj with IC50 values of 0.37 and 0.45 µM, respectively. Both compounds also contain small aromatic residues, which can function as hydrogen bond acceptors like the chromane or methylsulfanyl residue. This is in line with the previous observation that small basic residues lead to favorable additional interactions in this part of the active site (cf. 86b, 86j, 86l, and 86u). It should be noted that the change from the pyridyl substituent in 86a and 86b to the benzyl moiety in 86ac led to no dramatic loss in potency (IC50 = 2.7 µM), even though the presumed additional interaction in the water-ﬁlled subpocket was no longer possible. However, the lipophilicity as judged by the calculated partition coeﬃcient clogP increased by more than one unit, which was a promising improvement regarding the expected cell permeability of this compound. Taking together the data from the SAR observed so far and the structural information from the docking poses, a last optimized set of inhibitors was suggested and acquired, i. e. compounds 86am–ar. Indeed, this set again contained derivatives of remarkable in vitro potency with

IC50 values in the low single-digit micromolar range. Of particular interest were compounds

86an and 86ar with bridged catechol structures leading to IC50 values of 0.92 µM and 1.3 µM,

140 4.5. Aminopyrimidylpyridine-Based Inhibitors

Table 4.22.: IC50 values of aminopyrimidylpyridine-based inhibitors of the general lead structure 86 against JMJD2A in LANCEUltra assays. Data shown are mean ± s. d. of two independent experiments.

R = IC50 / µM R = IC50 / µM

86a 3.98 ± 1.26 86m 6.58 ± 1.48

86b 0.941 ± 0.052 86n 4.76 ± 2.33

86c 2.71 ± 0.36 86o 3.27 ± 0.04

86d 18.9 ± 0.5 86p 5.89 ± 4.13

86e 3.12 ± 0.38 86q 28.5 ± 4.2

86f 3.92 ± 0.09 86r 1.60 ± 0.24

86g 8.07 ± 0.80 86s 7.02 ± 0.22

86h 3.89 ± 0.97 86t 3.64 ± 0.14

86i 5.10 ± 1.10 86u 1.57 ± 0.09

86j 1.83 ± 0.37 86v 7.42 ± 1.84

86k 7.16 ± 0.89 86w 1.95 ± 0.11

86l 1.40 ± 0.22 86x 4.24 ± 0.36

141 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.22.: (continued)

R = IC50 / µM R = IC50 / µM

86y 5.70 ± 0.55 86ai 2.03 ± 0.26

86z 3.26 ± 0.30 86aj 0.446 ± 0.138

86aa 10.2 ± 3.5 86ak 4.11 ± 1.41

86ab 9.53 ± 0.59 86al 0.840 ± 0.051

86ac 2.70 ± 0.29 86am 1.04 ± 0.04

86ad 2.41 ± 0.47 86an 0.918 ± 0.214

86ae 2.06 ± 0.27 86ao 0.415 ± 0.093

86af 2.36 ± 0.23 86ap 7.31 ± 0.76

86ag 0.370 ± 0.028 86aq 5.77 ± 0.08

86ah 1.06 ± 0.09 86ar 1.30 ± 0.28

respectively, reinforcing the notion that hydrogen bond acceptor moieties in this position can lead to favorable interactions. Remarkably, derivatives, in which this motif is reversed, like 86ap and 86aq, were signiﬁcantly less potent inhibitors. On the other hand, these are again chiral compounds, where the eﬀects of the individual enantiomers could not be elucidated. In summary, a large panel of potent in vitro JMJD2A inhibitors based on the aminopyrimidylpyridine-carboxylate motif was discovered and its structure-activity relationship elucidated with a wide range of diverse substituents. Limitations were observed for excessively large

142 4.5. Aminopyrimidylpyridine-Based Inhibitors residues and aliphatic side chains, while small aromatic residues, in particular those comprising a hydrogen bond or dipole-dipole interaction acceptor moiety in a favorable orientation, usually resulted in improved potency. Of note are the most potent inhibitors 86ag, 86aj, and the thiophene compound 86ao, each with IC50 values as low as ca. 400 nM.

4.5.3. Modiﬁcations to the General Lead Structure

In order to test the relevance of the aminopyrimidylpyridine-4-carboxylate motif in the newly discovered lead structure 86, several commercially available control compounds were acquired with deviations from this general binding motif and evaluated for their JMJD2A inhibitory potential in the LANCEUltra assays. As was established from the structure-activity relationship (SAR) observed for 86, the distal aromatic substituents had only a small influence on potency and, generally speaking, all gave highly potent inhibitors (cf. Table 4.22). The control compounds all contain different distal side chains on the secondary amine. However, as the differences in potency induced by these are comparably small, they can still serve as suitable controls. The results of these test compounds are summarized in Table 4.23. Notably, a compound, in which the secondary amine present in lead structure 86 is incorporated into a piperazine ring as in 87, was dramatically less potent compared to the series in Table 4.22 by a factor of at least 4- to 5-fold. This is likely due to the loss of the hydrogen bond donor function of the secondary amine and the presumed interaction that this can undergo with the active site residue glutamate E190. Likewise, the shortened compound 88 lacking the amine altogether exhibited reduced potency considering its small basic aromatic residue, which should usually give quite favorable inhibitory properties. Compound 89 lacking the 4-carboxylic acid was dramatically less potent by around 10-fold compared to similarly substituted derivatives as this compound was no longer able to engage in the predicted ionic interaction with active site residue lysine K206. Given that ionic interactions largely contribute to the binding enthalpy, loss of this interaction dramatically reduced the potency of inhibition. Consequently, a test compound 90 lacking both the 4-carboxylate as well as the hydrogen bond capacity on the amine was devoid of any inhibitory activity at the highest tested concentration of 250 µM.

143 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.23.: IC50 values of control compounds to validate the lead structure 86 against JMJD2A. Data shown are mean ± s. d. of two independent experiments.

LANCEUltra assay Compound Structure IC50 / µM

87 18.2 ± 7.4

88 7.47 ± 3.35

89 34.8 ± 7.9

90 > 250

91 > 250

92 202 ± 16

As expected, shifting the substitution pattern of the pyrimidylpyridine system away from the known 2,2’ pattern as in mocetinostat 91, a known inhibitor of histone deacetylases (HDACs),[202, 203] likewise led to complete loss of inhibition as this compound cannot chelate the central ferrous ion present in the active site of JumonjiC domain-containing histone demethylases. In addition, it lacks the highly important carboxylate group. Compound 92 illustrates the importance of the ﬂexible methylene bridge linking the secondary amine to the aromatic substituents in 86 as rigidiﬁcation to the sulfonamide linker together with omission of the carboxylate led to a near complete loss of potency.

144 4.5. Aminopyrimidylpyridine-Based Inhibitors

Table 4.24.: IC50 values of synthesized analogs and prodrugs of structure 86 against JMJD2A. Data shown are mean ± s. d. of two independent experiments.

LANCEUltra assay Compound Structure IC50 / µM

93b 163 ± 11

93ac 14.6 ± 1.4

93ag 3.38 ± 0.37

93aj > 250

94b 67.7 ± 1.5

94ac > 500

The establishment of a synthetic platform towards prodrugs and commercially not available analogs of lead structure 86 (cf. Section 4.5.6) allowed for the generation of more analogs with which to test the predicted SAR of these inhibitors. The structures and in vitro potencies of the synthesized analogs are summarized in Table 4.24. Expectedly, all methyl ester prodrugs 93 exhibited drastically reduced in vitro potency by at least 5-fold when compared to their free carboxylate analogs (compare 93b to 86b and so forth). This can be explained with the loss of the beneﬁcial ionic interaction of this carboxylate with the highly conserved lysine residue K206. Likewise, analogs, in which the carboxylate substituent was omitted entirely (compounds 94b and 94ac) also exhibited markedly reduced inhibition when compared to their carboxylate counterparts (86b and 86ac, Table 4.22).

145 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Taking all these in vitro data together, it can be concluded that the main pharmacophore required for efficient inhibition is indeed the 2-(2-aminopyrimidin-4-yl)pyridine-4-carboxylate core, thus confirming the predicted docking pose generated by virtual screening. This means that the pyrimidylpyridine core is required for bidentate chelation of the central ferrous ion, while the 4-carboxylic acid is engaged in an ionic interaction with the active site residue lysine K206 and the secondary amine is engaged in a hydrogen bond with glutamate E190. Modula- tion of any of these structural features led to a stark drop in potency. The aromatic substituents attached to the secondary amine via a flexible methylene (or longer) spacer confer additional lipophilic interactions or possibly hydrogen bonds with the water-filled subpocket, but do not dramatically change the observed in vitro potency. The inherent flexibility of the water network, however, made it impossible to predict these interactions in the docking procedure, so that a clearer structure-based understanding of the SAR for this distal part of the molecule was obscured. Lastly, a set of putative test compounds was acquired, which do not follow the lead structure 86, but should be able to yield quite potent in vitro inhibitors based on the SAR observed so far. These can be used to test the scope of the established SAR. Their structures and IC50 values in LANCEUltra assays against JMJD2A are summarized in Table 4.25. It should be noted that compounds 95 through 98 are chiral, but were available only as racemates so that the individual contributions of either enantiomer could not be assessed. Compounds 95 through 97 demonstrate that also bridged derivatives with an additional alkyl substituent on the otherwise unsubstituted methylene spacer could quite potently inhibit JMJD2A with only a slight drop in potency. Curiously, for the pyridyl derivative 96, introduction of the ethyl residue led to a loss in potency compared to 86b, while 97 exhibited slightly improved potency when compared to its unalkylated analog 86a. In any case, rotational flexibility around this spacer is still preserved so that the aromatic ring system can orient properly into the water-filled subpocket. Compound 98 demonstrates that the hydrogen bond donor capacities fulfilled by the secondary amine in lead structure 86 could also be replaced by a hydroxyl group with only a slight decrease in potency. This may also be due to the fact that the H-bond donor position was shifted by one bond length, which may require a slight reorientation of the accepting amino acid residue, glutamate E190.

146 4.5. Aminopyrimidylpyridine-Based Inhibitors

Table 4.25.: IC50 values of analogs with deviations from the lead structure 86 against JMJD2A. Data shown are mean ± s. d. of two independent experiments.

LANCEUltra assay Compound Structure IC50 / µM

95 4.78 ± 0.71

96 16.3 ± 2.4

97 2.64 ± 0.09

98 7.32 ± 0.74

99 3.94 ± 0.26

100 12.3 ± 3.9

101 0.800 ± 0.080

Compound 99 was also a potent JMJD2A inhibitor even though the free methylene bridge is rigidiﬁed in this ring system. Compound 100 showed the expected drop in potency observed for compounds lacking the secondary amine. Surprisingly, the conformationally constrained test compound 101 was remarkably potent with an IC50 value in the sub-micromolar range. However, there was only one such compound available, which complicates the deduction of a clear structural explanation for this observation.

147 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

4.5.4. Structural Binding Model – X-Ray Crystallography

In order to better understand the structural binding mode of inhibitors of the lead structure 86 in the active site of JMJD2A, it was attempted to obtain a co-crystal x-ray structure of one such inhibitor with the catalytic domain of the enzyme. Such co-crystals were obtained by the team of Dr. Ralf Flaig (Diamond Light Source, Oxford), analyzed by x-ray to a resolution of 2.0 A,˚ and a structural model was established, which is depicted in Figure 4.11. Experi- mental details for these crystallization experiments and structure analysis have been published in Ref. 201 and the coordinates of the structure model deposited to the Protein Data Bank (www.pdb.org) under accession code 5ANQ. Figure 4.11(a) shows an overview over the entire catalytic domain of JMJD2A in complex with the inhibitor 86a. As expected from the literature and numerous previously published crystal structures of this enzyme, the catalytic JMJC domain folds into the jelly-roll-like all-β-barrel fold typical for enzymes of the cupin superfamily (cf. Section 1.2.2.3).[64, 72] Figure 4.11(b) shows a detailed view of the catalytic domain, which contains the central Fe2+ ion bound by two histidine and one glutamate residue as well as inhibitor 86a and a water molecule. Importantly, the detailed analysis of the binding situation of 86a and the adjacent amino acids (cf. Figure 4.11(c)) coincides with the predicted binding mode from docking and con- firms the structure-activity relationship (SAR) derived from the in vitro inhibitory data of all compounds. In particular, it shows how the nitrogen atoms of both the pyridine and the pyrimidine ring chelate the central ferrous ion in a bidentate manner. At the same time, the 4-carboxylic acid is oriented towards the highly conserved residue lysine K206, with which it undergoes an ionic interaction. Moreover, the phenolic OH group of tyrosine Y132 is oriented towards the carboxylate allowing for an additional hydrogen bond. As expected, the secondary amine is involved in another hydrogen bond with iron-complexing glutamate E190. The distal aromatic ring is twisted away by the flexible methylene bridge and protrudes into another subpocket shaped by residues aspartate D191, valine V171, and tyrosine Y175. In this pocket, likely only unspecific hydrophobic interactions are formed. However, the typical water network observed in this subpocket in other crystal structures was not resolved in this case. The phenylalanine residue F185 is oriented such that it can engage in a favorable π-stacking interaction with the pyridine ring, further stabilizing its binding.

148 4.5. Aminopyrimidylpyridine-Based Inhibitors

(a) Overview over the JMJC domain (b) 86a in the active site of JMJD2A

Figure 4.11.: Co-crystal structure of 86a in the active site of JMJD2A. Carbon atoms are depicted in green, nitrogen in blue, and oxygen in red. The central ferrous ion is shown as a golden sphere. Adapted under license from Ref. 201.

149 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Figure 4.12.: Comparison of the binding mode of 86a (green) in the active site of JMJD2A to that of previously published pyridine-2,4-dicarboxylic acid 6[126] (pink) and 22b[157] (cyan). Adapted under license from Ref. 201.

The surface representation of the active site in Figure 4.11(d) emphasizes the necessity of the rotationally ﬂexible methylene linker in 86a as this molecule needs to adopt the depicted bent conformation in order to avoid steric clashes with the edge of the catalytic pocket. This explains why compounds with excessively large substituents or conformationally constrained linkers like the sulfonamide 92 were dramatically less potent. The obtained co-crystal structure of 86a in complex with JMJD2A allows for the comparison of the binding mode of these compounds to that of previously published inhibitors like pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6[126] or the closely related bipyridyl inhibitors like 22b.[157] A superposition of these published co-crystal structures with the one in complex with 29a is depicted in Figure 4.12.

150 4.5. Aminopyrimidylpyridine-Based Inhibitors

The pyridine ring adopts the same position complexing the ferrous ion for all three inhibitors as does the 4-carboxylic acid in contact with lysine K206. The second atom complexing Fe2+ is the nitrogen atom of the second pyridine ring in 22b, respectively the pyrimidine ring in 86a, or an oxygen atom of the carboxylate group in 6. The pyrimidylpyridine system is slightly twisted out of the planar conformation in order to adopt this binding pose and to position the secondary amine in proximity to glutamate E190. The novel inhibitor 86a is the ﬁrst such compound to address this residue in a speciﬁc binding interaction and to protrude all the way to the edge of this catalytic pocket.

4.5.5. Competitivity of Inhibition Towards 2-Oxoglutarate

The binding mode that is suggested by the x-ray co-crystal structure (cf. Figure 4.11) and the predicted binding from docking is corroborated further by enzyme kinetic experiments. The structure of 86a in complex with JMJD2A suggests that compounds of this type inhibit the enzyme by bidentate chelation of the central ferrous ion and, thus, competitive displacement of the enzyme co-substrate 2-oxoglutarate (2-OG). This was tested by employing the LANCEUltra assay for JMJD2A with the most potent inhibitor 86ag in the presence of varying amounts of 2-OG. The results of such experiments are summarized in Figure 4.13. As outlined in detail for the tetrazolyl hydrazide inhibitors (cf. Section 4.2.2), a Michaelis-Menten type saturation curve is observed for JMJD2A with regard to its co-substrate in the absence of any inhibitor (cf. Figure 4.13(a)). With increasing amounts of the competitive inhibitor 86ag, these saturation curves shift to the right, i. e. to higher apparent KM values, while they all approach the same maximum velocity vmax. This is also illustrated by the double reciprocal representation of the data set (Figure 4.13(b)). This yields a linear trend for every inhibitor concentration, while all lines intersect in one point on the 1/v-axis, a characteristic of competitive inhibition.

The increase in apparent KM values is linear with regard to the concentration of inhibitor 86ag (cf. Figure 4.13(c)) and can be used to determine an assay-independent inhibition constant

Ki, which is as low as 186 nM for this compound, making it not only the most potent in vitro inhibitor described in this study, but also one of the most potent JMJD2A inhibitors described in the literature to date.

151 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

(a) Inhibition of JMJD2A by 86ag (b) Lineweaver-Burk representation of data

Figure 4.13.: Inhibition of JMJD2A by optimized pyrimidylpyridine inhibitor 86ag is competitive to co-substrate 2-OG. See text for explanations. Shown are mean ± s. d. of duplicates. Modiﬁed under license from Ref. 201.

The observation that inhibition of JMJD2A is competitive with regard to the co-substrate 2-OG can also be reproduced for the two original screening hits, compounds 86a and 86b as depicted in Figure 4.14. As compound 86b was also found to be a ca. 4-fold more potent inhibitor than 86a in the LANCEUltra assay, the observed eﬀects are markedly more pronounced for this compound and the increase in apparent KM values is steeper. For these compounds, inhibition constants Ki of 1.80 µM and 579 nM are obtained, respectively. Thus, the inhibition constants also reﬂect the relative trend in potencies as judged from the IC50 values.

152 4.5. Aminopyrimidylpyridine-Based Inhibitors

(a) Inhibition of JMJD2A by 86a (b) Ki = 1.80 µM

Figure 4.14.: Inhibition of JMJD2A by pyrimidylpyridine inhibitors 86a and 86b is competitive to co-substrate 2-OG. See text for explanations. Shown are mean ± s. d. of duplicates. Modiﬁed under license from Ref. 201.

4.5.6. Synthesis of Derivatives and Prodrugs

Encouraged by the remarkable in vitro potency of many inhibitors of the lead structure 86, it was attempted to also elucidate their in vivo eﬀects in cell culture experiments. However, the quite polar nature of many molecules and their side chains together with the carboxylate function, which can be expected to carry a negative charge under physiological conditions, led

153 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Figure 4.15.: Synthesis scheme for methyl ester prodrugs 93 and other derivatives of pyrimidylpyridine inhibitors 86. Modiﬁed under license from Ref. 201.

to the assumption that these compounds were likely not going to be very well cell-permeable and, therefore, not able to exert any in vivo potency. Therefore, a prodrug strategy was envisaged by synthesizing the corresponding methyl esters of these inhibitor compounds. These esters would mask the negative charge of the carboxylate and, thereby, allow the compounds to pass the lipid bilayer cell membrane. Inside the cell, they would be cleaved by unspeciﬁc intracellular esterases releasing the free and potent carboxylate. The use of esters as prodrugs of carboxylic acids is very well documented in the literature with a number of clinically used examples.[204] The synthesis of the methyl ester prodrugs could be accomplished according to the synthesis plan depicted in Figure 4.15. The key step in this multi-step synthesis is the formation of the aminopyrimidylpyridine system by condensation of the corresponding guanidines 104 with enaminone 107. This was inspired by the synthesis of previously published HDAC inhibitor mocetinostat (91, Table 4.23).[202, 203] This would allow for a modular synthetic approach, in which diﬀerent residues could be introduced into the lead structure 86 through synthesis of the required guanidines all the while using the same pyridylenaminone 107. Therefore, this synthetic platform allowed not only for the synthesis of methyl ester prodrugs of 86, but also

154 4.5. Aminopyrimidylpyridine-Based Inhibitors for the creation of novel inhibitors with different side chain substitutions that were commercially not available, like e. g. the benzyl derivative 86ac. The guanidines 104 could be synthesized from readily available primary amines 102 using the well-established guanidinylation reagent 1H -pyrazole-1-carboxamidine hydrochloride 103.[205] Using this procedure, the guanidines could be obtained in good to excellent yields and crystallized during the work-up procedure to yield 104 in sufficient purity for the subsequent coupling step. Other guanidinylation reagents like cyanamide never gave any detectable product under the tested conditions. The pyridylenaminone 107 was built up in two steps from inexpensive starting materials. Methyl isonicotinate 105 was acetylated in a radical iron(II)-catalyzed Minisci-type reaction following literature procedures.[206, 207] This acetyl derivative 106 was then transformed into the dimethylamino enaminone 107 in good yield by aldol condensation with N,N -dimethylformamide-dimethylacetal (DMF-DMA) under microwave conditions requiring virtually no work-up or purification.[208] The condensation of enaminone 107 and guanidines 104 to yield the aminopyrimidine heterocycle was performed in analogy to the reported synthesis of mocetinostat 91.[202] This involved reaction in refluxing iso-propanol under addition of molecular sieve granules in order to cap- ture the dimethylamine, which is released as a coupled product during the condensation. The yields for these reactions varied from only 10% to more acceptable 42%, in large parts due to the troublesome work-up of the reaction mixture and isolation of the cyclized products 93. This generally involved several chromatographic steps, because compounds 93 needed to be isolated from often rather viscous byproducts. Moreover, as the reaction was carried out at elevated temperatures in iso-propanol as solvent, in some cases, a substantial amount of the iso-propyl esters of 93 was also formed as revealed by 1H-NMR spectroscopic analyses of the reaction products. As the iso-propyl and methyl esters were virtually impossible to separate due to their similar physicochemical properties (two very close spots in thin-layer chromatography (TLC)), the reaction products then needed to be transesterified by refluxing again in methanol. Methanol as solvent for the condensation reaction would not reach a high enough reflux temperature. The reduced yields for the condensation reaction are certainly not satisfying especially considering that this is the key merging step of the modular synthesis. However, the focus in

155 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases this study was on the generation of derivatives that would meet the stringent purity criteria required for in vitro and in vivo testing (≥ 94%) rather than on the optimization of the reaction protocol towards higher yields. Moreover, the literature yield for a reaction with similar starting materials was also only 52%.[202] When novel side chain substitutions were introduced via this synthetic platform like the benzyl derivative 86ac, the corresponding free carboxylic acids could be generated from the methyl esters 93 by a standard saponification approach using aqueous lithium hydroxide and precipitation by addition of HCl yielding these acids 86 in excellent yields and purity. Furthermore, this modular synthetic strategy also allowed for the generation of control compounds lacking the carboxylate moiety altogether like derivatives 94 (cf. Table 4.24) when the sequence was started with simple 2-acetylpyridine. This was reacted with DMF-DMA and then coupled with the corresponding guanidines under the same conditions. As several of the methyl ester prodrugs only exhibited limited solubility in aqueous media required for cell culture experiments (cf. Section 4.5.7), a different prodrug was synthesized of the optimized inhibitor 86ag using the N -morpholinoethyl ester moiety, which can be expected to be much better soluble in aqueous media. As outlined in Figure 4.16, this was obtained by simple base-catalyzed transesterification of the methyl ester 93ag by stirring in an excess of N -morpholinoethanol yielding 108ag in excellent yield.

Figure 4.16.: Synthesis of the N -morpholinoethyl ester prodrug of pyrimidylpyridine inhibitor 86ag. Modiﬁed under license from Ref. 201.

156 4.5. Aminopyrimidylpyridine-Based Inhibitors

4.5.7. Cellular Eﬀects

In order to determine the biological effects of the highly potent inhibitors of the lead structure 86 and their prodrugs, cell culture studies were performed by Dr. Inga Hoffmann. In particular, cell proliferation experiments were carried out using the esophageal cancer cell line KYSE-150.[193] Experimental procedures and more detailed results have been published[201] and are discussed in her thesis.[78] KYSE-150 cells were chosen because of their reported elevated levels of the very closely related [100] H3K9me3 histone demethylase JMJD2C and because they had been shown to be sensitive to JMJD2A inhibition.[124] The results for selected compounds of this series are summarized in Figure 4.17. KYSE-150 cells were treated with inhibitors at different concentrations in media for 72 hours and then their proliferation was evaluated and normalized to a control experiment, in which cells were treated only with the organic solvent DMSO.

Despite its remarkable in vitro potency (IC50 = 0.94 µM, cf. Table 4.22), compound 86b did not show any noteworthy effect on cell proliferation. This is likely due to its highly polar nature and the negative charge on the carboxylate function under physiological conditions, making such a compound unlikely to cross the cell membrane. Intriguingly, its methyl ester prodrug 93b was able to efficiently reduce proliferation of these cancer cells by 50% at a concentration of 50 µM. In an effort to optimize the physicochemical properties of these test compounds, the pyridyl- methyl unit in 86b was replaced by a benzyl moiety yielding compound 86ac with an increased lipophilicity by more than one clogP unit, but similar in vitro potency (cf. Table 4.22). Re- markably, this compound was also able to reduce KYSE-150 cell proliferation by 44% at the same concentration even though this is the free carboxylate compound. The decreased polarity of the distal aromatic substituent apparently allowed this compound to permeate cells in spite of the charged carboxylate group. Compound 93ac, the methyl ester of the benzyl derivative could potentially be even more potent as it combines both favorable structural components. However, the increased lipophilicity of 93ac also made it insoluble in the aqueous cell culture medium at a high enough concentration to yield an effect. At a lower test concentration of 10 µM, only a 20% reduction in cell proliferation was observed.

157 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

The same problem of insolubility was also observed for the methyl ester prodrugs of the optimized in vitro inhibitors 86ag and 86aj, i. e. 93ag and 93aj. When the free carboxylates were tested, no eﬀect on cell proliferation was observed, which is likely due to their limited cell penetration. Therefore, the prodrug strategy was changed and the more hydrophilic N -morpholinoethyl ester prodrug 108ag was synthesized. Indeed, it was much better soluble in aqueous media allowing for cell culture experiments at concentrations of up to 250 µM, where it, expectedly, exhibited the highest reduction in cell proliferation (cf. Figure 4.17). In summary, these data demonstrate that aminopyrimidylpyridine-based inhibitors like the lead structure 86 can signiﬁcantly reduce proliferation of particularly sensitive KYSE-150 cells

Figure 4.17.: Effect of pyrimidylpyridine-based inhibitors 86 and their prodrugs 93 on proliferation of KYSE-150 cells. Data are mean ± s. e. m. of at least five wells per concentration. *** – significant decrease (p < 0.0001). Modified under license from Ref. 201.

158 4.5. Aminopyrimidylpyridine-Based Inhibitors when appropriate prodrugs are used to enable their cell penetration. The physicochemical properties of these compounds need to be fine-tuned in a narrow window between cell permeability and solubility. At the time of publication, these were the first optimized cell-permeable pyridine carboxylate-based JMJC demethylase inhibitors. Since then, a collaborative effort of the Structural Genomics Consortium (SGC) has developed compounds like 26, in which the carboxylate motif is masked in an annulated pyrimidinone system, which retains the binding properties, but does not carry a negative charge (cf. Section 1.4.5). These were optimized to be more potent and selective cell-permeable inhibitors than 86.[164]

4.5.8. Selectivity

Lastly, the selectivity of the newly discovered inhibitors of the lead structure 86 against a panel of JumonjiC domain-containing histone demethylases was investigated using the newly established LANCEUltra assays (cf. Section 3.3.6). The IC50 values obtained in these assays can be used as an approximate measure of a compound’s selectivity and are summarized in Table 4.26. As controls, the previously reported JMJD3-selective inhibitor GSK-J1 23a[158] and the relatively unselective inhibitor pyridine-2,4-dicarboxylic acid (2,4-PDCA) 6[126] are included for comparison.

Table 4.26.: Selectivity of pyrimidylpyridine-based inhibitors and their prodrugs against JMJD2A, JMJD3, and JARID1A in LANCEUltra assays. Data are mean ± s. d. of duplicates.

IC50 / µM Compound JMJD2A JMJD3 JARID1A 2,4-PDCA 6 0.0337 ± 0.0070 36.0 ± 4.7 0.100 ± 0.028 GSK-J1 23a 54.4 ± 1.6 0.128 ± 0.018 0.418 ± 0.153 86a 3.98 ± 1.26 16.2 ± 1.7 1.04 ± 0.22 86b 0.941 ± 0.052 36.5 ± 3.6 0.442 ± 0.067 93b 163 ± 11 387 ± 19 112 ± 35 94b 67.7 ± 1.5 ∼500 165 ± 17

159 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

The original screening hits 86a and 86b were potent inhibitors of JMJD2A and exhibited inhibition in a comparable range also of JARID1A. A slight preference for JARID1A by 2- to 3-fold could be observed. However, the enzyme concentrations and assay conditions also differ for the different JMJC demethylase subtypes. Notably, however, their inhibitory potential on JMJD3 was dramatically reduced with a difference of nearly 40-fold for 86b. This makes compounds of the general lead structure 86 potent inhibitors with dual selectivity for the KDM4 and KDM5 subfamilies over the KDM6 subfamily. This behavior coincides with the selectivity of the simple reference inhibitor 2,4-PDCA (cf. Table 4.26) and has recently also been observed for other more elaborate pyridine carboxylate-based inhibitors (cf. Section 1.4.5).[164–166] The methyl ester 93b and control compound 94b lacking the carboxylate moiety are, expectedly, much less potent inhibitors of JMJD2A and, consequently, also of JARID1A. With regard to JMJD3, their inhibitory potential is nearly not detectable, which coincides with the selectivity profile observed for the potent free carboxylates. Considering the similarity in potency of inhibition of JMJD2A and JARID1A, it should be pointed out that the observed cellular effects (cf. Section 4.5.7) may very well also be the result of inhibition of both or other JMJC demethylases.

In conclusion, by using virtual screening of a database of commercially available compounds, a novel lead structure scaﬀold 86 was identiﬁed for highly potent in vitro inhibitors of JMJD2A and JARID1A with remarkable selectivity over JMJD3. These have been optimized iteratively culminating in the discovery of 86ag with an inhibition constant Ki of only 186 nM. By establishing a synthetic platform towards non-commercial analogs as well as ester prodrugs of this lead structure, highly potent derivatives with antiproliferative cellular activity were obtained. The mode of binding in the active site of JMJD2A was elucidated by an x-ray co-crystal structure and can aid in the structure-based design of yet more potent analogs for therapeutic candidates and valuable biological tool compounds.

160 4.6. Clinically Used Iron Chelators

4.6. Clinically Used Iron Chelators

4.6.1. Discovery

In this section, the discovery and characterization of a novel JMJC demethylase inhibitor is described based on rational design. As JumonjiC domain-containing histone demethylases depend on a central Fe2+ ion in their active site for their enzymatic activity (cf. Section 1.2.2.2), it appeared interesting to test clinically used iron chelators for their potential inhibition of these enzymes. A number of small molecules have been developed and are used in the clinic for chelation of iron in the blood for the treatment of certain diseases related to iron overload (hemochromatosis). This can occur either as a hereditary disease or as a result from frequent blood transfusions. As humans have no mechanism to excrete excess iron, this accumulation of iron in the blood can have detrimental eﬀects on many organs, often related to the increased occurrence of reactive oxygen species promoted by redox-active metal ions in the blood. A summary of the chemical structures of commonly used iron chelators is given in Figure 4.18. The ﬁrst such drug was deferoxamine mesylate 37, which binds excess iron in the bloodstream and leads to excretion of the complexed iron. However, due to its low bioavailability, it needs to be administered by daily injections. More recently, deferasirox 109a and deferiprone 110 have been approved for treatment, which can be taken orally. As these molecules can potently complex iron species in solution in the blood, it seemed possible that they could also inhibit JMJC demethylases by complexation of the central ferrous ion in the active site and, thus, competitive displacement of the co-substrate 2-oxoglutarate (2-OG). This hypothesis was tested using all three iron chelators in the established JMJD2A inhibition assays. The results of both the FDH-coupled as well as LANCEUltra assays are summarized in Table 4.27. Surprisingly, all three iron chelators appeared to be remarkably potent in vitro inhibitors of

JMJD2A, with IC50 values in the single-digit micromolar range in the LANCEUltra assay.

Deferoxamine 37 appeared to be the most potent inhibitor with IC50 values as low as 3 µM.

However, for this compound, the typical diﬀerence between IC50 values measured in the FDH- coupled and LANCEUltra assay was not observed, which may already be an indication that the

161 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

(a) Deferoxamine mesylate 37 (b) Deferasirox 109a (c) Deferiprone 110

Figure 4.18.: Chemical structures of clinically used iron chelators.

Table 4.27.: Inhibition data of clinically used iron chelators as potential inhibitors of

JMJD2A. IC50 values are mean ± s. d. of duplicate experiments. FDH-coupled assay LANCEUltra assay Compound IC50 / µM IC50 / µM Deferoxamine 37 3.22 ± 0.30 3.33 ± 0.48 Deferasirox 109a 7.37 ± 1.58 4.76 ± 0.23 Deferiprone 110 17.4 ± 0.8 3.87 ± 0.27

apparent inhibition arose by mechanisms other than active site chelation within the enzyme. Alternatively, these compounds may also have inhibited the enzyme by sequestration of iron ions that are added to the buﬀer solution or by removal of iron ions from the enzyme if they form more stable complexes with the Fe2+ ion than does the enzyme. This seems particularly plausible for the hexadentate ligand deferoxamine 37 for entropic reasons. Such a sequestration mechanism would, however, not be a desirable starting point for the development of a drug candidate based on these motifs.

For deferasirox 109a and deferiprone 110, however, the typical trend was observed that IC50 values from the FDH-coupled assay were several times higher than those in the LANCEUltra assay due to the higher concentrations of enzyme and the competitive co-substrate 2-OG.

162 4.6. Clinically Used Iron Chelators

4.6.2. Elucidation of the Mode of Action

Based on the known properties of these compounds to be strong iron chelators, several diﬀerent modes of action can be envisaged to explain why they exhibited such high apparent in vitro potency as JMJD2A inhibitors. This includes the desired binding of the compound in the active site of the enzyme and competitive displacement of the co-substrate but also sequestration of Fe2+ ions in solution, which would make the observed inhibition an experimental artifact with no physiological relevance. In order to elucidate the mode of action, several further experiments were undertaken as outlined in the following.

4.6.2.1. Variation of Iron Content

Firstly, the in vitro inhibition experiments were repeated in the presence of different concentrations of iron(II) added to the assay buffer. If these compounds acted solely by sequestration of iron(II) in solution, the varying amounts should have a dramatic effect on the observed

IC50 values. Such experiments were already performed in the very ﬁrst collection of inhibitors for JMJD2E[126] to rule out alternative modes of actions of hydroxamate-based inhibitors and in the elucidation of the mode of binding of hydroxypyrazole-based JMJD2C inhibitors.[167]

However, no clear criteria have been established in the literature as to how variable IC50 values can be with regard to diﬀering iron concentrations.

Table 4.28.: Variation of potency of clinically used iron chelators with iron content in

JMJD2A assays. IC50 values are mean ± s. d. of duplicate experiments.

Assay type [Fe2+] Deferoxamine Deferasirox Deferiprone 5 µM 2.72 ± 0.03 4.89 ± 0.13 13.3 ± 1.3 FDH-coupled 10 µM 3.22 ± 0.30 7.37 ± 1.58 17.4 ± 0.8 20 µM 4.36 ± 0.24 11.9 ± 1.7 26.0 ± 1.4 2.5 µM 1.55 ± 0.02 1.64 ± 0.16 1.68 ± 0.07 LANCEUltra 5 µM 3.33 ± 0.48 4.76 ± 0.23 3.87 ± 0.27 10 µM 6.12 ± 0.22 9.79 ± 0.92 7.31 ± 2.84

163 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

The results for the eﬀect of iron content variation on JMJD2A inhibition both in the FDH- coupled as well as in the LANCEUltra assays are summarized in Table 4.28. Upon analysis of the trends in IC50 values of the same compound in the same assay, it becomes evident that there is some variation. Assays using less iron(II) in the assay buﬀer also yielded lower

IC50 values and experiments, in which more iron(II) was used, gave greater IC50 values. In many cases, the variation seems to follow exactly the change in iron content, i. e. doubling the iron(II) concentration also led to double the IC50 value. This would mean that the predo- minant reason for inhibition is by iron sequestration and that these structures should not be pursued further. For other examples (e. g. deferasirox 109a in the FDH-coupled assay), the change in IC50 values is not as pronounced, meaning that there may be other mechanisms at play that are independent of iron(II) content in the assay. On the other hand, these results are difficult to interpret as the enzymatic activity also depends on the iron content of the assay buffer and the uninhibited maximum enzyme activity varies between the different assays. Therefore, no mechanism of inhibition can be ruled out based on these data.

4.6.2.2. Competitivity Towards 2-Oxoglutarate

As the results from experiments varying the iron(II) content were mainly inconclusive, the dependence of inhibition of JMJD2A on the enzyme co-substrate 2-oxoglutarate (2-OG) was tested in LANCEUltra assays. This was outlined in detail for other lead structures (cf. Section 4.2.2). The results for deferoxamine 37 and deferasirox 109a are summarized in Figure 4.19. As becomes evident from the kinetic data for deferasirox 109a (cf. Figure 4.19(a)), inhibition of JMJD2A by this compound is competitive with regard to the co-substrate 2-OG as increasing concentrations of the co-substrate could overcome inhibition and all curves reached the same maximum velocity. The only diﬀerence is that they were shifted gradually to the right and yielded increasing apparent KM values for 2-OG. This is also illustrated by the double- reciprocal representation of the data set (cf. Figure 4.19(b)), highlighting competitive behavior as all lines intersect in one point on the 1/v-axis. A very diﬀerent behavior was observed for deferoxamine 37 (cf. Figure 4.19(c)). This shows that even by increasing the concentration of the co-substrate, inhibition of JMJD2A could not

164 4.6. Clinically Used Iron Chelators

(a) Inhibition of JMJD2A by deferasirox 109a (b) Lineweaver-Burk representation of data

Figure 4.19.: Inhibition of JMJD2A by clinically used iron chelator deferasirox 109a is competitive to co-substrate 2-oxoglutarate, while deferoxamine 37 and the general metal ion chelator EDTA 38 show no competitive behavior. See text for explanations. Shown are mean ± s. d. of duplicates.

be overcome and that all curves yielded diﬀerent maximum velocities vmax of the enzyme. On the other hand, the apparent binding constants KM of 2-OG to the enzyme did not change dramatically. This shows that the observed inhibition of JMJD2A by deferoxamine 37 was in fact independent of 2-OG and, therefore, very likely not due to complexation of Fe2+ in the active site, but due to other mechanisms like sequestration of iron from the assay buﬀer.

165 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

A similar behavior was also observed when the general metal ion chelator ethylenediaminetetraacetic acid (EDTA) 38 was tested for comparison. This hexadentate ligand can be expected to complex iron(II) ions in solution, but is unlikely to bind in a speciﬁc interaction in the enzyme active site. This compound exhibited the same behavior as deferoxamine 37 (cf. Figure 4.19(d)). The apparent inhibition of JMJD2A by EDTA 38 was independent of the co-substrate and increasing its concentration did not overcome the inhibitory eﬀect. This can likely be attributed to the fact that EDTA 38 complexes all Fe2+ ions in solution or removes them from the enzyme. In conclusion, out of the investigated metal ion chelating molecules, only deferasirox 109a was able to potently inhibit JMJD2A in a co-substrate-competitive manner based on these enzyme kinetic data. This suggests that it can actually bind into the active site of JMJD2A and be considered a veritable small molecule enzyme inhibitor as opposed to the other compounds who produce an apparent inhibition by ferrous ion sequestration.

4.6.2.3. EPR Spectroscopy

In order to further investigate the hypothesis that deferasirox 109a binds in the active site of JMJD2A and is, thus, a veritable small molecule inhibitor, electron paramagnetic resonance spectroscopy was employed. This method allowed for the characterization of the chemical environment of iron ions in complex with deferasirox and the enzyme. Measurements were performed and data analyzed by Kerstin Serrer (Schleicher group, Institute of Physical Chemistry, Albert-Ludwigs-University Freiburg). While further investigations are on-going, preliminary data are summarized in Figure 4.20. To this end, three diﬀerent samples were prepared and measured. One contained the enzyme

JMJD2A and a constant amount of FeSO4 in order to evaluate the effect of complexation of the iron ion by the enzyme. Another sample contained the same components and the putative inhibitor deferasirox 109a. In order to compare it to the effect if binding were to take place in solution outside of the enzyme, a third sample containing only FeSO4 and 109a in enzyme buffer was produced. In contrast to enzyme activity or kinetic experiments, these samples did not contain any ascorbic acid as reducing agent, so that a partial oxidation from Fe2+ to Fe3+ was possible. It is likely that the iron(III) species are observed in this spectrum.

166 4.6. Clinically Used Iron Chelators

Figure 4.20.: Electron paramagnetic resonance spectra of iron in complex with putative

inhibitor deferasirox 109a and JMJD2A. [FeSO4] = 100 µM in all experiments. If included, [JMJD2A] = 65 µM and [deferasirox] = 200 µM. Data

were collected using a two-pulse sequence, π/2 – τ – π (16 ns – 200 ns – 32 ns) at a microwave frequency of 33.815 GHz and at T = 4.5 K.

As becomes clear from Figure 4.20, samples containing iron complexed by the enzyme JMJD2A yielded a signal at ca. 5600 G, characteristic for the iron species bound in the protein structure. The sample, which additionally contained deferasirox 109a, gave a novel signal at ca. 9400 G. Importantly, this resonance was not observed in the sample, in which the iron-deferasirox complex is formed in solution in the absence of enzyme. This shows that there is indeed a novel interaction with an altered chemical environment around the iron species, which is only observed in the sample where enzyme and deferasirox were present. This can be taken as indirect evidence that binding of deferasirox does indeed take place within the active site of JMJD2A and not in solution as already assumed based on the kinetic data (cf. Section 4.6.2.2). Further experiments are currently on-going in order to better characterize these novel signals and to test whether the resonance at 9400 G can be attenuated in the presence of an excess of 2-oxoglutarate (2-OG), which should be possible given that deferasirox is a 2-OG-competitive inhibitor (cf. Figure 4.19).

167 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

4.6.3. Structural Binding Model

The mode of action of deferasirox 109a that is suggested by the enzyme kinetic experiments (cf. Section 4.6.2.2) and the EPR spectroscopic investigations (cf. Section 4.6.2.3) is further corroborated by a molecular docking study revealing a plausible binding mode of deferasirox in the active site of JMJD2A. Such a docking pose was generated by the Sippl group (Institute of Pharmacy, Martin-Luther- University Halle-Wittenberg) and is depicted in Figure 4.21. It shows the binding mode of deferasirox 109a in green and of another analog that was suggested in silico, which is lacking

Figure 4.21.: Proposed binding mode of deferasirox 109a (green) and a decarboxylated analog 109b (orange) in the active site of JMJD2A generated by molecular docking. The Fe2+ ion is shown as a brown sphere, oxygen atoms in red, and nitrogen atoms in blue. The shape of the active site is highlighted by a surface plot.

168 4.6. Clinically Used Iron Chelators the carboxylate moiety 109b, in orange. Both compounds adopt virtually the same binding mode in the active site and overlay perfectly. The docking pose suggests tridentate chelation of the central ferrous ion by deferasirox 109a via a nitrogen atom from the triazole ring and oxygen atoms from both phenol rings. The ring system needs to be twisted out of a planar conformation in order to adopt this binding pose. Importantly, the result from docking suggests that deferasirox can readily be accommodated in the active site of JMJD2A without any steric clashes with other residues. However, it also does not form any speciﬁc interactions with any residue, which would further improve its binding potency. As such, the observed micromolar potency of this compound is remarkable. The phenyl ring of deferasirox, which is not involved in metal chelation, points away from the catalytic site into the peptide substrate binding pocket, while the terminal carboxylate moiety protrudes into the tunnel, through which the peptide enters into the active site. However, the COOH group also does not undergo any relevant interactions with any active site residues. Based on this docking pose, several structural analogs could be envisaged, which may improve the in vitro potency of deferasirox 109a. They were prepared by chemical synthesis (cf. Section 4.6.5) and these derivatives tested in the established JMJD2A assays (cf. Section 4.6.6). Impor- tantly, this image suggests that additional interactions may be possible by substitutions on the phenol rings, in particular with the highly conserved active site residue lysine K206. Moreover, this binding pose suggests that the carboxylate moiety is not necessary for potent inhibition as the decarboxylated analog adopts precisely the same binding position. This was validated by synthesis of such an analog 109b.

4.6.4. Selectivity

As the last step in the in vitro characterization of clinical iron chelators as novel JMJC demethylase inhibitors, their selectivity towards two other enzyme subtypes was investigated, i. e. JMJD3 (KDM6B) and JARID1A (KDM5A). The results of LANCEUltra assays for these enzymes are summarized in Table 4.29. As becomes clear from these data, neither deferoxamine 37 nor deferasirox 109a exhibited any selectivity towards any of the three demethylases tested. This was to be expected for deferoxamine 37 as it was shown to inhibit JMJD2A by mechanisms other than actual binding

169 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.29.: Selectivity of clinically used iron chelators against JMJD2A, JMJD3, and JARID1A in LANCEUltra assays. Data are mean ± s. d. of duplicates.

IC50 / µM Compound JMJD2A JMJD3 JARID1A Deferoxamine 37 3.33 ± 0.48 2.36 ± 0.23 4.62 ± 0.27 Deferasirox 109a 4.76 ± 0.23 3.95 ± 0.65 5.00 ± 0.62

into the active site. The metal sequestration taking place in solution in the JMJD2A assays can occur for the other enzymes as well. As the iron(II) content was equal in all assay buffers, it is not surprising that this compound gave virtually identical results in all assays. On the other hand, for deferasirox 109a, direct binding within the active site of JMJD2A could be shown through indirect methods and rationalized in a molecular docking procedure. However, this revealed that 109a likely only functions as a tridentate iron chelator without forming any other specific interactions with any residues in the active site of JMJD2A. As- suming a similar binding pose also for JMJD3 and JARID1A, it is comprehensible that the potency does not differ much across the different demethylases.

4.6.5. Synthesis of Deferasirox Analogs

In order to elaborate on the lead structure of deferasirox 109a and to test the proposed binding pose suggested by molecular docking (cf. Section 4.6.3), several analogs of deferasirox were synthesized. The general synthetic route is outlined in Figure 4.22. This synthesis allowed for the introduction of several modiﬁcations to the lead structure both on the phenyl ring not involved in metal binding as well as on both phenols. As the entire synthesis is only a two-step procedure, it was acceptable that the substitutions needed to be introduced through the starting materials. The synthetic route as outlined in Figure 4.22 was based on the technical synthesis[209] of deferasirox 109a and literature reports.[210] Analogs with variations on the distal phenyl ring 109a–c had already been reported in the scientiﬁc literature,[210, 211] however only as metal- complexing ligands and their pharmacological properties were not investigated. While 109d

170 4.6. Clinically Used Iron Chelators

Figure 4.22.: Synthetic route towards analogs of the lead structure deferasirox 109a.

is included in some patents, a methoxy substitution on the metal-binding phenol rings 109e is novel to the literature and this was prepared in analogy to unsubstituted deferasirox 109a.[209] The first step of the synthesis is the condensation of (un)substituted salicylamides 111 and salicylic acids 112 to the corresponding benzoxazinones 113. This could readily be achieved at elevated reflux temperature using thionyl chloride as activating agent. The products 113 precipitated at the end of the SOCl2 addition and required virtually no purification. The crystalline benzoxazinones 113, however, can be observed to slowly decompose to bis(salicylamides) in the presence of moisture as was previously observed in the literature[210] and evidenced by 1 H-NMR spectroscopic investigations in DMSO-d 6 containing traces of water. Therefore, these intermediates needed to be stored in a dry place or used for the second coupling step quickly. Benzoxazinones 113 could be coupled to the final deferasirox-like products by refluxing with phenylhydrazines 114 in a polar solvent. While the unsubstituted phenylhydrazine 114b as well as the carboxy derivative 114a were commercially available, the methyl ester 114c was prepared from the acid by acid-catalyzed esterification by stirring in an excess of methanol as reported previously.[212] The methyl ester of deferasirox 109a was an interesting target as this would allow for cellular studies assuming it may function as a prodrug of the carboxylate, which is likely too polar to pass the cell membrane. The concept of prodrugs of carboxylic acids[204] was discussed in more detail for pyrimidylpyridine inhibitors (cf. Section 4.5.6).

171 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

The yields for the formation of the triazole ring reached from 18% to 58%, in large parts depending on the required purification method. While deferasirox 109a and some analogs readily precipitated at the end of the reaction sequence or could be purified by re-crystallization, methyl ester 109c needed to be purified by column chromatography.

4.6.6. Elaboration of the Deferasirox Lead Structure

The novel derivatives of deferasirox 109a–e that were synthesized (cf. Section 4.6.5) were also tested for their in vitro inhibitory potential on JMJD2A. The results from both the FDH- coupled as well as LANCEUltra assays are summarized in Table 4.30. These inhibition data show that all derivatives of the general lead structure 109 potently inhibited JMJD2A with very little variation in their IC50 values. The fact that all derivatives were nearly equally potent inhibitors is in line with the docking structure that was proposed for binding of deferasirox 109a in the active site of JMJD2A (cf. Figure 4.21). This suggested that the carboxylic acid in 109a was not essential to the binding. Consequently, a derivative without the acid 109b or with an ester prodrug 109c inhibited JMJD2A with equal potency.

Table 4.30.: In vitro characterization of deferasirox analogs as inhibitors of JMJD2A. Data are mean ± s. d. of duplicates.

FDH-coupled assay LANCEUltra assay R1 R2 IC50 / µM IC50 / µM 109a COOH H 7.37 ± 1.58 4.76 ± 0.23 109b HH 6.17 ± 1.43 5.27 ± 1.00

109c COOCH3 H 7.77 ± 1.32 4.12 ± 0.13 109d COOH Cl 14.0 ± 3.1 5.12 ± 0.05

109e COOH OCH3 5.51 ± 0.37 3.66 ± 0.11

172 4.6. Clinically Used Iron Chelators

The proposed binding mode of 109a in JMJD2A suggested that there may be a possibility for further interactions of the lead structure with conserved active site residues like lysine K206 through substitutions on the phenol rings. For synthetic reasons, it was only possible to synthesize derivatives with symmetrical substitutions using the simple two-step procedure (cf. Section 4.6.5). The chloro derivative 109d, in which only a lipophilic residue was attached, showed somewhat reduced potency compared to 109a, while the methoxy analog 109e was slightly more potent than 109a. This may be the case because it is able to engage in a hydrogen bond from positively charged lysine residue K206 to the oxygen atom of the methoxy group. However, it should be noted that the diﬀerences in potency are indeed very small.

4.6.7. Cellular Eﬀects of Deferasirox

Encouraged by the remarkable in vitro potency of deferasirox 109a and its derivatives and the discovery that they were in fact veritable small molecule inhibitors of JMJC demethylases, it was attempted to elucidate their biological eﬀects in cell culture studies.

4.6.7.1. Eﬀects on Cell Proliferation

Firstly, the eﬀect of deferasirox 109a and its analogs on cell proliferation was tested in two well-established cancer cell lines. The esophageal cancer cell line KYSE-150[193] was used because of its reported sensitivity to JMJC demethylase inhibition.[124] For comparison, the leukemia cell line HL-60 was also tested. Cell culture experiments were performed by Dr. Inga Hoﬀmann assisted by Karin Schmidtkunz (Jung group) and the experimental procedures and data in detail have been reported in her thesis.[78] To illustrate the potency of deferasirox- based inhibitors, their GI50 values are given in Table 4.31. This represents the concentrations, at which cell proliferation was reduced by 50%. Remarkably, all compounds were able to potently inhibit cancer cell growth at very low concentrations in the single-digit micromolar range or less. This is relevant as deferasirox 109a is a licensed drug on the market for the treatment of an entirely unrelated condition (iron overload or hemochromatosis) and was revealed by this experiment to be a lead structure for potent anti-cancer treatment. In general, KYSE-150 cells were more sensitive to treatment with deferasirox 109a and its analogs than HL-60 cells.

173 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

Table 4.31.: Cellular potency of deferasirox-based inhibitors on KYSE-150 esophageal cancer and HL-60 leukemia cells.

KYSE-150 cells[193] HL-60 cells GI50 / µM GI50 / µM 109a 3.27 ± 0.62 5.5 ± 0.4 109b 0.45 ± 0.06 1.7 ± 0.3 109c 0.46 ± 0.06 1.7 ± 0.2 109d 0.85 ± 0.21 5.7 ± 1.3 109e 2.24 ± 0.06 7.4 ± 6.8

As deferasirox 109a is usually used to chelate iron ions in the bloodstream, it does not need to be cell-permeable, but the focus is on aqueous solubility for the high doses typically administered. However, for inhibition of JMJC demethylases, these compounds need to be able to penetrate the cancer cell. As was revealed by the structure-activity relationship (SAR) of derivatives (cf. Table 4.30) and predicted by molecular docking (cf. Figure 4.21), the carboxylic acid is not required for potent inhibition. Importantly, derivatives, which do not contain this charged structural moiety, but are unsubstituted like 109b or have the carboxylate masked as a methyl ester like 109c, were about 7-fold more potent in inhibiting cancer cell proliferation. This can likely be attributed to their increased cell permeability and, thus, the higher effective concentration within the cell. In line with this observation, the substituted derivative with the lipophilic chloro substituent 109d also exhibited somewhat improved potency on KYSE-150 cells despite its carboxylate function, while the potency of the more polar methoxy derivative 109e was more similar to the unsubstituted deferasirox 109a. Within the cell, methyl ester 109c may be cleaved by unspecific esterases to the free carboxylate 109a. However, this is not relevant as both species were shown to exhibit the same in vitro potency (cf. Table 4.30). The difference in in vivo effects, thus, can arise only through higher effective on-target concentrations. For HL-60 leukemia cells, the trend regarding the in vivo potency is similar to that for KYSE- 150 cells although these cells were generally less susceptible to growth inhibition by these

174 4.6. Clinically Used Iron Chelators

Figure 4.23.: Cellular potency of deferasirox 109a and its methyl ester 109c on a variety of esophageal cancer cell lines and healthy esophagus epithelial Het-1A cells. Shown are mean ± s. d. of three independent experiments.

compounds. The most potent derivatives were, once again, the uncharged and more lipophilic analogs 109b and 109c, likely because of their improved cell permeability. Encouraged by the remarkable growth inhibitory potency of deferasirox 109a and its more cell-permeable ester analog 109c when tested on KYSE-150 esophageal cancer cells, the action of these two compounds was evaluated on a larger panel of esophageal cancer cell lines. These tests were performed by Dr. Theresa Ahrens (Laßmann group, Institute of Clinical Pathology, University Medical Center Freiburg) and the results summarized graphically in Figure 4.23. Notably, all esophageal cancer cells were sensitive to treatment with deferasirox 109a and even more so with its methyl ester 109c. For 109a, GI50 values ranged from 3.1 µM to 23 µM, but were generally in the single-digit micromolar range. The general trend that the methyl ester was signiﬁcantly more potent was observed for all cell lines, with improvements ranging from a factor of 6-fold for KYSE-410 cells to nearly 40-fold for OE19 cells. As for KYSE-150

175 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases and HL-60 cells, this can likely be attributed to the improved cell permeability of the methyl ester. For many cell lines, the GI50 value of 109c is well below 1 µM, making this compound a remarkably potent anti-cancer cytotoxic agent. It should be noted, however, that healthy epithelial cells Het-1A, which were used as a control, were also negatively aﬀected by these two compounds with little diﬀerence compared to the cancer cell lines.

4.6.7.2. On-Target Eﬀects

Deferasirox-based inhibitors of JMJC demethylases exhibited remarkable in vitro potency and potent inhibition of cell proliferation in preliminary experiments using a variety of esophageal cancer cell lines as well as leukemia cells. In order to investigate whether this was a specific on-target effect caused by JMJC demethylase inhibition or only general toxicity possibly by off-target inhibition, further experiments were performed investigating the histone methylation levels of treated cells. These experiments were performed by Dr. Inga Hoffmann and the methodology and results outlined in detail in her thesis.[78] Representative examples of immunofluorescence microscopy images are summarized in Figures 4.24 and 4.25. For these experiments, KYSE-150 cells were treated with the test compounds or DMSO as control for 72 hours, fixed, and the histone methylation level at H3K9me3 evaluated by staining with a specific primary antibody and visualized using a fluorescence-labeled secondary antibody (right panels). In order to identify cell nuclei, cells were also stained with DAPI (left panels). The control experiment, in which cells were treated with DMSO only (cf. Figure 4.24(a)), showed only very faint fluorescence when stained with the anti-H3K9me3 antibody. This represents the basal H3K9me3 level present in all cells. When cells were treated with deferasirox 109a (cf. Figure 4.24(b)), fluorescence dramatically increased due to inhibition of demethylases like JMJD2A, resulting in hypermethylation in the cell. Likely, the observed effect is the result of inhibition of several H3K9 demethylases as deferasirox 109a was shown not to be selective even across subfamilies (cf. Section 4.6.4). Moreover, the cell number that was observable in the plate was also reduced compared to DMSO treatment coinciding with the inhibition of cell proliferation (cf. Table 4.31).

176 4.6. Clinically Used Iron Chelators

(a) DMSO Control

(b) Deferasirox 109a 60 µM

(d) 109c 10 µM

Figure 4.24.: Eﬀect of deferasirox-based inhibitors on histone methylation in KYSE-150 cells as judged by immunoﬂuorescence microscopy. Left panels: DAPI

staining to identify cell nuclei; right panels: anti-H3K9me3 antibody staining. Reproduced with permission from Ref. 78.

177 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

(a) 109d 10 µM

(b) 109e 5 µM

Figure 4.25.: Eﬀect of deferasirox-based inhibitors on histone methylation in KYSE- 150 cells as judged by immunoﬂuorescence microscopy (cont’d.). Left

panels: DAPI staining to identify cell nuclei; right panels: anti-H3K9me3 antibody staining. Reproduced with permission from Ref. 78.

The same increase in trimethylation at the H3K9 mark was also observed when cells were treated with the less polar deferasirox analogs 109b and 109c (cf. Figures 4.24(c) and 4.24(d), respectively). Importantly, for these compounds, a much smaller concentration of test compound was required to reach the same hypermethylation effect, correlating with the assumed improved cell permeability and increased inhibition of cell proliferation. When these experiments were performed in a dose-dependent manner, an effective concentration of 41 µM was required for 109a to reach 50% of the maximum increase in fluorescence. Conversely, for 109b, only 3.3 µM was required as expected based on its improved in vivo potency.[78] This means that when KYSE-150 cells are treated with a low-micromolar dose of these compounds, a dramatic global hypermethylation occurs and these inhibitors can, therefore, be considered bona fide epigenetic regulators.

178 4.6. Clinically Used Iron Chelators

The same eﬀect could also be observed for the related substituted derivatives 109d and 109e as is depicted in Figure 4.25. This shows that deferasirox 109a and all its derivatives are indeed on-target JMJC demethylase inhibitors and that the observed reduction in cell proliferation is not simply a result of general cytotoxicity.

4.6.8. Discussion

The data presented in this section revealed that the clinically used iron chelator deferasirox 109a and novel derivatives based on its lead structure were potent inhibitors of JumonjiC domain-containing histone demethylases both in vitro as well as in cell culture studies. They were able to reduce cell proliferation of a number of cancer cell lines with outstanding in vivo potency. Moreover, it could be shown that due to JMJC demethylase inhibition, hypermethylation at the H3K9me3 mark occurred, making these compounds veritable epigenetic regulators. The fact that such molecules have already been approved for human use in an unrelated disorder opens the possibility to develop these derivatives further into potent anticancer medications with limited concerns about safety issues or side effects. In fact, there have been sporadic observations of the effect of deferasirox 109a on cancer cells in the literature. Due to lack of understanding of their mechanism of action, they remained mainly descriptive. In 2013, Ford et al. described that iron chelators like deferoxamine 37 and deferasirox 109a inhibited proliferation of esophageal cancer cell lines OE33, OE21, and OE19, both in cell culture studies as well as a xenograft model in mice.[213] Moreover, treatment of cells with deferasirox 109a made them more susceptible to treatment with standard chemotherapeutic agents like fluorouracil and cisplatin and managed to overcome cisplatin resistance. This study discussed that the mode of action of iron chelators in anticancer efficacy was elusive and assumed it was related to iron depletion of the cells.[213] In a different study using breast cancer cells, treatment with deferoxamine 37 led to changes in the epigenetic landscape. However, this was due to downregulation of expression of LSD1 and JMJC demethylases like JMJD2A, which resulted in a slight increase in H3K9 di- and trimethylation. Treatment induced apoptosis in these cells and enhanced their sensitivity to cisplatin and doxorubicin.[214]

179 Chapter 4. Novel Inhibitors of JMJC Histone Demethylases

The antiproliferative effect of deferasirox 109a on murine leukemia cell lines has also been described. Importantly, this effect was independent of the presence of exogenously supplied iron ions, suggesting that other mechanisms would have been at play. In mouse experiments, tumor size was dramatically reduced for animals treated with deferasirox and their survival prolonged, while treatment with 37 shortened survival compared to the control group.[215] Similarly, treatment with deferasirox 109a was also shown to be effective in inhibiting cell proliferation in a panel of gastric cancer cell lines. While there was no mechanistic explanation, it was found that treatment interfered with cell cycle regulation and induced apoptosis.[216] In terms of human use, a recent study found that in patients with myelodysplastic syndromes frequently receiving blood transfusions, their mortality was significantly reduced when they were additionally treated with deferasirox 109a.[217] Taking together these observations, the use of iron chelators and in particular 109a has been proposed as a novel avenue in anti-tumor treatment either alone[218] or in combination treat- ments.[219] In particular, its good side effect profile in clinical practice for treatment of iron overload and the fact that a number of pharmacological and safety data are already available make this compound an attractive starting point to develop further in clinical trials as anticancer agents.[220] While these biological data are of great value in demonstrating the potent anti-cancer efficacy of deferasirox 109a and the survival benefit in mouse models and human studies, they lack a mechanistic explanation. The data generated in this thesis revealed 109a to be a potent inhibitor of JMJC demethylases and to modify histone methylation patterns in cells. Moreover, novel derivatives with improved cell permeability and, thus, greater in vivo efficacy were developed. This suggests that the observed antiproliferative effects are related to changes in the epigenetic landscape caused by deferasirox-based inhibitors. Future investigations will need to dissect more precisely, to which extent JMJC demethylase inhibition and the consequently altered regulation of gene expression are the cause or an effect of the anti-tumor activity. This new insight into the mode of action of deferasirox 109a will drive the development of highly sought after novel potent anti-cancer drugs.

180 5. Outlook

For many forms of cancer, there is still an unmet clinical need for novel efficacious forms of treatment. While epigenetic mechanisms and, in particular, histone demethylases have been recognized as vital drivers of such diseases, highly potent and selective inhibitors remain largely elusive. This work provides several novel starting points for the development of inhibitors of JumonjiC domain-containing histone demethylases (JMJC demethylases). In this thesis, a screening platform has been established for the discovery of inhibitors of JMJC demethylases using different orthogonal in vitro assays. While some were adapted from the literature or commercially available, for other enzyme subtypes, these were novel assays. In the future, this screening platform should be extended to include even more enzymes of other JMJC demethylase subfamilies to better characterize the selectivity of the different inhibitors discovered in this project. Future research will focus on the optimization of these lead structures, in particular with regard to their physicochemical properties in order to improve on their pharmacokinetic characteristics to obtain better in vivo applicability. This is especially true for the class of aminopyrimidylpyridines (cf. Section 4.5), for which only a narrow window was observed between solubility of these compounds and cell permeability. Moreover, while for some lead structure classes like hydroxamic acid-based inhibitors (cf. Section 4.4), the structure-activity relationship (SAR) has been investigated extensively, for others, only partial SARs were obtained as a proof-of-principle study. This applies for example to tropolone-based inhibitors (cf. Section 4.3) and deferasirox-based inhibitors (cf. Section 4.6). However, this project has laid the groundwork for more exhaustive investigations in the future into possible routes of optimization of these compounds. With regard to tropolones, a number of natural products containing this structural motif are known and it may be worthwhile to investigate their effects. For deferasirox as well as aminopyrimidylpyridines, synthetic platforms towards analogs and prodrugs have been established in

181 Chapter 5. Outlook this project so that more derivatives can readily be generated and tested to further investigate this structure space. While some inhibitors discovered in this thesis such as the fragment-like tetrazolyl hydrazide- based structures (cf. Section 4.2) rather serve as in vitro reference compounds, for many other structures, biological effects could already be shown or can likely be obtained. Future research must, therefore, focus on the functional characterization of biological effects of JMJC demethylase inhibitor treatment of different cell lines or potentially even animals. Once selective histone demethylase inhibition can be linked to clear biological effects in animal models of disease states, the lead structures discovered herein will become valuable tool compounds for further elucidation of the biological role of these enzymes and as therapeutic candidates.

182 6. Conclusion

Iron(II)- and 2-oxoglutarate-dependent JumonjiC domain-containing histone demethylases have been recognized as vital players in epigenetic gene regulation. While aberrant expression of these enzymes has been linked to a number of diseases, in particular cancer, their precise physiological role is not yet fully understood. In this project, several in vitro screening assays for representative enzyme subtypes JMJD2A (KDM4A), JARID1A (KDM5A), and JMJD3 (KDM6B) were established and validated using reference inhibitors. Using this in vitro assay platform, several lead discovery campaigns were initiated employing a variety of methodologies such as virtual screening, fragment-based drug discovery, focused library screening, and rational design. A total of ﬁve diﬀerent lead structures for potent inhibitors of JMJC demethylases could be discovered. For all lead compounds, analogs were either generated using newly established synthetic platforms or acquired commercially or from collaboration partners allowing for iterative optimization of these compounds in feedback with biological testing. This has allowed for the development of a remarkably KDM4A-selective fragment-like in vitro inhibitor 44k and more elaborate compounds like aminopyrimidylpyridine-based inhibitors

86 with Ki values in the nanomolar range with dual selectivity for the KDM4 and KDM5 subfamilies. For these, a prodrug strategy could be devised allowing for cell-permeable derivatives with antiproliferative activity on esophageal cancer cells. Moreover, a known inhibitor of zinc-dependent histone deacetylases 57a could be modified structurally improving its in vitro potency on JMJC demethylases and culminating in the development of a series of selective derivatives with no HDAC inhibition, whose prodrugs exhibited antileukemic activity. Lastly, the clinically used iron chelator deferasirox 109a was discovered to also inhibit JMJC demethylases and cell-permeable derivatives could specifically alter histone methylation in treated cells while also exhibiting highly potent antiproliferative efficacy on a panel of esophageal cancer cells, suggesting the repurposing of this molecule and its analogs as an anticancer medication.

183 Chapter 6. Conclusion

In conclusion, the establishment of an in vitro assay platform for three subtypes of JMJC demethylases will allow for further screening for novel lead structures. In addition, the vast collection of highly diverse novel inhibitors discovered in the present study, some with subfamily selectivity, some with pan-JMJC activity, some exclusively for in vitro applications, some even suitable for in vivo studies, lays the foundation for further biological experiments to elucidate the physiological role of these enzymes and a potential advancement into pre-clinical studies for anticancer treatment.

184 7. Experimental Section

7.1. Biochemical Assays

7.1.1. General Remarks

All reagents, chemicals, and buffer components were obtained from commercial sources and used without further purification. Test compounds were dissolved in DMSO to a concentration of 20, 10, or 5 mM according to their solubility and these stock solutions stored in a freezer at –20◦C. Before use, these stock solutions were thawed, allowed to equilibrate to room temperature, and thoroughly homogenized on a vortex shaker. Water for buffers and solutions was double distilled to Milli-Q purity. Standard laboratory glassware and plastic apparatus was used, in particular micro-reaction vials (“Eppendorf tubes”) in 1.5 mL (Sarstedt) and 5 mL (VWR) size as well as pipette tips in 10 µL and 200 µL size (Sarstedt) and 1000 µL size (ULPlast Omnitip). All dilution and transfer steps were pipetted with appropriate Abimed/Kinesis HTL Discovery Comfort single-channel pipettes (ranges: 0.5–10 µL, 10–100 µL, 20–200 µL, 100–1000 µL), which were regularly calibrated. Final dispension steps of detection solutions were performed using an automated Brand HandyStep electronic dispenser pipette with appropriate PD tips. For fluorescence-based assays, white opaque OptiPlate-384 microtiter plates from PerkinElmer (catalog #: 6007299) with Lid-384 clear non-sterile lids (catalog #: 6007617) were used. For absorbance-based assays, clear SpectraPlate-96 MB microtiter plates from PerkinElmer (catalog #: 6005649) with Lid-96 clear non-sterile lids (catalog #: 6005617) were used. For the last step of the LANCEUltra assays, plates were sealed with airtight PerkinElmer TopSeal- A 384 (catalog #: 6005250) or TopSeal-A PLUS (catalog #: 6050185) adhesive sealing tapes instead of lids.

7.1.1.1. Reagents and Chemicals

2-[4-(2-Hydroxyethyl)piperazin-1-yl] ethanesulfonic acid (HEPES) for buffers was obtained from VWR (catalog #: 441485H). The LANCEUltra assay uses a proprietary Detection Buffer in its final step (delivered at 10X concentration, Perkin Elmer, catalog #: CR97), which was freshly diluted with Milli-Q water

185 Chapter 7. Experimental Section prior to use and stored on ice. The detection mix contains the proprietary dye ULight coupled to streptavidin, which was also obtained from PerkinElmer (catalog #: TRF0102). Ferrous sulfate heptahydrate (VWR, Normapur purity, catalog #: 24244.232), l(+)-ascorbic acid (AppliChem, BioChimica purity, catalog #: A1052), and α-ketoglutarate disodium salt dihydrate (AppliChem, BioChmica purity, catalog #: A6408) as components for enzyme inhibition assays were equally obtained from commercial sources. Aqueous solutions at a concentration of 10 mM were prepared in 50 mL conical tubes and diluted with Milli-Q water to the working concentrations of 1 mM, 3 mM, and 10 or 0.1 mM, respectively. These stock solutions were stored at room temperature and replaced with freshly prepared solutions at least once a week to avoid oxidation or other degradation of the reagents. The oxidized form of nicotinamide adenine dinucleotide (NAD+) was from different commercial sources at different times including Sigma-Aldrich, AppliChem, and Carl Roth. In any case, stock solutions were regularly prepared fresh by dissolving a defined amount in Milli-Q water to a concentration of 80 mM. These stocks were aliquoted, shock-frozen in liquid nitrogen, and stored in a freezer at –20◦C. Ethylenediaminetetraacetic acid (EDTA) disodium salt dihydrate and 37% aqueous formaldehyde (HCHO) solution were from bulk chemical stocks as supplied from the chemical store and used without further purification. Ferrozine (5,6-diphenyl-3-(2-pyridyl)-1,2,4-triazine-4,4”-disulfonic acid monosodium salt hydrate) was obtained from Alfa Aesar (catalog #: B24066). BSA for the LANCEUltra assay buffer was from PAA Laboratories (Fraction V, catalog #: K45-001). Full-length histone proteins for enzyme activity assays were obtained as histone from calf thymus (type II-A, Sigma-Aldrich, catalog #: H9250).

186 7.1. Biochemical Assays

7.1.1.2. Peptides

Peptide Sequence Sourcea Catalog #

H3K9me3 1–15 ARTKQTARK(me3)STGGKA CC 10530

H3K9me3 7–14 ARK(me3)STGGK-NH2 PSL custom synthesis

H3K4me3K9me3 1–24 ARTK(me3)QTARK(me3)STGG- PSL custom synthesis

KAPRKQLATKA-NH2

biot-H3K4me2 1–21 ARTK(me2)QTARKSTGG- AS 64356 KAPRKQLA-GGK(biotin)

biot-H3K4me3 1–21 ARTK(me3)QTARKSTGG- AS 64192 KAPRKQLA-GGK(biotin)

biot-H3K9me2 1–21 ARTKQTARK(me2)STGG- BPS 50351 KAPRKQLA-GGK(biotin)

biot-H3K9me3 1–21 ARTKQTARK(me3)STGG- BPS 50350 KAPRKQLA-GGK(biotin)

biot-H3K27me2 21–44 ATKAARK(me2)SAPATGGVK- AS 64366 KPHRYRPG-GK(biotin)

biot-H3K27me3 21–44 ATKAARK(me3)SAPATGGVK- PSL custom synthesis KPHRYRPG-GK(biotinNovaTag) a: CC – Cayman Chemical, Ann Arbor, MI; PSL – Peptide Specialty Laboratories, Heidelberg; AS – Eurogentec AnaSpec, Fremont, CA; BPS – BPS BioScience, San Diego, CA.

Peptides from BPS were obtained as 40 µM aqueous solutions, aliquoted, and stored in a freezer at –80◦C. All other peptides were obtained as lyophilized powders and reconstituted in Milli-Q water according to the declared amount and molecular weight supplied by the manufacturer to an appropriate concentration (e. g. 10 mM or 1 mM), shock-frozen in liquid nitrogen, and stored in a freezer at –20◦C or –80◦C.

7.1.1.3. Antibodies

Antibody Source Catalog #

Eu-labeled α-H3K4me1/2 PerkinElmer TRF0402 Eu-labeled α-H3K9me2 PerkinElmer TRF0403

Eu-labeled α-H3K27me2 PerkinElmer TRF0406

Mouse α-H3K9me2 abcam ab1220 Eu-labeled α-Mouse IgG PerkinElmer AD0124

187 Chapter 7. Experimental Section

Antibodies were stored either in a refrigerator or freezer according to the manufacturer’s instructions and used at the declared concentration.

7.1.1.4. Buﬀers

The optimized buﬀer compositions used for the diﬀerent assays were as follows.

FDH-coupled assay: 50 mM HEPES, pH 7.50, 0.01% (v/v) Tween-20. FDH counterscreen assay: 50 mM HEPES, pH 7.50, 0.01% (v/v) Tween-20. LANCEUltra assay: 50 mM HEPES, pH 7.50, 0.01% (v/v) Tween-20, 0.01% (m/m) BSA. LANCE Detection Buﬀer 10X: Unknown. From PerkinElmer (catalog #: CR97). Ferrozine assay: 25 mM sodium acetate, 25 mM acetic acid, pH 4.50.

The standard HEPES buffer was regularly prepared fresh and stored at room temperature. During the preparation procedure, it was filtered through 0.2 µm sterile cellulose acetate syringe-tip filters (VWR, catalog #: 514-0061) in order to sterilize it. Bovine serum albumin (BSA) was obtained from PAA, dissolved in water to 0.05 mg/µL, aliquoted and shock-frozen in liquid nitrogen, stored in a freezer at –20◦C, and added to the standard HEPES buffer on the day of use as required to reach the final concentration of 0.01% (m/m), i. e. 0.10 mg/mL.

7.1.1.5. Machines and Software

The following measurement devices were used to make optical measurements of absorbance, ﬂuorescence intensity, or time-resolved ﬂuorescence.

“BMG” Plate Reader BMG Labtech POLARstar Optima microplate reader “EnVision” Plate Reader PerkinElmer EnVision 2102 multilabel plate reader UV/Vis Spectrometer PerkinElmer Lambda 25 UV/VIS spectrometer Software for dose-response curves GraphPad Prism 4.00 (GraphPad Software, Inc.)

7.1.1.6. Enzymes

The primary target of this thesis work, the histone demethylase JMJD2A (KDM4A) was obtained from Dr. Henriette Franz (research group of Roland Sch¨ule,Central Clinical Research, University Medical Center, Freiburg i. Br.). Using standard protein expression and puriﬁcation methods, the catalytic domain of this enzyme (i. e. residues 1–359) was expressed in E. coli

188 7.1. Biochemical Assays bacteria and purified via its His-Tag on Nickel chelate beads. The approximate protein concentration of these enzyme batches was determined by a simple Bradford assay. The expression method followed that previously published by Ng et al.[64] Shortly, the plasmid pNIC28-Bsa4 JMJD2A encoding human KDM4A residues 1–359 was transformed in BL21-CodonPlus-Ril competent cells. Six liters of TB media containing kana- mycin (50 µg · mL−1) and chloramphenicol (34 µg · mL−1) were inoculated with a 15 mL · L−1 overnight culture and grown at 37◦C. Expression was induced by addition of 0.2 mM IPTG ◦ at A260 = 0.6. Then the culture was incubated at 18 C for another 18 h. After harvesting and lysis of the bacteria, the protein was purified by a Talon bead column. The purity of JMJD2A estimated by sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE) was above 90%. As a control for continually equal enzyme quality, every new batch of this enzyme (approx. two to three per year) was tested in activity assays (FDH-coupled kinetic measurements and LANCEUltra) to verify that the delivered fractions were active enzyme (cf. Section 7.1.2.2). It was verified that every batch, when diluted to the same nominal working concentration, showed comparable activity to previous batches or the fractions were rejected. Furthermore, every new batch of enzyme was tested in an FDH-coupled inhibition assay against the reference inhibitor 2,4-PDCA, which should yield an IC50 value of around 1 µM.

FDH was obtained commercially and is sold by activity units rather than weight so as to ensure equal activity of the enzyme even though the actual amount and purity may vary from batch to batch. Upon arrival, this enzyme was reconstituted in Milli-Q water to a concentration of 0.1 Unit/µL (based on the declared activity of the particular batch), aliquoted and shock-frozen in liquid nitrogen, and stored in a freezer at –20◦C.

Other enzymes that were obtained from commercial sources are summarized in the table.

Enzyme Systematic Name Tag Residues Sourcea Catalog #

JMJD2A KDM4A His6 1–350 CC 10336 JMJD2C KDM4C Strep II 2–372 CC 10776 JARID1A KDM5A FLAG 1–1090 BPS 50110 JMJD3 KDM6B FLAG 1043–end BPS 50115 FDH – full-length SA F1879 a: CC – Cayman Chemical, Ann Arbor, MI; BPS – BPS BioScience, San Diego, CA; SA – Sigma-Aldrich, St. Louis, MO.

189 Chapter 7. Experimental Section

All demethylases were aliquoted and stored in a freezer at –80◦C. Before use, they were slowly thawed on ice and gently homogenized on a vortex shaker. After use, aliquots were sometimes shock-frozen in liquid nitrogen again and stored in a freezer at –80◦C, but never more than three freeze-thaw cycles before a dramatic loss in enzymatic activity was observed.

7.1.2. FDH-Coupled Enzyme Activity Assays

7.1.2.1. Inhibition Curves

Unless otherwise noted, a typical experiment to test the inhibitory potency of a given compound in the FDH-coupled assay was performed as follows. The FDH enzyme-coupled demethylase activity assay was performed in a total volume of 20 µL on white OptiPlate-384 microtiter plates (PerkinElmer) using a 50 mM HEPES buﬀer at pH = 7.50 containing 0.01% (v/v) Tween-20. The enzyme JMJD2A (KDM4A) 1–359 was pre-diluted to the working concentration of 0.10 mg/mL (2.40 µM), which results in a ﬁnal assay concentration of 1.75 µM. A 13-point dilution series (0. . . 20 mM) of the test compound in DMSO was prepared (typically in 1:2 or 1:3 steps). A total amount of 100 or 200 µL of a substrate solution was prepared as required according to the following table.

assay conc sub soln conc stock solution for 100 µL for 200 µL ascorbic acid 100 µM 400 µM 3 mM 13.3 µL 26.7 µL

FeSO4 10 µM 40 µM 1 mM 4.0 µL 8.0 µL FDH 0.001 Unit/µL 0.004 Unit/µL 0.1 Unit/µL 4.0 µL 8.0 µL NAD+ 500 µM 2000 µM 80 mM 2.5 µL 5.0 µL 2-oxoglutarate 50 µM 200 µM 10 mM 2.0 µL 4.0 µL

H3K9me3 7–14 35 µM 140 µM 1 mM 14.0 µL 28.0 µL buﬀer 60.2 µL 120.3 µL

The assay mixtures for every well were prepared in 1.5 mL reaction tubes according to the following pipetting scheme. The ﬁnal DMSO concentration was 2% in all wells.

190 7.1. Biochemical Assays

Blank Pos Control Neg Control Sample

enzyme (cﬁnal × 1.37) – 17.5 µL 17.5 µL 17.5 µL buﬀer 17.5 µL – 6.0 µL–

compound (cﬁnal × 48) – – – 0.5 µL DMSO 0.5 µL 0.5 µL 0.5 µL–

sub soln (cﬁnal × 4) 6.0 µL 6.0 µL – 6.0 µL TOTAL 24.0 µL 24.0 µL 24.0 µL 24.0 µL

After dispensing the enzyme and buﬀer into the appropriate tubes and addition of compound pre-dilution or DMSO, this mixture was pre-incubated at room temperature for 10 minutes. Then, the substrate solution was pipetted into all the corresponding tubes and 20 µL of the mixture immediately transferred into the wells of a 384-well plate. FI of the forming product

NADH was measured at λex = 330 nm and λem = 460 nm on a POLARstar Optima microplate reader (BMG Labtech) immediately after addition (t = 0) and after one hour of incubation on a horizontal shaker at 37◦C. The values for all wells were blank-corrected and the difference in intensity at t = 1 h and t = 0 was taken as a measurement of enzyme activity. Activity in % was then calculated from these differences in comparison to compound-free DMSO control (Pos Control, defined as 100%) and no-substrate negative control (Neg Control, defined as 0%) according to the following equation. ∆FI − ∆FI Activity = Sample Neg · 100% (7.1) ∆FIPos − ∆FINeg The activity values for all sample wells were used to construct an inhibition curve and, using GraphPad Prism 4.00, analyzed by sigmoidal curve fitting to the 4-parameter logistic curve function of the type y − y y = y + max min , (7.2) min H 1 + x EC50 where y values correspond to the percentage of activity, ymin and ymax its minimum and maximum as defined by the fit curve, and x to the concentrations used. EC50 represents the particular concentration at which half the effect is observed, in this case half the reduction in activity between ymax and ymin. H is termed the Hill slope parameter and defines the ymax−ymin slope of the curve between ymax and ymin running through the point (EC50 ; ymin + 2 ). Furthermore, it is a measure of the stoichiometry of binding of inhibitor molecules to enzyme molecules and can, thus, be expected to be around 1 in the case of direct binding of one molecule of inhibitor per enzyme molecule.

191 Chapter 7. Experimental Section

From these curve parameters, IC50 values were determined, which denote the concentration, at which enzymatic activity was reduced to 50% or, in other words, 50% inhibition took place. They were calculated from the ﬁt parameters as follows:

1 ymax − ymin H IC50 = − 1 · EC50. (7.3) 50 − ymin Every compound was tested in at least two independent experiments and the values shown correspond to their mean ± s. d.

7.1.2.2. Kinetic Activity Measurements

Kinetic activity measurements using the FDH-coupled assay were performed essentially as for the inhibition curves (cf. Section 7.1.2.1), but in a simplified setting. For quality control purposes of every new batch of enzyme (cf. Section 7.1.1.6), a sample of the new batch and the current batch of enzyme were both diluted to the working concentration of 0.10 mg/mL (2.40 µM), which results in a final assay concentration of 1.75 µM. To these samples, 0.5 µL DMSO was added and the mixture pre-incubated at room temperature for 10 minutes. Then, 6.0 µL of the substrate solution were added and 20 µL of the total mixture transferred into the wells of a 384-well plate. This was measured at the same conditions on the BMG plate reader as for the inhibition experiments, but in kinetic mode. This means that the plate reader itself was pre-heated to 37◦C and the fluorescence intensity continuously measured (e. g. every 30 seconds) over a total time of 1.5 hours. This was used to construct a fluorescence-over-time curve, which should give similar results for both the old and the new batch of enzyme. Blank and negative control measurements resulted in a flat line with no change of fluorescence over time. These assays were typically performed in triplicate.

7.1.3. FDH-Counterscreening Assay

In order to rule out inhibition of formaldehyde dehydrogenase (FDH) as a confounder in the FDH-coupled JMJD2A asasy (cf. Section 7.1.2.1), the enzyme formaldehyde dehydrogenase alone was also tested in the presence of the potential inhibitor and its substrate, formaldehyde (HCHO), added separately. A typical FDH-counterscreening assay was performed in a total volume of 20 µL on white OptiPlate-384 microtiter plates (PerkinElmer) using a 50 mM HEPES buﬀer at pH = 7.50 containing 0.01% (v/v) Tween-20.

192 7.1. Biochemical Assays

To this end, 37% aqueous formaldehyde solution (13.4 mol/L) was diluted over several steps with Milli-Q water to a concentration of 200 µM, which corresponds to a ﬁnal assay concentration of 40 µM. This is an amount of HCHO comparable to that, which would be formed in case of full conversion during the FDH-coupled activity assay. Several dilution steps of the compound solution in DMSO were prepared (e. g. 400 µM, 100 µM, and 10 µM ﬁnal concentrations). A total amount of 50 or 100 µL of a substrate solution was prepared as required according to the following table. assay conc sub soln conc stock solution for 50 µL for 100 µL ascorbic acid 100 µM 400 µM 3 mM 6.67 µL 13.33 µL

FeSO4 10 µM 40 µM 1 mM 2.00 µL 4.00 µL FDH 0.001 Unit/µL 0.004 Unit/µL 0.1 Unit/µL 2.00 µL 4.00 µL NAD+ 500 µM 2000 µM 80 mM 1.25 µL 2.50 µL 2-oxoglutarate 50 µM 200 µM 10 mM 1.00 µL 2.00 µL buﬀer 37.1 µL 74.2 µL The assay mixtures for every well were prepared in 1.5 mL reaction tubes according to the following pipetting scheme. Blank Pos Control Neg Control Sample assay buﬀer 23.5 µL 12.7 µL 18.7 µL 12.7 µL

HCHO 200 µM(cﬁnal × 5) – 4.80 µL 4.80 µL 4.80 µL

compound (cﬁnal × 48) – – – 0.5 µL DMSO 0.5 µL 0.5 µL 0.5 µL–

sub soln (cfinal × 4) – 6.0 µL – 6.0 µL TOTAL 24.0 µL 24.0 µL 24.0 µL 24.0 µL After dispensing the formaldehyde solution and buffer into the appropriate tubes and addition of compound pre-dilution or DMSO, this mixture was pre-incubated at room temperature for 10 minutes. The final DMSO concentration was 2% in all wells. Then, 6.0 µL of the substrate solution were pipetted into all the corresponding tubes and 20 µL of the mixture immediately transferred into the wells of a 384-well plate. The plates were incubated at 37◦C on a horizontal shaker for 30 minutes. Then, fluorescence intensity

(FI) of the formed product NADH was measured at λex = 330 nm and λem = 460 nm on a POLARstar Optima microplate reader (BMG Labtech). All values were blank-corrected and activity of FDH was calculated in % relative to the positive and negative control according to FI − FI Activity = Sample Neg · 100%. (7.4) FIPos − FINeg

193 Chapter 7. Experimental Section

For a test compound to be considered a veritable JMJD2A inhibitor rather than an FDH- inhibiting confounder, the relative FDH activity should not be less than 80% even at the highest test concentration. These assays were typically performed in duplicate per inhibitor and concentration.

7.1.4. LANCEUltra Assays

7.1.4.1. Inhibition Curves for JMJD2A

Unless otherwise noted, a typical experiment to test the inhibitory potency of a given compound in the LANCEUltra assay was performed as follows. The commercial antibody-based LANCEUltra demethylase activity assay (PerkinElmer) was performed in a total volume of 10 µL on white OptiPlate-384 microtiter plates using a 50 mM HEPES buffer at pH = 7.50 containing 0.01% (v/v) Tween-20 and 0.01% (m/m) BSA. The enzyme JMJD2A (KDM4A) 1–359 was first pre-diluted to a concentration of 0.10 mg/mL (2.40 µM) and then in two further steps to the working concentration of 85.7 nM, which results in a final assay concentration of 60 nM. A 13-point dilution series (0. . . 20 mM) of the test compound in DMSO was prepared (typically in 1:2 or 1:3 steps). A total amount of 100 µL of a substrate solution was prepared according to the following table.

assay conc sub soln conc stock solution for 100 µL

FeSO4 5 µM 20 µM 1 mM 2.0 µL ascorbic acid 100 µM 400 µM 3 mM 13.3 µL 2-oxoglutarate 1 µM 4 µM 0.1 mM 4.0 µL

biot-H3K9me3 1–21 400 nM 1600 nM 40 µM 4.0 µL buﬀer 76.7 µL

The assay mixtures for every well were prepared directly in the wells of the 384-well plate according to the following pipetting scheme.

Blank Pos Control Neg Control Sample

enzyme (cﬁnal × 10/7) – 7.0 µL – 7.0 µL buﬀer 9.5 µL – 7.0 µL–

compound (cﬁnal × 20) – – – 0.5 µL DMSO 0.5 µL 0.5 µL 0.5 µL–

sub soln (cﬁnal × 4) – 2.5 µL 2.5 µL 2.5 µL TOTAL 10.0 µL 10.0 µL 10.0 µL 10.0 µL

194 7.1. Biochemical Assays

After dispensing the enzyme and buﬀer into the appropriate wells and addition of compound pre-dilution or DMSO, the plates were pre-incubated at room temperature for 10 minutes. The ﬁnal DMSO concentration was 5% in all wells. Then, the substrate solution was pipetted into all the corresponding wells, the plate gently agitated to ensure mixing, and incubated at room temperature on a horizontal shaker for 45 minutes. Shortly before the end of this incubation period, the dectection mix was freshly prepared according to the following table.

assay conc detect mix conc stock solution for 400 µL

Eu-labeled α-H3K9me2 Ab 2 nM 4 nM 0.625 µM 2.56 µL ULight-Streptavidin 50 nM 100 nM 10 µM 4.00 µL EDTA 1 mM 2 mM 10 mM 80.0 µL LANCE detection buﬀer 1X 313.4 µL

Once the first incubation period, i. e. the catalytic transformation catalyzed by JMJD2A, was over, 10 µL of detection mix were added to all wells using an electronic dispenser pipette. The detection mix contains a large concentration of EDTA as a general metal ion chelator to inhibit any further enzymatic reaction while antibody binding takes place. After addition, the plate was gently agitated to ensure sufficient mixing and incubated for another 60 minutes at room temperature on a horizontal shaker. Care must be taken as to the ‘room temperature’ as elevated temperatures have detrimental effects on antibody binding and lead to uninterpretable results. This is likely due to denaturing of the antibody above ∼25◦C. Then, time-resolved FRET intensity was measured on a PerkinElmer EnVision 2102 multilabel plate reader at λex = 340 nm and λem = 665 nm with a delay of 100 µs. The values for all wells were blank-corrected and activity in % is in comparison to compound- free DMSO control (Pos Control, defined as 100%) and no-enzyme negative control (Neg Con- trol, defined as 0%) according to the following equation.

LANCE − LANCE Activity = Sample Neg · 100% (7.5) LANCEPos − LANCENeg The activity values for all sample wells were used to construct an inhibition curve and, using GraphPad Prism 4.00, analyzed by sigmoidal curve ﬁtting to the 4-parameter logistic curve function (cf. eqn. 7.2) and IC50 values calculated as outlined in Section 7.1.2.1. Every compound was tested in at least two independent experiments and the values shown correspond to their mean ± s. d.

195 Chapter 7. Experimental Section

7.1.4.2. Inhibition Curves for JARID1A

The antibody-based LANCEUltra activity assay for JARID1A (KDM5A) was performed essentially as described for JMJD2A (KDM4A) (cf. Section 7.1.4.1) with the following modiﬁcations: a solution of 25 nM full-length JARID1A was used with 100 nM of biotinylated H3K4me3 1–21 substrate peptide. The detection mix contained the appropriate Europium-labeled α-

H3K4me2/1 LANCE antibody.

7.1.4.3. Inhibition Curves for JMJD3

The antibody-based LANCEUltra activity assay for JMJD3 (KDM6B) was performed essentially as described for JMJD2A (KDM4A) (cf. Section 7.1.4.1) with the following modiﬁcations: a solution of 50 nM catalytic domain JMJD3 was used with 400 nM of biotinylated H3K27me3 21–44 substrate peptide. The incubation time for the enzymatic reaction was 120 minutes and the detection mix contained the appropriate Europium-labeled α-H3K27me2 LANCE antibody.

7.1.4.4. Competitivity Investigations against 2-Oxoglutarate

In order to assess the competitivity of enzyme inhibition by test compounds to 2-oxoglutarate (2-OG), the LANCEUltra assay was performed as described in Section 7.1.4.1 with varying concentrations both of 2-OG (0. . . 5.0 µM) and inhibitor. This included at least two inhibitor concentrations below and one above the expected IC50 value. Each combination was tested in duplicate. The blank-corrected LANCE signal was compared to the pre-established calibration curve (cf. Section 3.3.1) to determine the amount of demethylated product formed, i. e.

LANCE − 200 [H3K9me ] = · 1 nM. (7.6) 2 275 From this, the reaction velocity v over the incubation time can be obtained as

∆ [H3K9me ] [H3K9me ] v = 2 = 2 . (7.7) ∆ t 45 min These velocity values were used to ﬁt them against the Michaelis-Menten equation v · [2-OG] v = max . (7.8) KM + [2-OG]

The apparent KM values thus obtained were plotted against the concentration of inhibitor to obtain the inhibition constant (Ki) by linear regression to the equation app 0 [inhibitor] KM = KM · 1 + . (7.9) Ki

196 7.1. Biochemical Assays

7.1.5. Ferrozine Assay

In order to assess the binding properties of a compound to iron(II) ions, it was tested in a displacement assay, where it would decolorize the intensely colored Fe(II)-Ferrozine complex. The Ferrozine assay was performed in a total volume of 100 µL on transparent SpectraPlate- 96 MB microtiter plates (PerkinElmer) using a 25 mM sodium acetate and 25 mM acetic acid buﬀer at pH = 4.50. Aqueous solutions of ascorbic acid, ferrous sulfate, and Ferrozine at their working concentrations of 1 M, 1 mM, and 0.6 mM, respectively, were prepared, which resulted in assay concentrations of 100 mM, 50 µM, and 300 µM, respectively. A 13-point dilution series (0. . . 20 mM) of the test compound in DMSO was prepared (typically in 1:2 or 1:3 steps). The assay mixtures for every well were prepared directly in the wells of the 96-well plate according to the following pipetting scheme.

Blank Pos Control Neg Control Sample ascorbic acid 1 M – 10.0 µL 10.0 µL 10.0 µL

FeSO4 1 mM – 5.0 µL 5.0 µL 5.0 µL compound (cﬁnal × 10) – – – 10.0 µL buﬀer 100 µL 35.0 µL 85.0 µL 25.0 µL Ferrozine 0.6 mM – 50.0 µL – 50.0 µL TOTAL 100 µL 100 µL 100 µL 100 µL

After addition of all the components, 50 µL of the Ferrozine stock solution were added into the appropriate wells using an electronic dispenser pipette. Then, absorbance of all wells was measured at λ = 570 nm on a POLARstar Optima microplate reader (BMG Labtech). Plates were stored on a horizontal shaker at room temperature for 30 minutes and measured again in order to ensure consistency in the case of slow chemical equilibria. These assays were typically performed in triplicate. Absorbance values were blank-corrected and used to calculate relative binding of Ferrozine to iron(II) according to the equation

A − A Binding = Sample Neg · 100%. (7.10) APos − ANeg These values were plotted against the concentration of the inhibitor and used to construct inhibition curves as outlined in Section 7.1.2.1, with the resulting IC50 value as a measure of the binding strength of the compound to iron(II).

197 Chapter 7. Experimental Section

7.2. Chemical Synthesis

7.2.1. General Remarks

Standard laboratory glassware and plastic apparatus was used. Starting materials, reagents, and analytical-grade solvents were obtained from commercial sources and used without further purification. Solvents for preparative chromatography were generally purified by distillation prior to use. Reaction progress was monitored by analytical thin-layer chromatography (TLC) on pre-coated silica gel plates with fluorescence indicator (Merck silica gel 60 F254) and spots were detected by ultraviolet light (λ = 254 nm or 365 nm). Flash column chromatography using gradient elution was performed using a Biotage IsoleraOne system with pre-packed Biotage SNAP cartridges or on manually packed column cartridges using Merck silica (particle size 40 – 63 µm) as indicated. Isocratic column chromatography was performed on a manually packed column using Merck silica (particle size 40 – 63 µm). Microwave-assisted syntheses were conducted on a CEM Discover microwave reactor in septum- sealed glass pressure vials at the indicated conditions. 1H-nuclear magnetic resonance (NMR) and proton-decoupled 13C-NMR spectra were recorded in the indicated deuterated solvents on a Bruker Avance DRX, Bruker Avance III HD, or Bruker Avance II+ 400 MHz spectrometer. Chemical shifts (δ) are expressed in parts per million (ppm), and coupling constants (J) in Hz. Chemical shifts are normalized to the expected 1 13 chemical shift of the residual solvent signal, i. e. H δ = 2.50 and C δ = 39.52 for DMSO-d 6 1 13 and H δ = 7.26 and C δ = 77.16 for CDCl3, respectively. The following abbreviations are used: br s (broad singlet), s (singlet), d (doublet), dd (doublet of doublets), ddd (doublet of doublets of doublets), t (triplet), m (multiplet). Where an assignment of resonances to a particular proton or carbon atom is given, this is deduced from two-dimensional NMR experiments such as 1H,1H-correlation spectroscopy (COSY), 1H,13C-heteronuclear multiple bond correlation (HMBC), or 1H,13C-heteronuclear single quantum coherence (HSQC) as appropriate as well as from the expected chemical shifts based on the chemical structure.[221] Mass spectra were recorded using electrospray ionization (ESI-MS) or chemical ionization at atmospheric pressure (APCI-MS) on a Thermo Electron LCQ Advantage or Thermo Scientific Exactive mass spectrometer in positive and/or negative ion mode as indicated. The purity of the final compounds was determined by high-performance liquid chromatography (HPLC) with ultraviolet light (UV) detection at 210 nm. HPLC analysis was performed on an Agilent Technologies 1260 Infinity system, using a Phenomenex Synergi 4µ Hydro-RP 80 A˚ column (250 nm × 4.60 nm). Elution was performed at 30◦C under gradient conditions.

198 7.2. Chemical Synthesis

Eluent A was water containing 0.05% triﬂuoroacetic acid (TFA). Eluent B was acetonitrile, also containing 0.05% TFA. Linear gradient conditions were as follows: 0 – 4 min: A = 90%, B = 10%; 4 – 29 min: linear increase to B = 100%; 29 – 31 min: B = 100%; 31 – 40 min: A = 10%, B = 90%. A ﬂow rate of 1 mL/min was maintained.

7.2.2. Synthesis of Deferasirox Analogs

7.2.2.1. Synthesis of Starting Materials

Methyl 4-hydrazinylbenzoate (114c).

According to the procedure reported in Ref. 212, 10.0 g (65.7 mmol, 1.0 eq) 4-hydrazinylbenzoic acid 114a were suspended in 300 mL methanol. 3.50 mL (6.45 g, 65.7 mmol, 1.0 eq) Con- centrated sulfuric acid were added dropwise, intensifying the reddish color. The mixture was brought to reﬂux, where it became a clear reddish solution and stirred at reﬂux overnight. After 18 h, stirring was suspended, the mixture cooled, and freed from all volatiles in vacuo. The remaining brown powder was resuspended in 700 mL dichloromethane, neutralized by the addition of 300 mL saturated NaHCO3 solution and the organic phase washed with another

2x 150 mL NaHCO3 solution and once with 150 mL brine. The organic phase was dried over

Na2SO4, ﬁltered, and freed of all volatiles in vacuo, leaving behind the title product as an orange powder. Yield: 3.98 g (23.9 mmol, 36 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 7.69 (d, J = 8.8 Hz, 2H), 7.59 (br s, 1H), 6.76 (d, J = 8.8 Hz, 2H), 4.20 (br s, 2H), 3.73 (s, 3H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 166.5, 156.1, 130.8, 116.2, 109.9, 51.2.

2-(2-Hydroxyphenyl)-4H -1,3-benzoxazin-4-one (113a).

199 Chapter 7. Experimental Section

According to the procedure reported in Ref. 210, 2.00 g (14.6 mmol, 1.0 eq) salicylamide 111a and 2.42 g (17.5 mmol, 1.2 eq) salicylic acid 112a were suspended in 3.00 mL of xylenes. 146 µL Dry pyridine were added and the mixture was brought to reflux, where it became a clear golden solution. 2.33 mL (3.82 g, 32.1 mmol, 2.2 eq) Thionyl chloride were added dropwise over 2 hours. An intense evolution of gases was noted after each addition and the solution became much darker. Upon completion of the addition, a yellowish solid material started to precipitate from the then viscous brown solution. Heating was turned off and the mixture stirred for another 30 min. Volatiles were removed in vacuo and the remaining yellow solid material resuspended in 8.80 mL ethanol and 146 µL glacial acetic acid. The mixture was brought to reflux and cooled. The crystalline yellow solid was collected by suction filtration, washed with cold ethanol and dried in vacuo yielding the title product as a yellow powder. When exposed to moisture, the product slowly decomposes to bis(salicyl)imide. 1H-NMR spectra in DMSO-d 6 containing traces of water show both products. Yield: 2.50 g (10.5 mmol, 72 %) 1 H-NMR (400 MHz, CDCl3, TMS, ppm): δ = 12.70 (br s, 1H), 8.19 (dd, J = 8.1 Hz, 1.7 Hz, 1H), 8.09 (dd, J = 8.1 Hz, 1.7 Hz, 1H), 7.79 (ddd, J = 8.5 Hz, 7.3 Hz, 1.7 Hz, 1H), 7.58–7.45 (m, 3H), 7.06 (dd, J = 8.5 Hz, 1.1 Hz, 1H), 6.98 (ddd, J = 8.2 Hz, 7.3 Hz, 1.1 Hz, 1H). 13 1 C{ H}-NMR (101 MHz, CDCl3, TMS, ppm): δ = 165.2, 164.0, 163.2, 154.2, 136.9, 135.7, 128.7, 128.0, 127.3, 119.5, 118.9, 118.3, 117.0, 111.3.

6-Chloro-2-(5-chloro-2-hydroxyphenyl)-4H -1,3-benzoxazin-4-one (113d).

As a variation of the procedure reported in Ref. 210, 2.00 g (11.7 mmol, 1.0 eq) 5-chlorosalicyl- amide 111d and 2.41 g (14.0 mmol, 1.2 eq) 5-chlorosalicylic acid 112d were suspended in 12.4 mL of xylenes. 720 µL Dry pyridine were added and the mixture was brought to reﬂux. 1.86 mL (3.05 g, 25.6 mmol, 2.2 eq) Thionyl chloride were added dropwise over 2 hours, during which the mixture became a clear solution. An intense evolution of gases was noted after each addition and the solution became much darker. Upon completion of the addition, a white solid material started to precipitate from the viscous mixture. The mixture was cooled to room temperature and volatiles were removed in vacuo. The remaining pale yellow solid material

200 7.2. Chemical Synthesis was resuspended in 10.0 mL ethanol and 120 µL glacial acetic acid. The mixture was brought to reﬂux and cooled. The crystalline solid was collected by suction ﬁltration, washed with cold ethanol, and dried in vacuo yielding the title product as a pale yellow powder. Yield: 1.99 g (6.46 mmol, 55 %) 1 H-NMR (400 MHz, CDCl3, TMS, ppm): δ = 12.52 (br s, 1H, OH), 8.18 (d, J = 2.6 Hz, 1H, benzox−H5), 8.04 (d, J = 2.7 Hz, 1H, phe−H6), 7.77 (dd, J = 8.9 Hz, 2.6 Hz, 1H, benzox−H7), 7.52 (d, J = 8.9 Hz, 1H, benzox−H8), 7.49 (dd, J = 9.0 Hz, 2.7 Hz, 1H, phe−H4), 7.06 (d, J = 9.0 Hz, 1H, phe−H3). 13 1 C{ H}-NMR (101 MHz, CDCl3, TMS, ppm): δ = 164.2 (O−C−N), 162.6 (C−O), 4 7 161.8 (C−OH), 152.4 (Cq−O), 137.1 (phe−C ), 136.2 (benzox−C ), 133.4 (Cq−C−O), 127.7 (phe−C6), 127.6 (benzox−C5), 124.6 (phe−C−Cl), 120.7 (phe−C3), 119.3 (benzox−C−Cl), 118.8 (benzox−C8), 111.8 (phe−C1).

HPLC: tR = 24.96 min, 99% purity.

2-(2-Hydroxy-5-methoxyphenyl)-6-methoxy-4H -1,3-benzoxazin-4-one (113e).

As a variation of the procedure reported in Ref. 210, 900 mg (5.38 mmol, 1.0 eq) 5-methoxy- salicylamide 111e and 1.09 g (6.46 mmol, 1.2 eq) 5-methoxysalicylic acid 112e were suspended in 2.20 mL of xylenes. 60.0 µL Dry pyridine were added and the mixture was brought to reflux, where it became a clear orange solution. Under vigorous stirring, 860 µL (1.41 g, 11.8 mmol, 2.2 eq) thionyl chloride were added dropwise over 1.5 hours. An intense evolution of gases was noted after each addition and the solution became much darker. Upon completion of the addition, a solid material started to precipitate from the viscous dark mixture. The mixture was refluxed for another 30 min, cooled to room temperature, and volatiles were removed in vacuo. The remaining brown oil was resuspended in 5.00 mL ethanol and 60.0 µL glacial acetic acid. The mixture was brought to reflux, cooled, and the precipitation completed in an ice bath for one hour. The brown solid material was collected by suction filtration, washed with cold ethanol, and dried in vacuo yielding the title product as a brown powder. Yield: 628 mg (2.10 mmol, 39 %) 1 H-NMR (400 MHz, CDCl3, TMS, ppm): δ = 12.26 (br s, 1H), 7.54 (d, J = 3.0 Hz, 1H), 7.48–7.42 (m, 2H), 7.34 (dd, J = 9.1 Hz, 3.0 Hz, 1H), 7.14 (dd, J = 9.1 Hz, 3.1 Hz, 1H),

201 Chapter 7. Experimental Section

7.00 (d, J = 9.1 Hz, 1H), 3.91 (s, 3H), 3.85 (s, 3H). 13 1 C{ H}-NMR (101 MHz, CDCl3, TMS, ppm): δ = 164.7, 164.3, 158.3, 157.8, 152.3, 148.7, 125.2, 125.2, 119.9, 118.9, 118.5, 110.8, 110.4, 107.6, 56.2, 56.1.

7.2.2.2. Couplings

4-[3,5-Bis(2-hydroxyphenyl)-1H -1,2,4-triazol-1-yl]benzoic acid (109a).

According to the procedure reported in Ref. 209, 1.00 g (4.18 mmol, 1.0 eq) 2-(2-hydroxyphenyl)- 4H -1,3-benzoxazin-4-one 113a and 700 mg (4.60 mmol, 1.1 eq) 4-hydrazinylbenzoic acid 114a were suspended in 15.0 mL of methanol. The mixture was brought to reflux, where it became a clear orange solution. After ca. 45 min, the mixture became turbid and stirring was stopped after 2 hours. Upon cooling, larger amounts of white solid formed, whose precipitation was completed in an ice bath. The off-white solid was collected by suction filtration and the raw material recrystallized from 65 mL of methanol, yielding the title product as pale brown crystals. Yield: 711 mg (1.90 mmol, 46 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 13.21 (br s, 1H), 10.81 (s, 1H), 10.06 (s, 1H), 8.05 (dd, J = 7.8 Hz, 1.6 Hz, 1H), 8.02–7.96 (m, 2H), 7.59–7.53 (m, 3H), 7.43–7.34 (m, 2H), 7.06–6.95 (m, 3H), 6.86 (d, J = 8.2 Hz, 1H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 166.5, 159.9, 156.4, 155.2, 152.1, 141.2, 132.6, 131.5, 131.1, 130.6, 130.3, 126.8, 123.4, 119.7, 119.5, 117.1, 116.2, 114.4, 113.7. m + + ESI-MS positive (MeOH): /z = 374.1137 (100%, [M+H] ). Calcd. for [C21H15N3O4+H] : 374.1135. m − ESI-MS negative (MeOH): /z = 372.0993 (100%, [M − H] ), 328.1094 (19%, [M − CO2 − − − H] ). Calcd. for [C21H15N3O4 − H] : 372.0990. HPLC: tR = 21.75 min, 98% purity.

202 7.2. Chemical Synthesis

4-[3,5-Bis(5-chloro-2-hydroxyphenyl)-1H -1,2,4-triazol-1-yl]benzoic acid (109d).

As a variation of the procedure reported in Ref. 209, 700 mg (2.27 mmol, 1.0 eq) 6-chloro- 2-(5-chloro-2-hydroxyphenyl)-4H -1,3-benzoxazin-4-one 113d and 380 mg (2.50 mmol, 1.1 eq) 4-hydrazinylbenzoic acid 114a were suspended in 52.4 mL of methanol. The mixture was brought to reflux, where it became a clear orange solution over time. After 4.5 hours, stirring was stopped and the mixture cooled to room temperature, when large amounts of white solid material formed, whose precipitation was completed in an ice bath. The material was collected by suction filtration and dried, but by 1H-NMR spectral analysis, this was identified to be starting material 113d. The mother liquor was left standing in a refrigerator for three days until another white solid crystallized, which was collected by suction filtration, dried in vacuo, and identified to be the title product. Yield: 254 mg (0.573 mmol, 25 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 13.23 (br s, 1H), 10.67 (s, 1H), 10.37 (s, 1H), 8.04–7.99 (m, 2H), 7.99 (d, J = 2.7 Hz, 1H), 7.67 (d, J = 2.7 Hz, 1H), 7.62–7.55 (m, 2H), 7.47–7.38 (m, 2H), 7.08 (d, J = 8.8 Hz, 1H), 6.86 (d, J = 8.8 Hz, 1H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 166.5, 158.8, 155.0, 154.2, 151.0, 141.0, 132.4, 131.1, 130.8, 130.5, 130.4, 126.1, 123.4, 123.3, 122.8, 119.2, 117.9, 116.0, 115.4. m + ESI-MS positive (MeOH): /z = 442.03558 (65%, [M+H] ). Calcd. for [C21H13Cl2N3O4 + H]+: 442.03559. m − ESI-MS negative (MeOH): /z = 440.02112 (100%, [M−H] ), 396.03128 (63%, [M−CO2 − − − H] ). Calcd. for [C21H13Cl2N3O4 − H] : 440.02104. Mass spectra for this compound show the expected isotope peak pattern with +2 and +4 shifts as expected for a dichloro compound.

HPLC: tR = 23.58 min, 98% purity.

203 Chapter 7. Experimental Section

4-[3,5-Bis(2-hydroxy-5-methoxyphenyl)-1H -1,2,4-triazol-1-yl]benzoic acid (109e).

As a variation of the procedure reported in Ref. 209, 300 mg (1.00 mmol, 1.0 eq) 2-(2-hydoxy-5- methoxyphenyl)-6-methoxy-4H -1,3-benzoxazin-4-one 113e and 168 mg (1.10 mmol, 1.1 eq) 4- hydrazinylbenzoic acid 114a were suspended in 3.60 mL of methanol. The mixture was brought to reflux, where it shortly became a clear reddish-brown solution. After 3.5 hours, stirring was stopped and the mixture cooled to room temperature, when little brown solid material formed. Crystallization was completed over night in a refrigerator. The brownish crystals were collected by suction filtration, washed with methanol, dried in vacuo, and identified to be the title product. Yield: 79.7 mg (0.184 mmol, 18 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 13.07 (br s, 1H), 10.35 (s, 1H), 9.53 (s, 1H), 8.00 (d, J = 8.6 Hz, 2H), 7.58 (d, J = 8.6 Hz, 2H), 7.52 (d, J = 2.4 Hz, 1H), 7.12 (d, J = 3.1 Hz, 1H), 7.04–6.92 (m, 3H), 6.78 (d, J = 8.9 Hz, 1H), 3.77 (s, 3H), 3.72 (s, 3H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 166.5, 159.8, 152.3, 152.0, 152.0, 150.4, 149.0, 141.2, 130.6, 130.3, 123.5, 118.7, 118.5, 118.1, 117.1, 115.2, 114.5, 113.6, 109.8, 55.6, 55.5. + + ESI-MS positive (MeOH): m/z = 434.13467 (100%, [M+H] ), 456.11658 (32%, [M+Na] ). + Calcd. for [C23H19N3O6 + H] : 434.13466. m − ESI-MS negative (MeOH): /z = 432.12024 (100%, [M−H] ), 388.13055 (57%, [M−CO2 − − − − H] ), 373.10703 (73%, [M − CO2 − CH3 − H] ). Calcd. for [C23H19N3O6 − H] : 432.12011. HPLC: tR = 21.71 min, 98% purity.

2,2’-(1-Phenyl-1H -1,2,4-triazole-3,5-diyl)diphenol (109b).

204 7.2. Chemical Synthesis

According to the procedure reported in Ref. 210, 145 mg (1.00 mmol, 1.2 eq) phenylhydrazine hydrochloride 114b were suspended in 8.00 mL of ethanol and brought to reflux, resulting in a pale yellow solution. To this were added 200 mg (0.836 mmol, 1.0 eq) 2-(2-hydroxyphenyl)- 4H -1,3-benzoxazin-4-one 113a and the mixture stirred at reflux for 2.5 hours. After cooling to room temperature, ∼10 mL of 1 mol/L hydrochloric acid were added dropwise, resulting in the formation of a white precipitate, which was filtered off, washed with more 1 mol/L HCl and dried in vacuo yielding the title product as a white powder. Yield: 155 mg (0.469 mmol, 56 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 10.89 (s, 1H), 10.06 (s, 1H), 8.04 (dd, J = 7.8 Hz, 1.7 Hz, 1H), 7.50–7.41 (m, 6H), 7.40–7.32 (m, 2H), 7.05–6.97 (m, 2H), 6.94 (ddd, J = 7.5 Hz, 7.5 Hz, 1.0 Hz, 1H), 6.87 (dd, J = 8.3 Hz, 0.8 Hz, 1H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 159.5, 156.3, 155.4, 151.7, 137.7, 132.3, 131.3, 131.0, 129.2, 128.6, 126.6, 123.8, 119.6, 119.2, 117.0, 116.1, 114.6, 113.8. m + APCI-MS positive (MeOH): /z = 330.1230 (100%, [M+H] ). Calcd. for [C20H15N3O2 + H]+: 330.1237. m − APCI-MS negative (MeOH): /z = 328.1092 (100%, [M−H] ). Calcd. for [C20H15N3O2 − H]−: 328.1092.

HPLC: tR = 22.94 min, 97% purity.

Methyl 4-[3,5-bis(2-hydroxyphenyl)-1H -1,2,4-triazol-1-yl]benzoate (109c).

According to the procedure reported in Ref. 211, 3.00 g (12.5 mmol, 1.0 eq) 2-(2-hydroxyphenyl)- 4H -1,3-benzoxazin-4-one 113a and 2.29 g (13.8 mmol, 1.1 eq) methyl 4-hydrazinylbenzoate 114c were suspended in 200 mL of ethanol. 1.77 mL (1.29 g, 12.8 mmol, 1.02 eq) Triethylamine were added dropwise and the mixture was brought to reﬂux, where it became a clear orange solution. After 2.5 hours, heating was stopped and stirring was continued for another hour. The mixture was diluted with 300 mL distilled water, when a white precipitate formed. The mixture was concentrated to about half its volume in vacuo and acidiﬁed by the addition of ∼10 mL glacial acetic acid. 300 mL Dichloromethane were added and the phases separated.

205 Chapter 7. Experimental Section

The aqueous phase was extracted with further 3x 150 mL dichloromethane. The combined organic layers were dried over Na2SO4, ﬁltered, and freed from all volatiles in vacuo. The remaining orange powder was further puriﬁed by column chromatography (100 g silica gel, manually packed Biotage cartridge using dry load sorbent, dichloromethane/methanol 100:0% → 99:1%), yielding the title product as a pale yellow powder. Yield: 2.83 g (7.31 mmol, 58 %) 1 H-NMR (400 MHz, CDCl3, TMS, ppm): δ = 11.28 (br s, 1H), 9.59 (br s, 1H), 8.27– 8.21 (m, 2H), 8.17–8.10 (m, 1H), 7.64–7.58 (m, 2H), 7.42–7.31 (m, 2H), 7.18–7.12 (m, 1H), 7.11–7.00 (m, 2H), 6.92 (dd, J = 8.0 Hz, 1.6 Hz, 1H), 6.71–6.62 (m, 1H), 4.00 (s, 3H). 13 1 C{ H}-NMR (101 MHz, CDCl3, TMS, ppm): δ = 165.9, 159.6, 158.1, 156.6, 152.2, 141.6, 133.3, 132.1, 131.8, 131.4, 127.8, 127.7, 126.2, 120.1, 119.2, 118.6, 117.3, 113.1, 109.9, 52.8. + + ESI-MS positive (MeOH): m/z = 388.1292 (100%, [M + H] ), 410.1111 (73%, [M + Na] ). + Calcd. for [C22H17N3O4 + H] : 388.1292. m − ESI-MS negative (MeOH): /z = 386.1150 (100%, [M − H] ). Calcd. for [C22H17N3O4 − H]−: 386.1146.

HPLC: tR = 25.39 min, 98% purity.

206 7.2. Chemical Synthesis

7.2.3. Synthesis of Aminopyrimidylpyridine-Based Inhibitors

7.2.3.1. Synthesis of Starting Materials

Methyl 2-acetylpyridine-4-carboxylate (106).

In accordance with the procedure reported in Refs. 206 and 207, 10.0 g (72.9 mmol, 1.0 eq) methyl isonicotinate 105 and 61.5 mL (48.2 g, 1.09 mol, 15 eq) acetaldehyde were dissolved in 150 mL acetonitrile at room temperature and 345 mg (1.24 mmol, 0.017 eq) ferrous sulfate heptahydrate and 5.73 mL (8.48 g, 74.4 mmol, 1.02 eq) trifluoroacetic acid were added. 18.8 mL (13.1 g, 146 mmol, 2.0 eq) of a 70% aqueous solution of tert-butyl hydroperoxide were added dropwise. The then intensely yellow solution was brought to reflux and stirred for 3.5 hours, after which the reaction was stopped, the mixture cooled to room temperature and freed from all volatiles in vacuo. The black residual oil was taken up in 120 mL saturated NaHCO3 solution until neutral and extracted with 4x 45 mL toluene. The combined organic layers were dried over Na2SO4, filtered, and solvents removed in vacuo. The remaining reddish-brown liquid was purified by column chromatography (manually packed 25 cm × 6 cm silica gel, cyclohexane/ethyl acetate 1:1) and the middle fractions combined and concentrated in vacuo. Upon treatment of the residue with light petroleum, a yellowish solid crystallized, which was filtered off and washed with light petroleum. The solid was dried in vacuo yielding the title product in sufficient purity to be used for subsequent syntheses. Yield: 5.84 g (32.6 mmol, 45 %) 1 6 H-NMR (400 MHz, CDCl3, TMS, ppm): δ = 8.82 (dd, J = 4.9 Hz, 0.9 Hz, 1H, H ), 8.54 (dd, J = 1.6 Hz, 0.9 Hz, 1H, H3), 8.01 (dd, J = 4.9 Hz, 1.6 Hz, 1H, H5), 3.97 (s, 3H, ester−CH3), 2.74 (s, 3H, acetyl−CH3). 13 1 C{ H}-NMR (101 MHz, CDCl3, TMS, ppm): δ = 199.4 (acetyl−CO), 165.2 (ester−CO), 2 6 4 3 5 154.6 (C ), 150.0 (C ), 138.7 (C ), 126.2 (C ), 121.1 (C ), 53.0 (ester−CH3), 26.0 (acetyl−CH3). m + + APCI-MS positive (MeOH): /z = 180.0658 (100%, [M+H] ). Calcd. for [C9H9NO3+H] : 180.0655.

207 Chapter 7. Experimental Section

(E)-3-(Dimethylamino)-1-(pyridin-2-yl)prop-2-en-1-one (115).

In accordance with the procedure reported in Ref. 208, 1.85 mL (2.00 g, 16.5 mmol, 1.0 eq) 2-acetylpyridine and 2.19 mL (1.97 g, 16.5 mmol, 1.0 eq) dimethylformamide-dimethylacetal (DMF-DMA) were placed in a 10 mL microwave vial and sealed. The mixture was heated at max. power of 50 W to 150◦C and the temperature held for 5 min (max. pressure = 3 bar). After cooling to room temperature, the brown oil was diluted with methanol, transferred, and freed from all volatiles in vacuo. The then black solid was washed extensively with xylenes, ﬁltered, and dried in vacuo yielding the title product as a dark green powder. Yield: 1.81 g (10.3 mmol, 62 %) 1 H-NMR (400 MHz, CDCl3, TMS, ppm): δ = 8.62 (ddd, J = 4.8 Hz, 1.8 Hz, 0.9 Hz, 1H), 8.18–8.06 (m, 1H), 7.91 (d, J = 12.6 Hz, 1H), 7.85–7.75 (m, 1H), 7.35 (ddd, J = 7.5 Hz, 4.8 Hz, 1.3 Hz, 1H), 6.44 (d, J = 12.6 Hz, 1H), 3.17 (s, 3H), 2.99 (s, 3H). 13 1 C{ H}-NMR (101 MHz, CDCl3, TMS, ppm): δ = 187.0, 156.4, 154.9, 148.4, 136.9, 125.5, 122.2, 91.3, 45.3, 37.6. m + APCI-MS positive (MeOH): /z = 177.1025 (100%, [M + H] ). Calcd. for [C10H12N2O + H]+: 177.1022.

Methyl 2-[(E)-3-(dimethylamino)prop-2-enoyl]pyridine-4-carboxylate (107).

As a variation of the procedure reported in Ref. 208, 2.50 g (14.0 mmol, 1.0 eq) methyl 2- acetylpyridine-4-carboxylate 106 and 1.85 mL (1.67 g, 14.0 mmol, 1.0 eq) dimethylformamide- dimethylacetal (DMF-DMA) were placed in a 10 mL microwave vial and sealed. The mixture was heated at max. power of 50 W to 150◦C and the temperature held for 5 min (max. pressure = 3 bar). After cooling to room temperature, another 1.50 mL of DMF-DMA were added and the irradiation cycle repeated. Then, the brown liquid was diluted with methanol,

208 7.2. Chemical Synthesis transferred, and freed from all volatiles in vacuo. The then black oil was washed extensively with xylenes and light petroleum, ﬁltered, and dried in vacuo yielding the title product as a brown powder. Yield: 2.01 g (8.58 mmol, 61 %) 1 6 H-NMR (400 MHz, CDCl3, TMS, ppm): δ = 8.76 (d, J = 4.9 Hz, 1H, H ), 8.64 (s, 1H, 3 5 H ), 7.93 (d, J = 12.5 Hz, 1H, CHNMe2), 7.90 (dd, J = 4.9 Hz, 1.5 Hz, 1H, H ), 6.43 (d, J =

12.5 Hz, 1H, COCH), 3.95 (s, 3H, ester−CH3), 3.18 (s, 3H, N(CH3)2), 2.99 (s, 3H, N(CH3)2). 13 1 C{ H}-NMR (101 MHz, CDCl3, TMS, ppm): δ = 185.9, 165.8, 157.5, 155.2, 149.2, 138.4, 124.6, 121.4, 91.1, 52.8, 45.4, 37.5. m + APCI-MS positive (MeOH): /z = 235.1077 (100%, [M+H] ). Calcd. for [C12H14N2O3 + H]+: 235.1077.

7.2.3.2. Synthesis of Guanidines

Pyridin-4-ylmethylguanidine (104b).

As a variation of the procedure reported in Refs. 202 and 205, 2.98 g (20.3 mmol, 1.1 eq) 1H - pyrazole-1-carboxamidine hydrochloride 103 were suspended in 10 mL dimethylformamide under nitrogen. 3.54 mL (2.63 g, 20.3 mmol, 1.1 eq) Diisopropylethylamine (DIPEA) were added, yielding a clear orange solution. 1.87 mL (2.00 g, 18.5 mmol, 1.0 eq) 4-Aminomethylpyridine 102b were added dropwise, immediately forming a precipitate. The mixture was stirred at room temperature under nitrogen for 4 hours until it became a clear solution again. The solvents were removed in vacuo and the remaining viscous orange oil was treated with ∼10 mL saturated NaHCO3 solution, immediately forming a white crystalline precipitate, which was

filtered off, washed with NaHCO3 solution and diethyl ether, and dried in vacuo yielding the title product as an off-white crystalline solid. Yield: 2.18 g (14.5 mmol, 78 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.53 (dd, J = 4.4 Hz, 1.6 Hz, 2H), 7.27 (dd, J = 4.4 Hz, 1.6 Hz, 2H), 4.38 (s, 2H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 159.9 (guanidine), 149.7, 147.2, 121.9, 42.8.

209 Chapter 7. Experimental Section

+ + APCI-MS positive (MeOH): m/z = 151.0979 (100%, [M+H] ), 301.1883 (6%, [2 M+H] ). + Calcd. for [C7H10N4 + H] : 151.0978.

Benzylguanidine (104ac).

As a variation of the procedure reported in Refs. 202 and 205, 2.25 g (15.3 mmol, 1.1 eq) 1H -pyrazole-1-carboxamidine hydrochloride 103 and 2.00 g (13.9 mmol, 1.0 eq) benzylamine hydrochloride 102ac were suspended in 8.0 mL dimethylformamide under nitrogen. 5.34 mL (3.96 g, 30.6 mmol, 2.2 eq) Diisopropylethylamine (DIPEA) were added, yielding a yellowish suspension. The mixture was stirred at room temperature under nitrogen for 4 hours until it became a clear solution. The solvents were removed in vacuo and the remaining viscous yellow oil was treated with ∼8 mL saturated NaHCO3 solution, allowing to slowly form a white crystalline precipitate, which was filtered off, washed with NaHCO3 solution and diethyl ether, and dried in vacuo yielding the title product as an off-white crystalline solid. Yield: 1.14 g (7.64 mmol, 55 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.39 (br s, 4H, guanidine), 7.43–7.23

(m, 5H), 4.32 (s, 2H, benzyl−CH2). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 157.4 (guanidine), 137.9, 128.5, 127.3, 127.2, 43.7. m + + ESI-MS positive (MeOH): /z = 150.1026 (100%, [M + H] ). Calcd. for [C8H11N3 + H] : 150.1026.

3,4-Dihydro-2H -chromen-6-ylmethylguanidine (104ag).

As a variation of the procedure reported in Refs. 202 and 205, 0.988 g (6.74 mmol, 1.1 eq) 1H - pyrazole-1-carboxamidine hydrochloride 103 were suspended in 3.5 mL dimethylformamide under nitrogen. 1.17 mL (0.871 g, 6.74 mmol, 1.1 eq) Diisopropylethylamine (DIPEA) were

210 7.2. Chemical Synthesis added, yielding a clear orange solution. 1.00 g (6.13 mmol, 1.0 eq) 3,4-Dihydro-2H -chromen-6- ylmethylamine 102ag were added dropwise, immediately forming a precipitate. The mixture was diluted with another 2.0 mL of DMF and stirred at room temperature under nitrogen for 4 hours until it became a clear solution again. The solvents were removed in vacuo and the remaining viscous orange oil was treated with ∼5 mL saturated NaHCO3 solution, immediately forming a white crystalline precipitate, which was filtered off, washed with NaHCO3 solution and diethyl ether, and dried in vacuo yielding the title product as an off-white crystalline solid. Yield: 1.06 g (5.16 mmol, 84 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.34 (br s, 4H, guanidine), 7.01–6.96 (m, 2H), 6.69 (d, J = 8.9 Hz, 1H), 4.17 (s, 2H), 4.10 (t, J = 5.0 Hz, 2H), 2.70 (t, J = 6.4 Hz, 2H), 1.93–1.85 (m, 2H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 159.9 (guanidine), 153.9, 129.0, 128.9, 126.4, 122.2, 116.3, 65.9, 43.3, 24.3, 21.8. m + + ESI-MS positive (MeOH): /z = 206.1288 (100%, [M+H] ). Calcd. for [C11H15N3O+H] : 206.1288.

3-(Methylsulfanyl)benzylguanidine (104aj).

As a variation of the procedure reported in Refs. 202 and 205, 680 mg (4.64 mmol, 1.1 eq) 1H - pyrazole-1-carboxamidine hydrochloride 103 and 800 mg (4.22 mmol, 1.0 eq) 3-(methylsulfanyl)- benzylamine hydrochloride 102aj were suspended in 2.5 mL dimethylformamide under nitrogen. 1.62 mL (1.20 g, 9.28 mmol, 2.2 eq) Diisopropylethylamine (DIPEA) were added, yielding a brownish suspension. The mixture was stirred at room temperature under nitrogen for 5.5 hours until it became a clear solution. The solvents were removed in vacuo and the remaining viscous brown oil was treated with ∼3 mL saturated NaHCO3 solution, allowing to slowly form a brown solid and precipitation was completed over ice. The solid was ﬁltered oﬀ, washed with NaHCO3 solution and diethyl ether, and dried in vacuo yielding the title product as a dark greyish solid. Yield: 978 mg (3.75 mmol, 89 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.62 (br s, 4H, guanidine), 7.32–7.23 (m, 1H), 7.20 (s, 1H), 7.16 (d, J = 7.6 Hz, 1H), 7.06 (d, J = 7.6 Hz, 1H), 4.28 (s, 2H), 2.45 (s, 3H).

211 Chapter 7. Experimental Section

13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 157.6 (guanidine), 138.9, 138.4, 129.1, 124.7, 124.6, 123.7, 43.4, 14.6. m + ESI-MS positive (MeOH): /z = 196.0905 (100%, [M + H] ), 137.0421 (68%, [H3CS− + + tropylium] ). Calcd. for [C9H13N3S + H] : 196.0903.

7.2.3.3. Couplings

4-(Pyridin-2-yl)-N -(pyridin-4-ylmethyl)pyrimidin-2-amine (94b).

As a variation of the procedure reported in Refs. 202 and 203, 200 mg (1.14 mmol, 1.0 eq) (E)-3-(dimethylamino)-1-(pyridin-2-yl)prop-2-en-1-one 115 and 205 mg (1.36 mmol, 1.2 eq) pyridin-4-ylmethylguanidine 104b were suspended in 4.0 mL iso-propanol under nitrogen. Approx. 150 mg of oven-heated molecular sieve granules were added and the mixture was brought to reflux under nitrogen and stirred for 3.5 hours, while occasionally adding more iso-propanol. Stirring under nitrogen was continued overnight at room temperature and the mixture refluxed again for another 30 min the next day. After cooling to room temperature, the mixture was diluted with 30 mL of methanol and refluxed for 15 min yielding a clear golden solution. The cold solution was filtered over a pad of celite to remove the molecular sieves and washed with methanol. The solution was freed of all volatiles in vacuo, yielding a brown solid, which was treated with ∼5 mL ethyl acetate. The remaining solid was filtered off, washed with little cold ethyl acetate, and dried in vacuo to yield the title product as a brown powder. Yield: 125 mg (0.475 mmol, 42 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.69 (d, J = 3.5 Hz, 1H), 8.48 (dd, J = 4.5 Hz, 1.5 Hz, 2H), 8.45 (d, J = 4.8 Hz, 1H), 8.25 (br s, 1H), 8.01–7.92 (m, 2H), 7.51 (d, J = 5.0 Hz, 2H), 7.40–7.31 (m, 2H), 4.61 (s, 2H). m + + APCI-MS positive (MeOH): /z = 264.1248 (100%, [M+H] ). Calcd. for [C15H13N5+H] : 264.1244.

HPLC: tR = 9.09 min, >99% purity.

212 7.2. Chemical Synthesis

Methyl 2-{2-[(pyridin-4-ylmethyl)amino]pyrimidin-4-yl}pyridine-4-carboxylate (93b).

As a variation of the procedure reported in Refs. 202 and 203, 300 mg (1.28 mmol, 1.0 eq) methyl 2-[(E)-3-(dimethylamino)prop-2-enoyl]pyridine-4-carboxylate 107 and 231 mg (1.54 mmol, 1.2 eq) pyridin-4-ylmethylguanidine 104b were suspended in 4.0 mL iso-propanol under nitrogen. Approx. 180 mg of oven-heated molecular sieve granules were added and the mixture was brought to reflux under nitrogen and stirred for 2.5 hours, while occasionally adding more iso-propanol. Stirring under nitrogen was continued overnight at room temperature and the mixture refluxed again for another 1.5 hours the next day. After cooling to room temperature, the mixture was diluted with 30 mL of methanol and refluxed for 30 min yielding a turbid brown solution. The cold solution was filtered over a pad of celite to remove the molecular sieves and washed with methanol. The solution was freed of all volatiles in vacuo, yielding a brown oil, which was treated with ∼5 mL ethyl acetate. However, no separation occurred and the mixture was again concentrated in vacuo. The remaining oil was purified by column chromatography (Telos-20 g pre-packed column, dichloromethane/methanol 2% → 20%), yielding a reddish oil. Upon treatment with ethyl acetate, a yellowish solid crystallized, which was filtered off and dried in vacuo to yield the title product as pale yellow crystals. Yield: 45 mg (0.140 mmol, 11 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.90 (d, J = 3.7 Hz, 1H), 8.73 (br s, 1H), 8.54–8.42 (m, 3H), 8.19–8.07 (m, 1H), 7.94 (s, 1H), 7.52 (d, J = 4.9 Hz, 1H), 7.36 (s, 2H), 4.58 (d, J = 6.2 Hz, 2H), 3.95 (s, 3H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 165.0, 162.4, 155.2, 150.8, 149.5, 138.2, 129.5, 124.9, 124.2, 122.2, 120.3, 119.6, 106.3, 53.0, 43.7. + + ESI-MS positive (MeOH): m/z = 322.1300 (100%, [M + H] ), 344.1119 (7%, [M + Na] ). + Calcd. for [C17H15N5O2 + H] : 322.1299. HPLC: tR = 13.09 min, 94% purity.

213 Chapter 7. Experimental Section

Methyl 2-[2-(benzylamino)pyrimidin-4-yl]pyridine-4-carboxylate (93ac).

As a variation of the procedure reported in Refs. 202 and 203, 229 mg (0.978 mmol, 1.0 eq) methyl 2-[(E)-3-(dimethylamino)prop-2-enoyl]pyridine-4-carboxylate 107 and 175 mg (1.17 mmol, 1.2 eq) benzylguanidine 104ac were suspended in 3.0 mL iso-propanol under nitrogen. Approx. 110 mg of oven-heated molecular sieve granules were added and the mixture was brought to reflux and stirred under nitrogen, while occasionally adding more iso-propanol. After 6 h, the reaction was stopped, the mixture cooled to room temperature and diluted with 30 mL of methanol and refluxed for 15 min yielding a turbid brown solution. The cold solution was filtered over a pad of celite to remove the molecular sieves and washed with methanol. The solution was freed of all volatiles in vacuo, yielding a yellow solid, which was treated with ∼3 mL ethyl acetate. However, no separation occurred and the mixture was again concentrated in vacuo. The remaining oil was purified by column chromatography (Telos-20 g pre-packed column, cyclohexane/ethyl acetate 7% → 60%), yielding a sticky yellow solid. Upon treatment with methanol, white crystals formed, which were filtered off, washed with methanol and dried in vacuo. NMR analysis revealed this compound, however, to be the iso-propyl ester of the title product. For the transesterification, all remaining solid was taken up in ∼20 mL of methanol [222] and 5.6 mg K2HPO4 · 3H2O were added as catalyst. The mixture was refluxed for 4 hours, filtered over a pad of celite to remove inorganic salt, and freed of all volatiles in vacuo yielding the title compound as a white powder. Yield: 73 mg (0.228 mmol, 23 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.90 (d, J = 4.8 Hz, 1H), 8.77 (s, 1H), 8.47 (d, J = 5.0 Hz, 1H), 8.06 (m, 1H), 7.95 (d, J = 4.1 Hz, 1H), 7.49 (d, J = 4.9 Hz, 1H), 7.39 (s, 2H), 7.31 (m, 2H), 7.25–7.16 (m, 1H), 4.57 (d, J = 6.3 Hz, 2H), 3.95 (s, 3H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 165.0, 162.4, 159.9, 155.3, 152.8, 150.7, 140.4, 138.1, 128.2, 127.2, 126.6, 124.1, 119.5, 105.9, 53.0, 44.3. m + APCI-MS positive (MeOH): /z = 321.1347 (100%, [M+H] ). Calcd. for [C18H16N4O2 + H]+: 321.1346.

HPLC: tR = 20.09 min, 96% purity.

214 7.2. Chemical Synthesis

N -Benzyl-4-(pyridin-2-yl)pyrimidin-2-amine (94ac).

As a variation of the procedure reported in Refs. 202 and 203, 300 mg (1.70 mmol, 1.0 eq) (E)- 3-(dimethylamino)-1-(pyridin-2-yl)prop-2-en-1-one 115 and 305 mg (2.04 mmol, 1.2 eq) benzylguanidine 104ac were suspended in 5.0 mL iso-propanol under nitrogen. Approx. 180 mg of oven-heated molecular sieve granules were added and the mixture was brought to reflux under nitrogen and stirred for 5 hours, while occasionally adding more iso-propanol. After cooling to room temperature, the mixture was diluted with 45 mL of methanol and refluxed for 30 min. The cold solution was filtered over a pad of celite to remove the molecular sieves and washed with methanol. The clear orange solution was freed of all volatiles in vacuo, yielding a brown solid, which was treated with ∼5 mL ethyl acetate, which resulted in the formation of a crystalline solid that could not be filtered off, however. Upon standing for a few days, a larger amount of solid material separated from the mother liquor, which was filtered off, dried, and purified by column chromatography (Telos-10 g pre-packed column, cyclohexane/ethyl acetate 7% → 60%), yielding the title product as a white powder. Yield: 45.3 mg (0.173 mmol, 10 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.69 (d, J = 4.1 Hz, 1H), 8.43 (d, J = 5.0 Hz, 1H), 8.32 (d, J = 7.9 Hz, 1H), 7.97 (m, 1H), 7.88 (m, 1H), 7.55–7.44 (m, 2H), 7.43–7.34 (m, 2H), 7.34–7.26 (m, 2H), 7.20 (m, 1H), 4.60 (d, J = 5.9 Hz, 2H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 162.4, 159.5, 157.2, 154.0, 149.4, 140.6, 137.4, 128.2, 127.2, 126.5, 125.5, 120.8, 105.8, 44.1. m + + APCI-MS positive (MeOH): /z = 263.1293 (100%, [M + H] ), 91.0542 (8%, [C7H7] ). + Calcd. for [C16H14N4 + H] : 263.1291. HPLC: tR = 17.46 min, >99% purity.

215 Chapter 7. Experimental Section

Methyl 2-{2-[(3,4-dihydro-2H -chromen-6-ylmethyl)amino]pyrimidin-4-yl}pyridine- 4-carboxylate (93ag).

As a variation of the procedure reported in Refs. 202 and 203, 400 mg (1.71 mmol, 1.0 eq) methyl 2-[(E)-3-(dimethylamino)prop-2-enoyl]pyridine-4-carboxylate 107 and 421 mg (2.05 mmol, 1.2 eq) 3,4-dihydro-2H -chromen-6-ylmethylguanidine 104ag were suspended in 5.0 mL iso-propanol under nitrogen. Approx. 180 mg of oven-heated molecular sieve granules were added and the mixture was brought to reflux under nitrogen and stirred for 5 hours, while occasionally adding more iso-propanol. After cooling to room temperature, the mixture was diluted with 45 mL of methanol and refluxed for 45 min yielding a turbid brown solution. The cold solution was filtered over a pad of celite to remove the molecular sieves and washed with methanol. The clear orange solution was freed of all volatiles in vacuo, yielding an orange- brown oil, which was taken up in 25 mL methanol. A few crystals of K2HPO4 · 3H2O were added[222] and the mixture brought to reflux. After 5 hours, the mixture had become very turbid, was filtered again over a pad of celite and concentrated in vacuo, yielding a brown oil, which was purified by column chromatography (Telos-20 g pre-packed column, cyclohexane/ethyl acetate 7% → 60%), yielding a yellow crystalline substance, which was dried in vacuo. NMR analysis revealed this, however, to still contain traces of the iso-propyl ester. The remaining solid material was, thus, suspended again in 45 mL of methanol and refluxed for three hours. Several spoons of celite were added to decolorize the solution and the mixture was filtered twice over a glass frit while hot and once, after cooling, over cotton to remove celite, resulting in a clear and only pale-yellow solution, which was freed from all volatiles in vacuo, yielding the title compound as an off-white powdery solid substance. Yield: 131 mg (0.349 mmol, 20 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.90 (d, J = 4.9 Hz, 1H), 8.80 (s, 1H), 8.47 (d, J = 4.9 Hz, 1H), 7.99–7.89 (m, 2H), 7.49 (d, J = 4.9 Hz, 1H), 7.08 (br s, 2H), 6.65 (d, J = 8.5 Hz, 1H), 4.44 (d, J = 6.2 Hz, 2H), 4.07 (t, J = 4.8 Hz, 2H), 3.95 (s, 3H), 2.69 (t, J = 6.3 Hz, 2H), 1.94–1.79 (m, 2H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 165.0, 162.4, 159.8, 155.3, 153.3, 150.8, 138.1, 131.6, 128.9, 126.3, 124.1, 121.8, 120.2, 119.5, 116.0, 105.7, 65.8, 53.0, 43.7, 24.3, 21.9.

216 7.2. Chemical Synthesis

m + APCI-MS positive (MeOH): /z = 377.1610 (100%, [M+H] ). Calcd. for [C21H20N4O3 + H]+: 377.1608.

HPLC: tR = 21.06 min, 98% purity.

Methyl 2-(2-{[3-(methylsulfanyl)benzyl]amino}pyrimidin-4-yl)pyridine-4-carboxylate (93aj).

As a variation of the procedure reported in Refs. 202 and 203, 400 mg (1.71 mmol, 1.0 eq) methyl 2-[(E)-3-(dimethylamino)prop-2-enoyl]pyridine-4-carboxylate 107 and 400 mg (2.05 mmol, 1.2 eq) 3-(methylsulfanyl)benzylguanidine 104aj were suspended in 5.0 mL iso-propanol under nitrogen. Approx. 180 mg of oven-heated molecular sieve granules were added and the mixture was brought to reflux under nitrogen and stirred for 4.5 hours, while occasionally adding more iso-propanol. After cooling to room temperature, the mixture was diluted with 50 mL of methanol and refluxed for 1 hour and 45 min yielding a turbid solution. The cold solution was filtered over a pad of celite to remove the molecular sieves and washed with methanol. The clear reddish solution was freed of all volatiles in vacuo, yielding a brown oil, which was dried in vacuo. The residue was purified by column chromatography (Telos-20 g pre-packed column, cyclohexane/ethyl acetate 7% → 60%), yielding the title product as an off-white powder. Yield: 173 mg (0.472 mmol, 28 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.90 (d, J = 4.9 Hz, 1H), 8.78 (s, 1H), 8.48 (d, J = 5.0 Hz, 1H), 8.05 (m, 1H), 7.95 (d, J = 3.7 Hz, 1H), 7.51 (d, J = 5.0 Hz, 1H), 7.32–7.21 (m, 2H), 7.22–7.13 (m, 1H), 7.14–7.06 (m, 1H), 4.55 (d, J = 6.3 Hz, 2H), 3.95 (s, 3H), 2.42 (s, 3H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 165.0, 162.4, 159.7, 155.2, 150.8, 141.2, 138.1, 138.0, 129.5, 128.9, 124.7, 124.1, 124.1, 120.3, 119.5, 106.0, 53.0, 44.1, 14.6. m + APCI-MS positive (MeOH): /z = 367.1224 (100%, [M+H] ). Calcd. for [C19H18N4O2S+ H]+: 367.1223.

HPLC: tR = 22.03 min, 96% purity.

217 Chapter 7. Experimental Section

7.2.3.4. Saponiﬁcations

2-[2-(Benzylamino)pyrimidin-4-yl]pyridine-4-carboxylic acid (86ac).

As a variation of the procedure reported in Ref. 202, 53.0 mg (0.165 mmol, 1.0 eq) of methyl 2-[2-(benzylamino)pyrimidin-4-yl]pyridine-4-carboxylate 93ac were suspended in a mixture of 1.4 mL methanol and 1.4 mL tetrahydrofuran. 1.4 mL of a 1 mol/L aqueous lithium hydroxide solution (1.40 mmol, 8.5 eq) were added and the mixture heated gently to 40◦C. After 6 hours, the mixture was cooled to room temperature and the reaction stopped by the dropwise addition of ∼2 mL 2 mol/L hydrochloric acid. Immediately, a white solid formed, whose precipitation was completed over ice. The solid was ﬁltered oﬀ, washed with little aqueous hydrochloric acid, and dried in vacuo yielding the title compound as a white powder. Yield: 38.3 mg (0.125 mmol, 77 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 10.02 (s, 1H), 8.92 (d, J = 4.9 Hz, 1H), 8.80 (s, 1H), 8.67 (br s, 1H), 8.53 (d, J = 3.3 Hz, 1H), 7.99 (dd, J = 4.9 Hz, 1.6 Hz, 1H), 7.64 (d, J = 5.1 Hz, 1H), 7.49–7.38 (m, 2H), 7.33 (m, 2H), 7.24 (m, 1H), 4.66 (s, 2H). 13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 165.8, 164.0, 156.3, 154.2, 150.8, 139.6, 139.5, 134.6, 129.5, 129.2, 128.4, 127.5, 126.9, 105.9, 44.4. m + + ESI-MS positive (MeOH): /z = 307.1191 (100%, [M+H] ). Calcd. for [C17H14N4O2+H] : 307.1190. m − ESI-MS negative (MeOH): /z = 305.1045 (100%, [M − H] ). Calcd. for [C17H14N4O2 − H]−: 305.1044.

HPLC: tR = 17.17 min, >99% purity.

2-(2-{[3-(methylsulfanyl)benzyl]amino}pyrimidin-4-yl)pyridine-4-carboxylic acid (86aj).

As a variation of the procedure reported in Ref. 202, 50.0 mg (0.136 mmol, 1.0 eq) of methyl 2- (2-{[3-(methylsulfanyl)benzyl]amino}pyrimidin-4-yl)pyridine-4-carboxylate 93aj were suspended

218 7.2. Chemical Synthesis

in a mixture of 1.0 mL methanol and 1.0 mL tetrahydrofuran. 1.0 mL of a 1 mol/L aqueous lithium hydroxide solution (1.00 mmol, 7.4 eq) were added and the mixture heated gently to 40◦C. Heating was stopped after 2.5 hours and stirring continued overnight. After a total of 16 hours, the mixture was diluted with 10 mL 1 mol/L NaOH and extracted with 2x 10 mL ethyl acetate. The organic phase was extracted again with 10 mL NaOH. The combined aqueous phases were treated with 2 mol/L hydrochloric acid until a white precipitate formed (pH 3), which was ﬁltered oﬀ, washed with 2 mol/L HCl, and dried in vacuo yielding the title product as a yellow crystalline substance. Yield: 47.9 mg (0.136 mmol, 99 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 9.98 (s, 1H), 8.92 (d, J = 4.5 Hz, 1H), 8.81 (s, 1H), 8.52 (br s, 2H), 7.98 (d, J = 3.9 Hz, 1H), 7.61 (s, 1H), 7.34–7.22 (m, 2H), 7.23–7.14 (m, 1H), 7.12 (d, J = 7.5 Hz, 1H), 4.62 (s, 2H), 2.43 (s, 3H). m + APCI-MS positive (MeOH): /z = 353.1070 (100%, [M+H] ). Calcd. for [C18H16N4O2S+ H]+: 353.1067. m − APCI-MS negative (MeOH): /z = 351.0922 (100%, [M−H] ). Calcd. for [C18H16N4O2S− H]−: 351.0921.

HPLC: tR = 18.0 min, 94% purity.

2-{2-[(3,4-dihydro-2H -chromen-6-ylmethyl)amino]pyrimidin-4-yl}pyridine-4-carboxylic acid (86ag).

As a variation of the procedure reported in Ref. 202, 50.0 mg (0.133 mmol, 1.0 eq) of methyl 2-{2-[(3,4-dihydro-2H -chromen-6-ylmethyl)amino]pyrimidin-4-yl}pyridine-4-carboxylate 93ag were suspended in a mixture of 600 µL methanol and 600 µL tetrahydrofuran. 300 µL of a 1 mol/L aqueous lithium hydroxide solution (0.300 mmol, 2.3 eq) were added and the mixture heated gently to 45◦C. The next day, another 1 mL 1 mol/L LiOH solution was added and stirring continued at 45◦C for another few hours. After cooling to room temperature, the reaction mixture was acidified by the dropwise addition of 2 mol/L hydrochloric acid, resulting in the formation of a white solid, whose precipitation was completed over ice. The solid was filtered off, washed with more aqueous 2 mol/L HCl and methanol, and dried in vacuo yielding the title product as yellow crystals.

219 Chapter 7. Experimental Section

Yield: 43.8 mg (0.121 mmol, 91 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.93 (d, J = 4.9 Hz, 1H), 8.83 (s, 1H), 8.65–8.41 (m, 2H), 7.99 (d, J = 3.8 Hz, 1H), 7.62 (d, J = 5.1 Hz, 1H), 7.12 (s, 2H), 6.67 (d, J = 8.9 Hz, 1H), 4.91 (br s, 1H), 4.52 (s, 2H), 4.07 (t, J = 4.9 Hz, 2H), 2.70 (t, J = 6.3 Hz, 2H), 1.91–1.83 (m, 2H). m + + ESI-MS positive (MeOH): /z = 363.1453 (100%, [M+H] ). Calcd. for [C20H18N4O3+H] : 363.1452. m − ESI-MS negative (MeOH): /z = 361.1310 (100%, [M − H] ). Calcd. for [C20H18N4O3 − H]−: 361.1306.

HPLC: tR = 18.22 min, 96% purity.

7.2.3.5. Transesteriﬁcation

N -Morpholinoethyl 2-{2-[(3,4-dihydro-2H -chromen-6-ylmethyl)amino]pyrimidin- 4-yl}pyridine-4-carboxylate (108ag).

25.0 mg (0.0664 mmol, 1.0 eq) of methyl 2-{2-[(3,4-dihydro-2H -chromen-6-ylmethyl)amino]- pyrimidin-4-yl}pyridine-4-carboxylate 93ag were suspended in 1.61 mL (1.74 g, 13.3 mmol, 200 eq) 4-(2-hydroxyethyl)morpholine and 8.0 µL of 1 mol/L sodium hydroxide solution (0.008 mmol, 0.12 eq) were added. The mixture was heated to 70◦C, resulting in a clear orange solution that was stirred at 70◦C overnight. The next day, the mixture was diluted with 15 mL water and extracted with 3x 10 mL ethyl acetate. The combined organic phases were washed with 2x 15 mL brine until colorless, dried over Na2SO4, ﬁltered, and freed from all volatiles in vacuo, leaving behind a pale yellow oil that was further dried in vacuo and identiﬁed as the title product. Yield: 31.0 mg (0.0652 mmol, 98 %) 1 H-NMR (400 MHz, DMSO-d 6, TMS, ppm): δ = 8.91 (d, J = 4.9 Hz, 1H), 8.81 (s, 1H), 8.47 (d, J = 4.9 Hz, 1H), 7.95 (dd, J = 4.9 Hz, 1.5 Hz, 1H), 7.91 (m, 1H), 7.49 (d, J = 5.0 Hz, 1H), 7.09 (br s, 2H), 6.65 (d, J = 8.4 Hz, 1H), 4.52–4.36 (m, 4H), 4.12–4.04 (m, 2H), 3.56–3.52 (m, 5H), 3.51–3.46 (m, 1H), 2.76–2.64 (m, 4H), 2.39–2.35 (m, 2H), 1.91–1.81 (m, 2H).

220 7.2. Chemical Synthesis

13 1 C{ H}-NMR (101 MHz, DMSO-d 6, TMS, ppm): δ = 164.5, 162.4, 155.3, 153.4, 150.8, 138.3, 131.6, 129.5, 128.9, 126.5, 124.9, 124.1, 121.9, 120.3, 119.5, 116.0, 66.2, 65.8, 62.9, 56.4, 53.4, 43.7, 24.3, 21.9. + APCI-MS positive (MeOH): m/z = 476.22885 (30%, [M + H] ), 377.16064 (100%, methyl + ester from standing in methanol solution). Calcd. for [C26H29N5O4 + H] : 476.22923. HPLC: tR = 16.92 min, 94% purity.

221

References

[1] M. R. Stratton, P. J. Campbell, P. A. Futreal, The cancer genome. Nature 458(7239), (2009), 719–724.

[2] R. H. Bradbury (Ed.), Cancer, vol. 1 of Topics in Medicinal Chemistry, Springer, Berlin, 2007.

[3] E. S. Lander, L. M. Linton, B. Birren, C. Nusbaum, M. C. Zody, J. Baldwin, K. Devon, K. Dewar, M. Doyle, W. FitzHugh, R. Funke, D. Gage, K. Harris, A. Heaford, J. How- land, et al., Initial sequencing and analysis of the human genome. Nature 409(6822), (2001), 860–921.

[4] K. A. Lipinski, L. J. Barber, M. N. Davies, M. Ashenden, A. Sottoriva, M. Gerlinger, Cancer Evolution and the Limits of Predictability in Precision Cancer Medicine. Trends Cancer 2(1), (2016), 49–63.

[5] C. D. Allis, T. Jenuwein, D. Reinberg (Eds.), Epigenetics, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 2007.

[6] W. Sippl, M. Jung (Eds.), Epigenetic Targets in Drug Discovery, vol. 42 of Methods and Principles in Medicinal Chemistry, Wiley-VCH, Weinheim, 2009.

[7] D. K. Morgan, E. Whitelaw, The case for transgenerational epigenetic inheritance in humans. Mamm. Genome 19(6), (2008), 394–397.

[8] K. Siklenka, S. Erkek, M. Godmann, R. Lambrot, S. McGraw, C. Laﬂeur, T. Cohen, J. Xia, M. Suderman, M. Hallett, J. Trasler, A. H. F. M. Peters, S. Kimmins, Disruption of histone methylation in developing sperm impairs oﬀspring health transgenerationally. Science 350(6261), (2015), aab2006.

[9] M. Xu, W. Wang, S. Chen, B. Zhu, A model for mitotic inheritance of histone lysine methylation. EMBO Rep. 13(1), (2012), 60–67.

[10] P. N. C. B. Audergon, S. Catania, A. Kagansky, P. Tong, M. Shukla, A. L. Pidoux, R. C. Allshire, Restricted epigenetic inheritance of H3K9 methylation. Science 348(6230), (2015), 132–135.

[11] D. J. Voet, J. G. Voet, C. W. Pratt, Principles of Biochemistry, John Wiley & Sons, Hoboken, NJ, 3rd ISE edn., 2008.

[12] S. B. Baylin, K. E. Schuebel, The epigenomic era opens. Nature 448(7153), (2007), 548–549.

223 References

[13] Wikipedia, Nucleosome. https://en.wikipedia.org/wiki/Nucleosome (accessed Feb 8, 2016). [14] J. M. Harp, B. L. Hanson, D. E. Timm, G. J. Bunick, Asymmetries in the nucleosome core particle at 2.5 A˚ resolution. Acta Crystallogr., Sect. D: Biol. Crystallogr. 56(12), (2000), 1513–1534. [15] C. B. Yoo, P. A. Jones, Epigenetic therapy of cancer: past, present and future. Nat. Rev. Drug Discov. 5(1), (2006), 37–50. [16] M. Tahiliani, K. P. Koh, Y. Shen, W. A. Pastor, H. Bandukwala, Y. Brudno, S. Agar- wal, L. M. Iyer, D. R. Liu, L. Aravind, A. Rao, Conversion of 5-methylcytosine to 5- hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science 324(5929), (2009), 930–935. [17] L. Tan, Y. G. Shi, Tet family proteins and 5-hydroxymethylcytosine in development and disease. Development 139(11), (2012), 1895–1902. [18] A. Maiti, A. C. Drohat, Thymine DNA Glycosylase Can Rapidly Excise 5-Formylcytosine and 5-Carboxylcytosine. Potential Implications for Active Demethylation of CpG Sites. J. Biol. Chem. 286(41), (2011), 35334–35338. [19] X. Hu, L. Zhang, S.-Q. Mao, Z. Li, J. Chen, R.-R. Zhang, H.-P. Wu, J. Gao, F. Guo, W. Liu, G.-F. Xu, H.-Q. Dai, Y. G. Shi, X. Li, B. Hu, et al., Tet and TDG Mediate DNA Demethylation Essential for Mesenchymal-to-Epithelial Transition in Somatic Cell Reprogramming. Cell Stem Cell 14(4), (2014), 512–522. [20] T. Pfaﬀeneder, F. Spada, M. Wagner, C. Brandmayr, S. K. Laube, D. Eisen, M. Truss, J. Steinbacher, B. Hackner, O. Kotljarova, D. Schuermann, S. Michalakis, O. Kosmatchev, S. Schiesser, B. Steigenberger, et al., Tet oxidizes thymine to 5- hydroxymethyluracil in mouse embryonic stem cell DNA. Nat. Chem. Biol. 10(7), (2014), 574–581. [21] T. Kouzarides, Chromatin Modiﬁcations and their Function. Cell 128(4), (2007), 693– 705. [22] M. Biel, V. Wascholowski, A. Giannis, Epigenetics – An Epicenter of Gene Regulation: Histones and Histone-Modifying Enzymes. Angew. Chem. Int. Ed. 44(21), (2005), 3186– 3216. [23] K. Keller, M. Jung, Lesezeichen im Buch des Lebens. Nachr. Chem. 59(9), (2011), 822– 827. [24] M. Rodr´ıguez-Paredes, M. Esteller, Cancer epigenetics reaches mainstream oncology. Nat. Med. 17(3), (2011), 330–339. [25] J. Fuhrmann, P. R. Thompson, Protein Arginine Methylation and Citrullination in Epi- genetic Regulation. ACS Chem. Biol. 11(3), (2016), 654–668.

224 References

[26] C. H. Arrowsmith, C. Bountra, P. V. Fish, K. Lee, M. Schapira, Epigenetic protein families – a new frontier for drug discovery. Nat. Rev. Drug Discov. 11(5), (2012), 384– 400. [27] T. Wagner, D. Robaa, W. Sippl, M. Jung, Mind the Methyl: Methyllysine Binding Proteins in Epigenetic Regulation. ChemMedChem 9(3), (2014), 466–483. [28] T. Wagner, D. Robaa, W. Sippl, M. Jung, Epigenetic Readers Interpreting the Lysine Methylome – Biological Roles and Drug Discovery, in: G. Egger, P. Arimondo (Eds.), Drug Discovery in Cancer Epigenetics, 273–304, Academic Press, Oxford, 2016. [29] T. Jenuwein, C. D. Allis, Translating the histone code. Science 293(5532), (2001), 1074– 1080. [30] S. Lanouette, V. Mongeon, D. Figeys, J.-F. Couture, The functional diversity of protein lysine methylation. Mol. Syst. Biol. 10(4), (2014), 724. [31] E. Dimitrova, A. H. Turberﬁeld, R. J. Klose, Histone demethylases in chromatin biology and beyond. EMBO Rep. 16(12), (2015), 1620–1639. [32] R. Lillico, M. Gomez Sobral, N. Stesco, T. M. Lakowski, HDAC inhibitors induce global changes in histone lysine and arginine methylation and alter expression of lysine demethylases. J. Proteomics 133, (2016), 125–133. [33] S. L. Berger, The complex language of chromatin regulation during transcription. Nature 447(7143), (2007), 407–412. [34] Z. Su, J. M. Denu, Reading the Combinatorial Histone Language. ACS Chem. Biol. 11(3), (2016), 564–574. [35] P. Gut, E. Verdin, The nexus of chromatin regulation and intermediary metabolism. Nature 502(7472), (2013), 489–498. [36] A. Salminen, A. Kauppinen, M. Hiltunen, K. Kaarniranta, Krebs cycle intermediates regulate DNA and histone methylation: Epigenetic impact on the aging process. Ageing Res. Rev. 16, (2014), 45–65. [37] P. Chi, C. D. Allis, G. G. Wang, Covalent histone modiﬁcations – miswritten, misinter- preted and mis-erased in human cancers. Nat. Rev. Cancer 10(7), (2010), 457–469. [38] A. E. Handel, G. C. Ebers, S. V. Ramagopalan, Epigenetics: molecular mechanisms and implications for disease. Trends Mol. Med. 16(1), (2010), 7–16. [39] A. Huston, C. H. Arrowsmith, S. Knapp, M. Schapira, Probing the epigenome. Nat. Chem. Biol. 11(8), (2015), 542–545. [40] J. S. Butler, E. Koutelou, A. C. Schibler, S. Y. R. Dent, Histone-modifying enzymes: regulators of developmental decisions and drivers of human disease. Epigenomics 4(2), (2012), 163–177.

225 References

[41] G. Egger, P. Arimondo (Eds.), Drug Discovery in Cancer Epigenetics, Academic Press, Oxford, 2016.

[42] P. Jones, Histone Deacetylase Inhibitors, in: W. Sippl, M. Jung (Eds.), Epigenetic Tar- gets in Drug Discovery, vol. 42 of Methods and Principles in Medicinal Chemistry, 185– 223, Wiley-VCH, Weinheim, 2009.

[43] R. P. Simon, D. Robaa, Z. Alhalabi, W. Sippl, M. Jung, KATching-Up on Small Molecule Modulators of Lysine Acetyltransferases. J. Med. Chem. 59(4), (2016), 1249–1270.

[44] T. Wagner, M. Jung, New lysine methyltransferase drug targets in cancer. Nat. Bio- technol. 30(7), (2012), 622–623.

[45] R. P. Clausen, M. T. Pedersen, K. Helin, Histone Demethylases, in: W. Sippl, M. Jung (Eds.), Epigenetic Targets in Drug Discovery, vol. 42 of Methods and Principles in Medic- inal Chemistry, 269–290, Wiley-VCH, Weinheim, 2009.

[46] A. Nebbioso, V. Carafa, R. Benedetti, L. Altucci, Trials with ‘epigenetic’ drugs: An update. Mol. Oncol. 6(6), (2012), 657–682.

[47] L. Morera, M. L¨ubbert, M. Jung, Targeting histone methyltransferases and demethylases in clinical trials for cancer therapy. Clin. Epigenet. 8, (2016), 57.

[48] J. C. Rice, C. D. Allis, Histone methylation versus histone acetylation: new insights into epigenetic regulation. Curr. Opin. Cell Biol. 13(3), (2001), 263–273.

[49] Y. Shi, F. Lan, C. Matson, P. Mulligan, J. R. Whetstine, P. A. Cole, R. A. Casero, Histone Demethylation Mediated by the Nuclear Amine Oxidase Homolog LSD1. Cell 119(7), (2004), 941–953.

[50] M. Yang, J. C. Culhane, L. M. Szewczuk, C. B. Gocke, C. A. Brautigam, D. R. Tomchick, M. Machius, P. A. Cole, H. Yu, Structural basis of histone demethylation by LSD1 revealed by suicide inactivation. Nat. Struct. Mol. Biol. 14(6), (2007), 535–539.

[51] E. Metzger, M. Wissmann, N. Yin, J. M. Müller,R. Schneider, A. H. F. M. Peters, T. Günther, R. Buettner, R. Schüle,LSD1 demethylates repressive histone marks to promote androgen-receptor-dependent transcription. Nature 437(7057), (2005), 436–439.

[52] I. Hoﬀmann, M. Roatsch, M. L. Schmitt, L. Carlino, M. Pippel, W. Sippl, M. Jung, The role of histone demethylases in cancer therapy. Mol. Oncol. 6(6), (2012), 683–703.

[53] Y.-i. Tsukada, J. Fang, H. Erdjument-Bromage, M. E. Warren, C. H. Borchers, P. Tempst, Y. Zhang, Histone demethylation by a family of JmjC domain-containing proteins. Nature 439(7078), (2006), 811–816.

[54] J. R. Whetstine, A. Nottke, F. Lan, M. Huarte, S. Smolikov, Z. Chen, E. Spooner, E. Li, G. Zhang, M. Colaiacovo, Y. Shi, Reversal of Histone Lysine Trimethylation by the JMJD2 Family of Histone Demethylases. Cell 125(3), (2006), 467–481.

226 References

[55] C. Johansson, A. Tumber, K. Che, P. Cain, R. Nowak, C. Gileadi, U. Oppermann, The roles of Jumonji-type oxygenases in human disease. Epigenomics 6(1), (2014), 89–120.

[56] S. Y. Park, J.-W. Park, Y.-S. Chun, Jumonji histone demethylases as emerging therapeutic targets. Pharmacol. Res. 105, (2016), 146–151.

[57] R. J. Klose, E. M. Kallin, Y. Zhang, JmjC-domain-containing proteins and histone demethylation. Nat. Rev. Genet. 7(9), (2006), 715–727.

[58] S. M. Kooistra, K. Helin, Molecular mechanisms and potential functions of histone demethylases. Nat. Rev. Mol. Cell Biol. 13(5), (2012), 297–311.

[59] C. D. Allis, S. L. Berger, J. Cote, S. Dent, T. Jenuwein, T. Kouzarides, L. Pillus, D. Rein- berg, Y. Shi, R. Shiekhattar, A. Shilatifard, J. Workman, Y. Zhang, New Nomenclature for Chromatin-Modifying Enzymes. Cell 131(4), (2007), 633–636.

[60] T. Lu, M. W. Jackson, B. Wang, M. Yang, M. R. Chance, M. Miyagi, A. V. Gudkov, G. R. Stark, Regulation of NF-κB by NSD1/FBXL11-dependent reversible lysine methylation of p65. Proc. Natl. Acad. Sci. U. S. A. 107(1), (2010), 46–51.

[61] P. M. Clissold, C. P. Ponting, JmjC: cupin metalloenzyme-like domains in jumonji, hair- less and phospholipase A2β. Trends Biochem. Sci. 26(1), (2001), 7–9. [62] R. P. Hausinger, Fe(II)/α-Ketoglutarate-Dependent Hydroxylases and Related Enzymes. Crit. Rev. Biochem. Mol. Biol. 39(1), (2004), 21–68.

[63] N. Mosammaparast, Y. Shi, Reversal of Histone Methylation: Biochemical and Molecular Mechanisms of Histone Demethylases. Annu. Rev. Biochem. 79, (2010), 155–179.

[64] S. S. Ng, K. L. Kavanagh, M. A. McDonough, D. Butler, E. S. Pilka, B. M. R. Lienard, J. E. Bray, P. Savitsky, O. Gileadi, F. von Delft, N. R. Rose, J. Oﬀer, J. C. Scheinost, T. Borowski, M. Sundstrom, et al., Crystal structures of histone demethylase JMJD2A reveal basis for substrate speciﬁcity. Nature 448(7149), (2007), 87–91.

[65] Y. Shi, J. R. Whetstine, Dynamic Regulation of Histone Lysine Methylation by Demethy- lases. Mol. Cell 25(1), (2007), 1–14.

[66] W. Aik, M. A. McDonough, A. Thalhammer, R. Chowdhury, C. J. Schoﬁeld, Role of the jelly-roll fold in substrate binding by 2-oxoglutarate oxygenases. Curr. Opin. Struct. Biol. 22(6), (2012), 691–700.

[67] W. A. Cortopassi, R. Simion, C. E. Honsby, T. C. C. Fran¸ca,R. S. Paton, Dioxygen Binding in the Active Site of Histone Demethylase JMJD2A and the Role of the Protein Environment. Chem. Eur. J. 21(52), (2015), 18983–18992.

[68] B. Lohse, C. Helgstrand, J. B. L. Kristensen, U. Leurs, P. A. C. Cloos, J. L. Kristensen, R. P. Clausen, Posttranslational Modiﬁcations of the Histone 3 Tail and Their Impact on the Activity of Histone Lysine Demethylases In Vitro. PLoS One 8(7), (2013), e67653.

227 References

[69] L. Hillringhaus, W. W. Yue, N. R. Rose, S. S. Ng, C. Gileadi, C. Loenarz, S. H. Bello, J. E. Bray, C. J. Schofield, U. Oppermann, Structural and Evolutionary Basis for the Dual Substrate Selectivity of Human KDM4 Histone Demethylase Family. J. Biol. Chem. 286(48), (2011), 41616–41625. [70] H. Hou, H. Yu, Structural insights into histone lysine demethylation. Curr. Opin. Struct. Biol. 20(6), (2010), 739–748. [71] S. Krishnan, S. Horowitz, R. C. Trievel, Structure and Function of Histone H3 Lysine 9 Methyltransferases and Demethylases. ChemBioChem 12(2), (2011), 254–263. [72] J.-F. Couture, E. Collazo, P. A. Ortiz-Tello, J. S. Brunzelle, R. C. Trievel, Specificity and mechanism of JMJD2A, a trimethyllysine-specific histone demethylase. Nat. Struct. Mol. Biol. 14(8), (2007), 689–695. [73] R. J. Klose, K. Yamane, Y. Bae, D. Zhang, H. Erdjument-Bromage, P. Tempst, J. Wong, Y. Zhang, The transcriptional repressor JHDM3A demethylates trimethyl histone H3 lysine 9 and lysine 36. Nature 442(7100), (2006), 312–316. [74] K. Agger, J. Christensen, P. A. C. Cloos, K. Helin, The emerging functions of histone demethylases. Curr. Opin. Genet. Dev. 18(2), (2008), 159–168. [75] J. W. Højfeldt, K. Agger, K. Helin, Histone lysine demethylases as targets for anticancer therapy. Nat. Rev. Drug Discov. 12(12), (2013), 917–930. [76] D. Rotili, A. Mai, Targeting Histone Demethylases: A New Avenue for the Fight against Cancer. Genes Cancer 2(6), (2011), 663–679. [77] W. L. Berry, R. Janknecht, KDM4/JMJD2 Histone Demethylases: Epigenetic Regulators in Cancer Cells. Cancer Res. 73(10), (2013), 2936–2942. [78] I. Hoffmann, Zelluläre Charakterisierung neuer Hemmstoffe der Jumonji C-Domäne- enthaltenden Demethylasen, Dissertation. Albert-Ludwigs-Universität, Freiburg im Breisgau, 2015. [79] Y. Liang, J. L. Vogel, J. H. Arbuckle, G. Rai, A. Jadhav, A. Simeonov, D. J. Maloney, T. M. Kristie, Targeting the JMJD2 Histone Demethylases to Epigenetically Control Herpesvirus Infection and Reactivation from Latency. Sci. Transl. Med. 5(167), (2013), 167RA5. [80] P.-C. Chang, L. D. Fitzgerald, D. A. Hsia, Y. Izumiya, C.-Y. Wu, W.-P. Hsieh, S.- F. Lin, M. Campbell, K. S. Lam, P. A. Luciw, C. G. Tepper, H.-J. Kung, Histone Demethylase JMJD2A Regulates Kaposi’s Sarcoma-Associated Herpesvirus Replication and Is Targeted by a Viral Transcriptional Factor. J. Virol. 85(7), (2011), 3283–3293. [81] Q.-J. Zhang, H.-Z. Chen, L. Wang, D.-P. Liu, J. A. Hill, Z.-P. Liu, The histone trimethyllysine demethylase JMJD2A promotes cardiac hypertrophy in response to hypertrophic stimuli in mice. J. Clin. Invest. 121(6), (2011), 2447–2456.

228 References

[82] T. Inagaki, M. Tachibana, K. Magoori, H. Kudo, T. Tanaka, M. Okamura, M. Naito, T. Kodama, Y. Shinkai, J. Sakai, Obesity and metabolic syndrome in histone demethylase JHDM2a-deﬁcient mice. Genes Cells 14(8), (2009), 991–1001.

[83] F. de Santa, M. G. Totaro, E. Prosperini, S. Notarbartolo, G. Testa, G. Natoli, The histone H3 lysine-27 demethylase Jmjd3 links inﬂammation to inhibition of polycomb- mediated gene silencing. Cell 130(6), (2007), 1083–1094.

[84] A. Salminen, A. Kauppinen, K. Kaarniranta, 2-Oxoglutarate-dependent dioxygenases are sensors of energy metabolism, oxygen availability, and iron homeostasis: potential role in the regulation of aging process. Cell. Mol. Life Sci. 72(20), (2015), 3897–3914.

[85] A. Spannhoﬀ, A.-T. Hauser, R. Heinke, W. Sippl, M. Jung, The Emerging Therapeu- tic Potential of Histone Methyltransferase and Demethylase Inhibitors. ChemMedChem 4(10), (2009), 1568–1582.

[86] C. C. Thinnes, K. S. England, A. Kawamura, R. Chowdhury, C. J. Schoﬁeld, R. J. Hopkinson, Targeting histone lysine demethylases – Progress, challenges, and the future. Biochim. Biophys. Acta, Gene Regul. Mech. 1839(12), (2014), 1416–1432.

[87] P. B. Rasmussen, P. Staller, The KDM5 family of histone demethylases as targets in oncology drug discovery. Epigenomics 6(3), (2014), 277–286.

[88] J. McGrath, P. Trojer, Targeting histone lysine methylation in cancer. Pharmacol. Ther. 150, (2015), 1–22.

[89] Y. Itoh, T. Suzuki, N. Miyata, Small-molecular modulators of cancer-associated epigenetic mechanisms. Mol. BioSyst. 9(5), (2013), 873–896.

[90] W. L. Berry, S. Shin, S. A. Lightfoot, R. Janknecht, Oncogenic features of the JMJD2A histone demethylase in breast cancer. Int. J. Oncol. 41(5), (2012), 1701–1706.

[91] N. Patani, W. G. Jiang, R. F. Newbold, K. Mokbel, Histone-modiﬁer gene expression proﬁles are associated with pathological and clinical outcomes in human breast cancer. Anticancer Res. 31(12), (2011), 4115–4125.

[92] S. Shin, R. Janknecht, Activation of androgen receptor by histone demethylases JMJD2A and JMJD2D. Biochem. Biophys. Res. Commun. 359(3), (2007), 742–746.

[93] K. Yamane, C. Toumazou, Y.-i. Tsukada, H. Erdjument-Bromage, P. Tempst, J. M. Wong, Y. Zhang, JHDM2A, a JmjC-Containing H3K9 Demethylase, Facilitates Tran- scription Activation by Androgen Receptor. Cell 125(3), (2006), 483–495.

[94] M. Wissmann, N. Yin, J. M. Müller,H. Greschik, B. D. Fodor, T. Jenuwein, C. Vogler, R. Schneider, T. Günther, R. Buettner, E. Metzger, R. Schüle,Cooperative demethylation by JMJD2C and LSD1 promotes androgen receptor-dependent gene expression. Nat. Cell Biol. 9(3), (2007), 347–353.

229 References

[95] F. A. Mallette, S. Richard, JMJD2A Promotes Cellular Transformation by Blocking Cellular Senescence Through Transcriptional Repression of the Tumor Suppressor CHD5. Cell Rep. 2(5), (2012), 1233–1243.

[96] M. Kogure, M. Takawa, H.-S. Cho, G. Toyokawa, K. Hayashi, T. Tsunoda, T. Kobayashi, Y. Daigo, M. Sugiyama, Y. Atomi, Y. Nakamura, R. Hamamoto, Deregulation of the histone demethylase JMJD2A is involved in human carcinogenesis through regulation of the G1/S transition. Cancer Lett. 336(1), (2013), 76–84. [97] T.-D. Kim, S. Shin, W. L. Berry, S. Oh, R. Janknecht, The JMJD2A demethylase regulates apoptosis and proliferation in colon cancer cells. J. Cell. Biochem. 113(4), (2012), 1368–1376.

[98] E. C. Kauﬀman, B. D. Robinson, M. J. Downes, L. G. Powell, M. M. Lee, D. S. Scherr, L. J. Gudas, N. P. Mongan, Role of androgen receptor and associated lysine-demethylase coregulators, LSD1 and JMJD2A, in localized and advanced human bladder cancer. Mol. Carcinog. 50(12), (2011), 931–944.

[99] L. Guerra-Calderas, R. González-Barrios,L. A. Herrera, D. Cantúde León,E. Soto- Reyes, The role of the histone demethylase KDM4A in cancer. Cancer Genet. 208(5), (2015), 215–224.

[100] P. A. C. Cloos, J. Christensen, K. Agger, A. Maiolica, J. Rappsilber, T. Antal, K. H. Hansen, K. Helin, The putative oncogene GASC1 demethylates tri- and dimethylated lysine 9 on histone H3. Nature 442(7100), (2006), 307–311.

[101] B. Berdel, K. Nieminen, Y. Soini, M. Tengstr¨om,M. Malinen, V.-M. Kosma, J. J. Palvimo, A. Mannermaa, Histone demethylase GASC1 – a potential prognostic and predictive marker in invasive breast cancer. BMC Cancer 12, (2012), 516.

[102] W. Luo, R. Chang, J. Zhong, A. Pandey, G. L. Semenza, Histone demethylase JMJD2C is a coactivator for hypoxia-inducible factor 1 that is required for breast cancer progression. Proc. Natl. Acad. Sci. U. S. A. 109(49), (2012), E3367–E3376.

[103] T.-D. Kim, S. Oh, S. Shin, R. Janknecht, Regulation of Tumor Suppressor p53 and HCT116 Cell Physiology by Histone Demethylase JMJD2D/KDM4D. PLoS One 7(4), (2012), e34618.

[104] J. S. Burchﬁeld, Q. Li, H. Y. Wang, R.-F. Wang, JMJD3 as an epigenetic regulator in development and disease. Int. J. Biochem. Cell Biol. 67, (2015), 148–157.

[105] P. Ntziachristos, A. Tsirigos, G. G. Welstead, T. Trimarchi, S. Bakogianni, L. Xu, E. Loizou, L. Holmfeldt, A. Strikoudis, B. King, J. Mullenders, J. Becksfort, J. Nedjic, E. Paietta, M. S. Tallman, et al., Contrasting roles of histone 3 lysine 27 demethylases in acute lymphoblastic leukaemia. Nature 514(7523), (2014), 513–517.

230 References

[106] A.-T. Hauser, M. Roatsch, J. Schulz-Fincke, D. Robaa, W. Sippl, M. Jung, Discovery of Histone Demethylase Inhibitors, in: Y. G. Zheng (Ed.), Epigenetic Technological Applications, 397–424, Academic Press, Amsterdam, 2015. [107] M. Gale, Q. Yan, High-throughput screening to identify inhibitors of lysine demethylases. Epigenomics 7(1), (2015), 57–65. [108] N. J. Martinez, A. Simeonov, Cell-based assays to support the profiling of small molecules with histone methyltransferase and demethylase modulatory activity. Drug Discovery Today: Technol. 18, (2015), 9–17. [109] C. Mannironi, M. Proietto, F. Bufalieri, E. Cundari, A. Alagia, S. Danovska, T. Rinaldi, V. Famiglini, A. Coluccia, G. La Regina, R. Silvestri, R. Negri, An High-Throughput In Vivo Screening System to Select H3K4-Specific Histone Demethylase Inhibitors. PLoS One 9(1), (2014), e86002. [110] A. Mulji, C. Haslam, F. Brown, R. Randle, B. Karamshi, J. Smith, R. Eagle, J. Munoz- Muriedas, J. Taylor, A. Sheikh, A. Bridges, K. Gill, R. Jepras, P. Smee, M. Barker, et al., Configuration of a High-Content Imaging Platform for Hit Identification and Pharmaco- logical Assessment of JMJD3 Demethylase Enzyme Inhibitors. J. Biomol. Screen. 17(1), (2012), 108–120. [111] A.-T. Hauser, E.-M. Bissinger, E. Metzger, A. Repenning, U.-M. Bauer, A. Mai, R. Schüle,M. Jung, Screening assays for epigenetic targets using native histones as substrates. J. Biomol. Screen. 17(1), (2012), 18–26. [112] M. Sakurai, N. R. Rose, L. Schultz, A. M. Quinn, A. Jadhav, S. S. Ng, U. Oppermann, C. J. Schofield, A. Simeonov, A miniaturized screen for inhibitors of Jumonji histone demethylases. Mol. BioSyst. 6(2), (2010), 357–364. [113] O. N. F. King, X. S. Li, M. Sakurai, A. Kawamura, N. R. Rose, S. S. Ng, A. M. Quinn, G. Rai, B. T. Mott, P. Beswick, R. J. Klose, U. Oppermann, A. Jadhav, T. D. Heightman, D. J. Maloney, et al., Quantitative High-Throughput Screening Identifies 8-Hydroxyquinolines as Cell-Active Histone Demethylase Inhibitors. PLoS One 5(11), (2010), e15535. [114] Q. Li, P. Sritharathikhun, S. Motomizu, Development of Novel Reagent for Hantzsch Re- action for the Determination of Formaldehyde by Spectrophotometry and Fluorometry. Anal. Sci. 23(4), (2007), 413–417. [115] L.-J. Liu, L. Lu, H.-J. Zhong, B. He, D. W. J. Kwong, D.-L. Ma, C.-H. Leung, An Irid- ium(III) Complex Inhibits JMJD2 Activities and Acts as a Potential Epigenetic Modu- lator. J. Med. Chem. 58(16), (2015), 6697–6703. [116] Y. Wang, X. Deng, J. Liu, H. Tang, J. Jiang, Surface enhanced Raman scattering based sensitive detection of histone demethylase activity using a formaldehyde-selective reactive probe. Chem. Commun. 49(76), (2013), 8489–8491.

231 References

[117] T. F. Brewer, C. J. Chang, An Aza-Cope Reactivity-Based Fluorescent Probe for Imaging Formaldehyde in Living Cells. J. Am. Chem. Soc. 137(34), (2015), 10886–10889.

[118] A. Roth, H. Li, C. Anorma, J. Chan, A Reaction-Based Fluorescent Probe for Imaging of Formaldehyde in Living Cells. J. Am. Chem. Soc. 137(34), (2015), 10890–10893.

[119] T. A. Egelhofer, A. Minoda, S. Klugman, K. Lee, P. Kolasinska-Zwierz, A. A. Alek- seyenko, M.-S. Cheung, D. S. Day, S. Gadel, A. A. Gorchakov, T. Gu, P. V. Kharchenko, S. Kuan, I. Latorre, D. Linder-Basso, et al., An assessment of histone-modiﬁcation antibody quality. Nat. Struct. Mol. Biol. 18(1), (2011), 91–93.

[120] A. L. Nielsen, L. H. Kristensen, K. B. Stephansen, J. B. L. Kristensen, C. Helgstrand, M. Lees, P. Cloos, K. Helin, M. Gajhede, L. Olsen, Identiﬁcation of catechols as histone- lysine demethylase inhibitors. FEBS Lett. 586(8), (2012), 1190–1194.

[121] L. Wang, J. Chang, D. Varghese, M. Dellinger, S. Kumar, A. M. Best, J. Ruiz, R. Bruick, S. Pe˜na-Llopis,J. Xu, D. J. Babinski, D. E. Frantz, R. A. Brekken, A. M. Quinn, A. Simeonov, et al., A small molecule modulates Jumonji histone demethylase activity and selectively inhibits cancer growth. Nat. Commun. 4, (2013), 2035.

[122] PerkinElmer, LANCE TR-FRET Epigenetics Tool Box. http://www.perkinelmer.com/Catalog/Category/ID/LANCE%20Ultra%20for %20Epigenetics%20Research (accessed Feb 12, 2016).

[123] N. Gauthier, M. Caron, L. Pedro, M. Arcand, J. Blouin, A. Labont´e,C. Normand, V. Paquet, A. Rodenbrock, M. Roy, N. Rouleau, L. Beaudet, J. Padr´os,R. Rodriguez- Suarez, Development of homogeneous nonradioactive methyltransferase and demethylase assays targeting histone H3 lysine 4. J. Biomol. Screen. 17(1), (2012), 49–58.

[124] X. Luo, Y. Liu, S. Kubicek, J. Myllyharju, A. Tumber, S. Ng, K. H. Che, J. Podoll, T. D. Heightman, U. Oppermann, S. L. Schreiber, X. Wang, A Selective Inhibitor and Probe of the Cellular Functions of Jumonji C Domain-Containing Histone Demethylases. J. Am. Chem. Soc. 133(24), (2011), 9451–9456.

[125] A. Kawamura, A. Tumber, N. R. Rose, O. N. F. King, M. Daniel, U. Oppermann, T. D. Heightman, C. Schoﬁeld, Development of homogeneous luminescence assays for histone demethylase catalysis and binding. Anal. Biochem. 404(1), (2010), 86–93.

[126] N. R. Rose, S. S. Ng, J. Mecinović,B. M. R. Liénard,S. H. Bello, Z. Sun, M. A. McDonough, U. Oppermann, C. J. Schofield, Inhibitor Scaffolds for 2-Oxoglutarate- Dependent Histone Lysine Demethylases. J. Med. Chem. 51(22), (2008), 7053–7056.

[127] S. T. Williams, L. J. Walport, R. J. Hopkinson, S. K. Madden, R. Chowdhury, C. J. Schoﬁeld, A. Kawamura, Studies on the catalytic domains of multiple JmjC oxygenases using peptide substrates. Epigenetics 9(12), (2014), 1596–1603.

232 References

[128] S. E. Hutchinson, M. V. Leveridge, M. L. Heathcote, P. Francis, L. Williams, M. Gee, J. Munoz-Muriedas, B. Leavens, A. Shillings, E. Jones, P. Homes, S. Baddeley, C.-w. Chung, A. Bridges, A. Argyrou, Enabling Lead Discovery for Histone Lysine Demethy- lases by High-Throughput RapidFire Mass Spectrometry. J. Biomol. Screen. 17(1), (2012), 39–48.

[129] M. Leveridge, R. Buxton, A. Argyrou, P. Francis, B. Leavens, A. West, M. Rees, P. Hard- wicke, A. Bridges, S. Ratcliﬀe, C.-w. Chung, Demonstrating Enhanced Throughput of RapidFire Mass Spectrometry through Multiplexing Using the JmjD2d Demethylase as a Model System. J. Biomol. Screen. 19(2), (2014), 278–286.

[130] N. R. Rose, E. C. Y. Woon, G. L. Kingham, O. N. F. King, J. Mecinovi´c, I. J. Clifton, S. S. Ng, J. Talib-Hardy, U. Oppermann, M. A. McDonough, C. J. Schoﬁeld, Selective inhibitors of the JMJD2 histone demethylases: combined nondenaturing mass spectrometric screening and crystallographic approaches. J. Med. Chem. 53(4), (2010), 1810– 1818.

[131] W. Xu, J. D. Podoll, X. Dong, A. Tumber, U. Oppermann, X. Wang, Quantitative Analysis of Histone Demethylase Probes Using Fluorescence Polarization. J. Med. Chem. 56(12), (2013), 5198–5202.

[132] W. Wang, L. J. Marholz, X. Wang, Novel Scaﬀolds of Cell-Active Histone Demethylase Inhibitors Identiﬁed from High-Throughput Screening. J. Biomol. Screen. 20(6), (2015), 821–827.

[133] I. K. H. Leung, M. Demetriades, A. P. Hardy, C. Lejeune, T. J. Smart, A. Szöllössi, A. Kawamura, C. J. Schofield, T. D. W. Claridge, Reporter Ligand NMR Screening Method for 2-Oxoglutarate Oxygenase Inhibitors. J. Med. Chem. 56(2), (2013), 547– 555.

[134] W. Yu, M. S. Eram, T. Hajian, A. Szykowska, N. Burgess-Brown, M. Vedadi, P. J. Brown, A scintillation proximity assay for histone demethylases. Anal. Biochem. 463, (2014), 54–60.

[135] B. Lohse, J. L. Kristensen, L. H. Kristensen, K. Agger, K. Helin, M. Gajhede, R. P. Clausen, Inhibitors of histone demethylases. Bioorg. Med. Chem. 19(12), (2011), 3625– 3636.

[136] T. Suzuki, N. Miyata, Lysine Demethylases Inhibitors. J. Med. Chem. 54(24), (2011), 8236–8250.

[137] Z. Wang, D. J. Patel, Small molecule epigenetic inhibitors targeted to histone lysine methyltransferases and demethylases. Q. Rev. Biophys. 46(4), (2013), 349–373.

[138] T. Maes, E. Carceller, J. Salas, A. Ortega, C. Buesa, Advances in the development of histone lysine demethylase inhibitors. Curr. Opin. Pharmacol. 23, (2015), 52–60.

233 References

[139] T. E. McAllister, K. S. England, R. J. Hopkinson, P. E. Brennan, A. Kawamura, C. J. Schoﬁeld, Recent Progress in Histone Demethylase Inhibitors. J. Med. Chem. 59(4), (2016), 1308–1329.

[140] C. J. Schoﬁeld, M. A. McDonough, N. R. Rose, A. Thalhammer, Histone lysine demethylase inhibitors. Patent WO(2010/043866), (2010), 1–138.

[141] K. Al-Qahtani, B. Jabeen, R. Sekirnik, N. Riaz, T. D. W. Claridge, C. J. Schoﬁeld, J. S. O. McCullagh, The broad spectrum 2-oxoglutarate oxygenase inhibitor N - oxalylglycine is present in rhubarb and spinach leaves. Phytochemistry 117, (2015), 456–461.

[142] S. Hamada, T.-D. Kim, T. Suzuki, Y. Itoh, H. Tsumoto, H. Nakagawa, R. Janknecht, N. Miyata, Synthesis and activity of N -oxalylglycine and its derivatives as Jumonji C-domain-containing histone lysine demethylase inhibitors. Bioorg. Med. Chem. Lett. 19(10), (2009), 2852–2855.

[143] W. Xu, H. Yang, Y. Liu, Y. Yang, P. Wang, S.-H. Kim, S. Ito, C. Yang, P. Wang, M.- T. Xiao, L.-x. Liu, W.-q. Jiang, J. Liu, J.-y. Zhang, B. Wang, et al., Oncometabolite 2-Hydroxyglutarate Is a Competitive Inhibitor of α-Ketoglutarate-Dependent Dioxyge- nases. Cancer Cell 19(1), (2011), 17–30.

[144] L. H. Kristensen, A. L. Nielsen, C. Helgstrand, M. Lees, P. Cloos, J. S. Kastrup, K. Helin, L. Olsen, M. Gajhede, Studies of H3K4me3 demethylation by KDM5B/Jarid1B/PLU1 reveals strong substrate recognition in vitro and identiﬁes 2,4-pyridine-dicarboxylic acid as an in vitro and in cell inhibitor. FEBS J. 279(11), (2012), 1905–1914.

[145] J. Sayegh, J. Cao, M. R. Zou, A. Morales, L. P. Blair, M. Norcia, D. Hoyer, A. J. Tackett, J. S. Merkel, Q. Yan, Identiﬁcation of Small Molecule Inhibitors of Jumonji AT-rich Interactive Domain 1B (JARID1B) Histone Demethylase by a Sensitive High Throughput Screen. J. Biol. Chem. 288(13), (2013), 9408–9417.

[146] A. Thalhammer, J. Mecinovi´c,C. Loenarz, A. Tumber, N. R. Rose, T. D. Heightman, C. J. Schoﬁeld, Inhibition of the histone demethylase JMJD2E by 3-substituted pyridine 2,4-dicarboxylates. Org. Biomol. Chem. 9(1), (2011), 127–135.

[147] M. Korczynska, D. D. Le, N. Younger, E. Gregori-Puigjan´e,A. Tumber, T. Krojer, S. Velupillai, C. Gileadi, R. P. Nowak, E. Iwasa, S. B. Pollock, I. Ortiz Torres, U. Opper- mann, B. K. Shoichet, D. G. Fujimori, Docking and Linking of Fragments To Discover Jumonji Histone Demethylase Inhibitors. J. Med. Chem. 59(4), (2016), 1580–1598.

[148] R. Schiller, G. Scozzafava, A. Tumber, J. R. Wickens, J. T. Bush, G. Rai, C. Lejeune, H. Choi, T.-L. Yeh, M. C. Chan, B. T. Mott, J. S. O. McCullagh, D. J. Maloney, C. J. Schoﬁeld, A. Kawamura, A Cell-Permeable Ester Derivative of the JmjC Histone Demethylase Inhibitor IOX1. ChemMedChem 9(3), (2014), 566–571.

234 References

[149] C. C. Thinnes, A. Tumber, C. Yapp, G. Scozzafava, T. Yeh, M. C. Chan, T. A. Tran, K. Hsu, H. Tarhonskaya, L. J. Walport, S. E. Wilkins, E. D. Martinez, S. Müller,C. W. Pugh, P. J. Ratcliffe, et al., Betti reaction enables efficient synthesis of 8-hydroxyquinoline inhibitors of 2-oxoglutarate oxygenases. Chem. Commun. 51(84), (2015), 15458–15461.

[150] L. Duan, G. Rai, C. Roggero, Q.-J. Zhang, Q. Wei, S. H. Ma, Y. Zhou, J. Santoyo, E. D. Martinez, G. Xiao, G. V. Raj, A. Jadhav, A. Simeonov, D. J. Maloney, J. Rizo, et al., KDM4/JMJD2 Histone Demethylase Inhibitors Block Prostate Tumor Growth by Suppressing the Expression of AR and BMYB-Regulated Genes. Chem. Biol. 22(9), (2015), 1185–1196.

[151] T. Feng, D. Li, H. Wang, J. Zhuang, F. Liu, Q. Bao, Y. Lei, W. Chen, X. Zhang, X. Xu, H. Sun, Q. You, X. Guo, Novel 5-carboxy-8-HQ based histone demethylase JMJD2A inhibitors: introduction of an additional carboxyl group at the C-2 position of quinoline. Eur. J. Med. Chem. 105, (2015), 145–155.

[152] D. Rotili, M. Altun, A. Kawamura, A. Wolf, R. Fischer, I. K. H. Leung, M. M. Mackeen, Y.-m. Tian, P. J. Ratcliﬀe, A. Mai, B. M. Kessler, C. J. Schoﬁeld, A Photoreactive Small- Molecule Probe for 2-Oxoglutarate Oxygenases. Chem. Biol. 18(5), (2011), 642–654.

[153] D. Rotili, S. Tomassi, M. Conte, R. Benedetti, M. Tortorici, G. Ciossani, S. Valente, B. Marrocco, D. Labella, E. Novellino, A. Mattevi, L. Altucci, A. Tumber, C. Yapp, O. N. F. King, et al., Pan-Histone Demethylase Inhibitors Simultaneously Targeting Jumonji C and Lysine-Speciﬁc Demethylases Display High Anticancer Activities. J. Med. Chem. 57(1), (2014), 42–55.

[154] S. Hamada, T. Suzuki, K. Mino, K. Koseki, F. Oehme, I. Flamme, H. Ozasa, Y. Itoh, D. Ogasawara, H. Komaarashi, A. Kato, H. Tsumoto, H. Nakagawa, M. Hasegawa, R. Sasaki, et al., Design, Synthesis, Enzyme-Inhibitory Activity, and Eﬀect on Human Cancer Cells of a Novel Series of Jumonji Domain-Containing Protein 2 Histone Demethy- lase Inhibitors. J. Med. Chem. 53(15), (2010), 5629–5638.

[155] T. Suzuki, H. Ozasa, Y. Itoh, P. Zhan, H. Sawada, K. Mino, L. Walport, R. Ohkubo, A. Kawamura, M. Yonezawa, Y. Tsukada, A. Tumber, H. Nakagawa, M. Hasegawa, R. Sasaki, et al., Identiﬁcation of the KDM2/7 Histone Lysine Demethylase Subfamily Inhibitor and its Antiproliferative Activity. J. Med. Chem. 56(18), (2013), 7222–7231.

[156] Y. Itoh, H. Sawada, M. Suzuki, T. Tojo, R. Sasaki, M. Hasegawa, T. Mizukami, T. Suzuki, Identiﬁcation of Jumonji AT-Rich Interactive Domain 1A Inhibitors and Their Eﬀect on Cancer Cells. ACS Med. Chem. Lett. 6(6), (2015), 665–670.

[157] K.-H. Chang, O. N. F. King, A. Tumber, E. C. Y. Woon, T. D. Heightman, M. A. Mc- Donough, C. J. Schoﬁeld, N. R. Rose, Inhibition of Histone Demethylases by 4-Carboxy- 2,2’-Bipyridyl Compounds. ChemMedChem 6(5), (2011), 759–764.

235 References

[158] L. Kruidenier, C.-w. Chung, Z. Cheng, J. Liddle, K. Che, G. Joberty, M. Bantscheﬀ, C. Bountra, A. Bridges, H. Diallo, D. Eberhard, S. Hutchinson, E. Jones, R. Katso, M. Leveridge, et al., A selective jumonji H3K27 demethylase inhibitor modulates the proinﬂammatory macrophage response. Nature 488(7411), (2012), 404–408.

[159] B. Heinemann, J. M. Nielsen, H. R. Hudlebusch, M. J. Lees, D. V. Larsen, T. Boesen, M. Labelle, L.-O. Gerlach, P. Birk, K. Helin, Inhibition of demethylases by GSK-J1/J4. Nature 514(7520), (2014), E1–E2.

[160] L. Kruidenier, C.-w. Chung, Z. Cheng, J. Liddle, K. Che, G. Joberty, M. Bantscheﬀ, C. Bountra, A. Bridges, H. Diallo, D. Eberhard, S. Hutchinson, E. Jones, R. Katso, M. Leveridge, et al., Inhibition of demethylases by GSK-J1/J4 Reply. Nature 514(7520), (2014), E2.

[161] J. Hu, X. Wang, L. Chen, M. Huang, W. Tang, J. Zuo, Y.-C. Liu, Z. Shi, R. Liu, J. Shen, B. Xiong, Design and discovery of new pyrimidine coupled nitrogen aromatic rings as chelating groups of JMJD3 inhibitors. Bioorg. Med. Chem. Lett. 26(3), (2016), 721–725.

[162] M. C. Chan, O. Atasoylu, E. Hodson, A. Tumber, I. K. H. Leung, R. Chowdhury, V. G´omez-P´erez,M. Demetriades, A. M. Rydzik, J. Holt-Martyn, Y.-M. Tian, T. Bishop, T. D. W. Claridge, A. Kawamura, C. W. Pugh, et al., Potent and Selective Triazole- Based Inhibitors of the Hypoxia-Inducible Factor Prolyl-Hydroxylases with Activity in the Murine Brain. PLoS One 10(7), (2015), e0132004.

[163] K. S. England, A. Tumber, T. Krojer, G. Scozzafava, S. S. Ng, M. Daniel, A. Szykowska, K. Che, F. von Delft, N. A. Burgess-Brown, A. Kawamura, C. J. Schoﬁeld, P. E. Brennan, Optimisation of a triazolopyridine based histone demethylase inhibitor yields a potent and selective KDM2A (FBXL11) inhibitor. MedChemComm 5(12), (2014), 1879–1886.

[164] V. Bavetsias, R. M. Lanigan, G. F. Ruda, B. Atrash, M. G. McLaughlin, A. Tumber, N. Y. Mok, Y.-V. Le Bihan, S. Dempster, K. J. Boxall, F. Jeganathan, S. B. Hatch, P. Savitsky, S. Velupillai, T. Krojer, et al., 8-Substituted Pyrido[3,4-d]pyrimidin-4(3H )- one Derivatives As Potent, Cell Permeable, KDM4 (JMJD2) and KDM5 (JARID1) Histone Lysine Demethylase Inhibitors. J. Med. Chem. 59(4), (2016), 1388–1409.

[165] Z. Nie, J. A. Staﬀord, J. M. Veal, M. B. Wallace, Histone Demethylase Inhibitors. Patent WO(2014/089364), (2014), 1–294.

[166] M. Labelle, T. Boesen, M. Mehrotra, Q. Khan, F. Ullah, Inhibitors of Histone Demethy- lases. Patent WO(2014/053491), (2014), 1–223.

[167] U. Leurs, R. P. Clausen, J. L. Kristensen, B. Lohse, Inhibitor scaﬀold for the histone lysine demethylase KDM4C (JMJD2C). Bioorg. Med. Chem. Lett. 22(18), (2012), 5811– 5813.

236 References

[168] N. R. Rose, E. C. Y. Woon, A. Tumber, L. J. Walport, R. Chowdhury, X. S. Li, O. N. F. King, C. Lejeune, S. S. Ng, T. Krojer, M. C. Chan, A. M. Rydzik, R. J. Hopkinson, K. H. Che, M. Daniel, et al., Plant Growth Regulator Daminozide Is a Selective Inhibitor of Human KDM2/7 Histone Demethylases. J. Med. Chem. 55(14), (2012), 6639–6643.

[169] S.-H. Kim, S. H. Kwon, S.-H. Park, J. K. Lee, H.-S. Bang, S.-J. Nam, H. C. Kwon, J. Shin, D.-C. Oh, Tripartin, a Histone Demethylase Inhibitor from a Bacterium Associated with a Dung Beetle Larva. Org. Lett. 15(8), (2013), 1834–1837.

[170] C.-H. Chu, L.-Y. Wang, K.-C. Hsu, C.-C. Chen, H.-H. Cheng, S.-M. Wang, C.-M. Wu, T.-J. Chen, L.-T. Li, R. Liu, C.-L. Hung, J.-M. Yang, H.-J. Kung, W.-C. Wang, KDM4B as a Target for Prostate Cancer: Structural Analysis and Selective Inhibition by a Novel Inhibitor. J. Med. Chem. 57(14), (2014), 5975–5985.

[171] A. K. Upadhyay, D. Rotili, J. W. Han, R. Hu, Y. Chang, D. Labella, X. Zhang, Y.- s. Yoon, A. Mai, X. Cheng, An Analog of BIX-01294 Selectively Inhibits a Family of Histone H3 Lysine 9 Jumonji Demethylases. J. Mol. Biol. 416(3), (2012), 319–327.

[172] R. Sekirnik, N. R. Rose, A. Thalhammer, P. T. Seden, J. Mecinovi´c,C. J. Schoﬁeld, Inhibition of the histone lysine demethylase JMJD2A by ejection of structural Zn(II). Chem. Commun. (42), (2009), 6376–6378.

[173] C. J. Schoﬁeld, N. Rose, R. Sekirnik, JMJD2 Demethylase Inhibitors. Patent WO(2011/030108), (2011), 1–63.

[174] B. Lohse, A. L. Nielsen, J. B. L. Kristensen, C. Helgstrand, P. A. C. Cloos, L. Olsen, M. Gajhede, R. P. Clausen, J. L. Kristensen, Targeting Histone Lysine Demethylases by Truncating the Histone 3 Tail to Obtain Selective Substrate-Based Inhibitors. Angew. Chem. Int. Ed. 50(39), (2011), 9100–9103.

[175] E. C. Y. Woon, A. Tumber, A. Kawamura, L. Hillringhaus, W. Ge, N. R. Rose, J. H. Y. Ma, M. C. Chan, L. J. Walport, K. H. Che, S. S. Ng, B. D. Marsden, U. Oppermann, M. A. McDonough, C. J. Schoﬁeld, Linking of 2-Oxoglutarate and Substrate Binding Sites Enables Potent and Highly Selective Inhibition of JmjC Histone Demethylases. Angew. Chem. Int. Ed. 51(7), (2012), 1631–1634.

[176] U. Leurs, B. Lohse, K. D. Rand, S. Ming, E. S. Riise, P. A. Cole, J. L. Kristensen, R. P. Clausen, Substrate- and Cofactor-independent Inhibition of Histone Demethylase KDM4C. ACS Chem. Biol. 9(9), (2014), 2131–2138.

[177] J. A. Day, S. M. Cohen, Investigating the Selectivity of Metalloenzyme Inhibitors. J. Med. Chem. 56(20), (2013), 7997–8007.

[178] Y. Chen, S. M. Cohen, Investigating the Selectivity of Metalloenzyme Inhibitors in the Presence of Competing Metalloproteins. ChemMedChem 10(10), (2015), 1733–1738.

237 References

[179] C. A. Lipinski, F. Lombardo, B. W. Dominy, P. J. Feeney, Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Delivery Rev. 46(1-3), (2001), 3–26.

[180] G. L. Patrick, An Introduction to Medicinal Chemistry, Oxford University Press, Oxford, 4th edn., 2009.

[181] P. D. Leeson, B. Springthorpe, The inﬂuence of drug-like concepts on decision-making in medicinal chemistry. Nat. Rev. Drug Discov. 6(11), (2007), 881–890.

[182] M. M. Hann, Molecular obesity, potency and other addictions in drug discovery. MedChemComm 2(5), (2011), 349–355.

[183] G. W. Langley, A. Brinkø, M. M¨unzel,L. J. Walport, C. J. Schoﬁeld, R. J. Hop- kinson, Analysis of JmjC Demethylase-Catalyzed Demethylation Using Geometrically- Constrained Lysine Analogues. ACS Chem. Biol. 11(3), (2016), 755–762.

[184] J.-H. Zhang, T. D. Y. Chung, K. R. Oldenburg, A Simple Statistical Parameter for Use in Evaluation and Validation of High Throughput Screening Assays. J. Biomol. Screen. 4(2), (1999), 67–73.

[185] P. M. Perrigue, J. Najbauer, J. Barciszewski, Histone demethylase JMJD3 at the inter- section of cellular senescence and cancer. Biochim. Biophys. Acta, Rev. Cancer 1865(2), (2016), 237–244.

[186] L. L. Stookey, Ferrozine – A New Spectrophotometric Reagent for Iron. Anal. Chem. 42(7), (1970), 779–781.

[187] C. R. Gibbs, Characterization and Application of FerroZine Iron Reagent as a Ferrous Iron Indicator. Anal. Chem. 48(8), (1976), 1197–1201.

[188] H. Jomaa, J. Wiesner, S. Sanderbrand, B. Altincicek, C. Weidemeyer, M. Hintz, I. T¨urba- chova, M. Eberl, J. Zeidler, H. K. Lichtenthaler, D. Soldati, E. Beck, Inhibitors of the Nonmevalonate Pathway of Isoprenoid Biosynthesis as Antimalarial Drugs. Science 285(5433), (1999), 1573–1576.

[189] N. R¨uger,M. Roatsch, T. Emmrich, H. Franz, R. Sch¨ule,M. Jung, A. Link, Tetrazolyl- hydrazides as Selective Fragment-Like Inhibitors of the JumonjiC-Domain-Containing Histone Demethylase KDM4A. ChemMedChem 10(11), (2015), 1875–1883.

[190] P. Lassalas, B. Gay, C. Lasfargeas, M. J. James, V. Tran, K. G. Vijayendran, K. R. Brunden, M. C. Kozlowski, C. J. Thomas, A. B. Smith, D. M. Huryn, C. Ballatore, Structure Property Relationships of Carboxylic Acid Isosteres. J. Med. Chem. 59(7), (2016), 3183–3203.

[191] S. N. Ononye, M. D. VanHeyst, E. Z. Oblak, W. Zhou, M. Ammar, A. C. Anderson, D. L. Wright, Tropolones As Lead-Like Natural Products: The Development of Potent

238 References

and Selective Histone Deacetylase Inhibitors. ACS Med. Chem. Lett. 4(8), (2013), 757– 761.

[192] C. Meck, M. P. D’Erasmo, D. R. Hirsch, R. P. Murelli, The biology and synthesis of α-hydroxytropolones. MedChemComm 5(7), (2014), 842–852.

[193] Y. Shimada, M. Imamura, T. Wagata, N. Yamaguchi, T. Tobe, Characterization of 21 newly established esophageal cancer cell lines. Cancer 69(2), (1992), 277–284.

[194] L. Morera, M. Roatsch, M. C. D. Fürst,I. Hoffmann, J. Senger, H. Franz, R. Schüle,M. R. Heinrich, M. Jung, 4-Biphenylalanine- and 3-Phenyltyrosine-Derived Hydroxamic Acids as Inhibitors of the JumonjiC Domain-Containing Histone Demethylase KDM4A. submitted .

[195] S. Wittich, H. Scherf, C. Xie, G. Brosch, P. Loidl, C. Gerh¨auser,M. Jung, Structure- Activity Relationships on Phenylalanine-Containing Inhibitors of Histone Deacetylase: In Vitro Enzyme Inhibition, Induction of Diﬀerentiation, and Inhibition of Proliferation in Friend Leukemic Cells. J. Med. Chem. 45(15), (2002), 3296–3309.

[196] S. Sch¨afer,L. Saunders, E. Eliseeva, A. Velena, M. Jung, A. Schwienhorst, A. Strasser, A. Dickmanns, R. Ficner, S. Schlimme, W. Sippl, E. Verdin, M. Jung, Phenylalanine- containing hydroxamic acids as selective inhibitors of class IIb histone deacetylases (HDACs). Bioorg. Med. Chem. 16(4), (2008), 2011–2033.

[197] R. Ortmann, J. Wiesner, A. Reichenberg, D. Henschker, E. Beck, H. Jomaa, M. Schlitzer, Acyloxyalkyl ester prodrugs of FR900098 with improved in vivo anti-malarial activity. Bioorg. Med. Chem. Lett. 13(13), (2003), 2163–2166.

[198] D. M. Fass, S. A. Reis, B. Ghosh, K. M. Hennig, N. F. Joseph, W.-N. Zhao, T. J. F. Nieland, J.-S. Guan, C. E. Groves Kuhnle, W. Tang, D. D. Barker, R. Mazitschek, S. L. Schreiber, L.-H. Tsai, S. J. Haggarty, Crebinostat: A novel cognitive enhancer that inhibits histone deacetylase activity and modulates chromatin-mediated neuroplasticity. Neuropharmacology 64, (2013), 81–96.

[199] E.-M. Herrlinger, Struktur-Wirkungsbeziehungen an Hemmstoﬀen von Histondemethy- lasen und -desacetylasen, Bachelor Thesis. Albert-Ludwigs-Universit¨at, Freiburg im Breisgau, 2013.

[200]S. Ogushi,¯ M. Ando, D. Tsuru, Formaldehyde Dehydrogenase from Pseudomonas putida: A Zinc Metalloenzyme. J. Biochem. 96(5), (1984), 1587–1591.

[201] M. Roatsch, D. Robaa, M. Pippel, J. E. Nettleship, Y. Reddivari, L. E. Bird, I. Hoﬀ- mann, H. Franz, R. J. Owens, R. Sch¨ule, R. Flaig, W. Sippl, M. Jung, Substi- tuted 2-(2-aminopyrimidin-4-yl)pyridine-4-carboxylates as potent inhibitors of JumonjiC domain-containing histone demethylases. Future Med. Chem. 8(13), (2016), in press, doi:10.4155/fmc.15.188.

239 References

[202] O. Moradei, I. Paquin, S. Leit, S. Frechette, A. Vaisburg, J. M. Besterman, P. Tessier, T. C. Mallais, Inhibitors of histone deacetylase. Patent WO(2005/030704), (2005), 151– 152. [203] N. Zhou, O. Moradei, S. Raeppel, S. Leit, S. Frechette, F. Gaudette, I. Paquin, N. Bernstein, G. Bouchain, A. Vaisburg, Z. Jin, J. Gillespie, J. Wang, M. Four- nel, P. T. Yan, et al., Discovery of N -(2-Aminophenyl)-4-[(4-pyridin-3-ylpyrimidin- 2-ylamino)methyl]benzamide (MGCD0103), an Orally Active Histone Deacetylase In- hibitor. J. Med. Chem. 51(14), (2008), 4072–4075. [204] H. Maag, Prodrugs of Carboxylic Acids, in: V. J. Stella, R. T. Borchardt, M. J. Hageman, R. Oliyai, H. Maag, J. W. Tilley (Eds.), Prodrugs: Challenges and Rewards Part 1, 703– 729, Springer, New York, 2007. [205] M. S. Bernatowicz, Y. Wu, G. R. Matsueda, 1H -Pyrazole-1-carboxamidine Hydro- chloride: An Attractive Reagent for Guanylation of Amines and Its Application to Pep- tide Synthesis. J. Org. Chem. 57(8), (1992), 2497–2502. [206] J. Dehaudt, J. Husson, L. Guyard, A more efficient synthesis of 4,4’,4”-tricarboxy- 2,2’:6’,2”-terpyridine. Green Chem. 13(12), (2011), 3337–3340. [207] S. H. Wadman, J. M. Kroon, K. Bakker, R. W. A. Havenith, G. P. M. van Klink, G. van Koten, Cyclometalated Organoruthenium Complexes for Application in Dye-Sensitized Solar Cells. Organometallics 29(7), (2010), 1569–1579. [208] K. M. Al-Zaydi, R. M. Borik, Microwave Assisted Condensation Reactions of 2-Aryl Hydrazonopropanals with Nucleophilic Reagents and Dimethyl Acetylenedicarboxylate. Molecules 12(8), (2007), 2061–2079. [209] R. S. Patil, K. Charugundla, P. K. Neela, N. S. Pradhan, J. Valgeirsson, Substantially pure Deferasirox and process for the preparation thereof. Patent WO(2009/147529), (2009), 16. [210] S. Steinhauser, U. Heinz, M. Bartholomä,T. Weyhermüller,H. Nick, K. Hegetschweiler, Complex Formation of ICL670 and Related Ligands with FeIII and FeII. Eur. J. Inorg. Chem. 2004(21), (2004), 4177–4192. [211] F. Kielar, Q. Wang, P. D. Boyle, K. J. Franz, A boronate prochelator built on a triazole framework for peroxide-triggered tridentate metal binding. Inorg. Chim. Acta 393, (2012), 294–303. [212] F. Stieber, U. Grether, H. Waldmann, Development of the Traceless Phenylhydrazide Linker for Solid-Phase Synthesis. Chem. Eur. J. 9(14), (2003), 3270–3281. [213] S. J. Ford, P. Obeidy, D. B. Lovejoy, M. Bedford, L. Nichols, C. Chadwick, O. Tucker, G. Y. L. Lui, D. S. Kalinowski, P. J. Jansson, T. H. Iqbal, D. Alderson, D. R. Richardson, C. Tselepis, Deferasirox (ICL670A) effectively inhibits oesophageal cancer growth in vitro and in vivo. Br. J. Pharmacol. 168(6), (2013), 1316–1328.

240 References

[214] I. P. Pogribny, V. P. Tryndyak, M. Pogribna, S. Shpyleva, G. Surratt, G. Gamboa da Costa, F. A. Beland, Modulation of intracellular iron metabolism by iron chelation aﬀects chromatin remodeling proteins and corresponding epigenetic modiﬁcations in breast cancer cells and increases their sensitivity to chemotherapeutic agents. Int. J. Oncol. 42(5), (2013), 1822–1832.

[215] D.-H. Lee, P. S. Jang, N. G. Chung, B. Cho, D. C. Jeong, H. K. Kim, Deferasirox shows in vitro and in vivo antileukemic eﬀects on murine leukemic cell lines regardless of iron status. Exp. Hematol. 41(6), (2013), 539–546.

[216] J. H. Choi, J. S. Kim, Y. W. Won, J. Uhm, B. B. Park, Y. Y. Lee, The potential of deferasirox as a novel therapeutic modality in gastric cancer. World J. Surg. Oncol. 14, (2016), 77.

[217] A. M. Zeidan, F. Hendrick, E. Friedmann, M. R. Baer, S. D. Gore, M. Sasane, C. Paley, A. J. Davidoﬀ, Deferasirox therapy is associated with reduced mortality risk in a medicare population with myelodysplastic syndromes. J. Comp. Eﬀ. Res. 4(4), (2015), 327–340.

[218] B. D. Keeler, M. J. Brookes, Iron chelation: a potential therapeutic strategy in oesophageal cancer. Br. J. Pharmacol. 168(6), (2013), 1313–1315.

[219] H. Nick, Combination Comprising an Iron Chelator and an Anti-Neoplastic Agent and Use Thereof. Patent WO(2007/128820), (2007), 1–55.

[220] M. R. Bedford, S. J. Ford, R. D. Horniblow, T. H. Iqbal, C. Tselepis, Iron Chelation in the Treatment of Cancer: A New Role for Deferasirox? J. Clin. Pharmacol. 53(9), (2013), 885–891.

[221] M. Hesse, H. Meier, B. Zeeh, Spektroskopische Methoden in der organischen Chemie, Georg Thieme Verlag, Stuttgart, 7th edn., 2005.

[222] T. Shinada, M. Hamada, K. Miyoshi, M. Higashino, T. Umezawa, Y. Ohfune, Mild and Catalytic Transesteriﬁcation Reaction Using K2HPO4 for the Synthesis of Methyl Esters. Synlett 2010(14), (2010), 2141–2145.

241

A. List of Publications

Certain parts of this thesis work have previously been published in scientiﬁc journals and presented at conferences according to the following list.

• Published Original Articles – N. Rüger,M. Roatsch, T. Emmrich, H. Franz, R. Schüle,M. Jung, A. Link, Tetra- zolylhydrazides as Selective Fragment-Like Inhibitors of the JumonjiC-Domain- Containing Histone Demethylase KDM4A. ChemMedChem 10(11), (2015), 1875– 1883. – M. Roatsch, D. Robaa, M. Pippel, J. E. Nettleship, Y. Reddivari, L. E. Bird, I. Hoff- mann, H. Franz, R. J. Owens, R. Schüle,R. Flaig, W. Sippl, M. Jung, Substituted 2-(2-aminopyrimidin-4-yl)pyridine-4-carboxylates as potent inhibitors of JumonjiC domain-containing histone demethylases. Future Med. Chem. 8(13), (2016), in press, doi:10.4155/fmc.15.188. • Original Articles in Preparation – L. Morera, M. Roatsch, M. C. D. Fürst,I. Hoffmann, J. Senger, H. Franz, R. Schüle, M. R. Heinrich, M. Jung, 4-Biphenylalanine- and 3-Phenyltyrosine-Derived Hydrox- amic Acids as Inhibitors of the JumonjiC Domain-Containing Histone Demethy- lase KDM4A. submitted . – Another original manuscript detailing the results for clinically used iron chelators (cf. Section 4.6) is currently in preparation. • Review Articles and Book Chapters – I. Hoffmann, M. Roatsch, M. L. Schmitt, L. Carlino, M. Pippel, W. Sippl, M. Jung, The role of histone demethylases in cancer therapy. Mol. Oncol. 6(6), (2012), 683– 703. – A.-T. Hauser, M. Roatsch, J. Schulz-Fincke, D. Robaa, W. Sippl, M. Jung, Discovery of Histone Demethylase Inhibitors, in: Y. G. Zheng (Ed.), Epigenetic Technological Applications, 397–424, Academic Press, Amsterdam, 2015. • Poster Presentations as Main Author – M. Roatsch, H. Franz, R. Schüle,M. Jung, “Screening for inhibitors of JumonjiC domain-containing histone demethylases using a fluorescence-based assay”, 6th Sum- mer School Medicinal Chemistry, Regensburg, Germany, 26-28 September 2012. (Poster prize)

243 Appendix A. List of Publications

– M. Roatsch, H. Franz, R. Schüle,M. Jung, “Development of inhibitors of JumonjiC domain-containing histone demethylases using fluorescence-based assays”, Spring School Epigenetics of Civilization Diseases, Leipzig, Germany, 27-31 May 2013. – M. Roatsch, N. Rüger,H. Franz, R. Schüle,A. Link, M. Jung, “Tetrazolyl Hydra- zides as Fragment-Like Inhibitors of the JumonjiC Domain-Containing Histone Demethylase JMJD2A”, GDCh-Fachgruppentagung Frontiers in Medicinal Chemistry, Tübingen,Germany, 16-19 March 2014. – M. Roatsch, L. Morera, I. Hoffmann, H. Franz, R. Schüle,M. Jung, “Biphenylalanine- Derived Hydroxamic Acids as Inhibitors of the JumonjiC Domain-Containing Histone Demethylase JMJD2A”, International Symposium on Medical Epigenetics, Freiburg i. Br., Germany, 7-9 April 2014. – M. Roatsch, D. Arican, K. Schmidtkunz, H. Franz, R. Schüle,R. Brückner, M. Jung, “The Natural Product β-Thujaplicin and Novel Benzotropolones as Leads for Inhibitors of JumonjiC Domain-Containing Histone Demethylases”, Tag der Forschung der Fakultät,Freiburg i. Br., Germany, 4 July 2014. – M. Roatsch, L. Morera, N. Rüger,H. Franz, R. Schüle,A. Link, M. Jung, “Hydrox- amic Acids and Tetrazolyl Hydrazides as Novel Inhibitors of the JumonjiC Domain- Containing Histone Demethylase JMJD2A”, XXIII International Symposium on Medicinal Chemistry EFMC-ISMC, Lisbon, Portugal, 7-11 September 2014. – M. Roatsch, I. Hoffmann, L. Morera, H. Franz, R. Schüle,M. Jung, “Biphenylalanine- Derived Hydroxamic Acids as Inhibitors of the JumonjiC Domain-Containing Histone Demethylase JMJD2A”, Max Planck Freiburg Epigenetics Meeting, Freiburg i. Br., Germany, 3-5 December 2014. – M. Roatsch, L. Morera, N. Rüger,H. Franz, J. Senger, R. Schüle,A. Link, M. Jung, “Hydroxamic Acids and Tetrazolyl Hydrazides as Novel Inhibitors of JumonjiC Domain-Containing Histone Demethylases”, GDCh-Fachgruppentagung Frontiers in Medicinal Chemistry, Marburg, Germany, 15-18 March 2015. – M. Roatsch, D. Robaa, M. Pippel, J. E. Nettleship, Y. Reddivari, L. E. Bird, I. Hoffmann, H. Franz, R. J. Owens, R. Schüle,R. Flaig, W. Sippl, M. Jung, “2- (2-Aminopyrimidin-4-yl)-pyridine-4-carboxylates as a Novel Scaffold for Potent and Selective Inhibitors of JumonjiC Domain-Containing Histone Demethylases Identi- fied by Virtual Screening”, 6th EFMC International Symposium on Advances in Synthetic and Medicinal Chemistry, Rehovot, Israel, 15-18 November 2015. – M. Roatsch, D. Robaa, M. Pippel, J. E. Nettleship, Y. Reddivari, L. E. Bird, I. Hoffmann, H. Franz, R. J. Owens, R. Schüle,R. Flaig, W. Sippl, M. Jung, “2- (2-Aminopyrimidin-4-yl)-pyridine-4-carboxylates as a Novel Scaffold for Potent and Selective Inhibitors of JumonjiC Domain-Containing Histone Demethylases Iden- tified by Virtual Screening”, 3rd Freiburg Epigenetic Spring Meeting: Chemical Biology of Epigenetics, Freiburg i. Br., Germany, 10-13 April 2016.

244 • Talks at Scientific Meetings – “Screening for inhibitors of JumonjiC domain-containing histone demethylases using a fluorescence-based assay”, Short Poster Presentation, 6th Summer School Medic- inal Chemistry, Regensburg, Germany, 26-28 September 2012. – “Development of inhibitors of JumonjiC domain-containing histone demethylases using fluorescence-based assays”, Short Poster Presentation, Spring School Epi- genetics of Civilization Diseases, Leipzig, Germany, 27-31 May 2013. – “Biphenylalanine-Derived Hydroxamic Acids as Inhibitors of the JumonjiC Domain- Containing Histone Demethylase JMJD2A”, Short Poster Presentation, Interna- tional Symposium on Medical Epigenetics, Freiburg i. Br., Germany, 7-9 April 2014.

245

B. List of Abbreviations

2-OG 2-oxoglutarate (α-ketoglutarate) 8-HQ 8-hydroxyquinoline AlphaScreen screening method using an amplified luminescent proximity homogeneous assay APCI-MS mass spectrometry using chemical ionization at atmospheric pressure AR androgen receptor BSA bovine serum albumin clogP calculated partition coefficient between n-octanol and water COSY correlation spectroscopy DELFIA dissociation-enhanced lanthanide fluorescent immunoassay DIPEA N,N -diisopropylethylamine DMF N,N -dimethylformamide DMF-DMA N,N -dimethylformamide-dimethylacetal DMSO dimethyl sulfoxide DNA deoxyribonucleic acid DNMT DNA methyltransferase EDTA ethylenediaminetetraacetic acid ELISA enzyme-linked immunosorbent assay ESI-MS mass spectrometry using electrospray ionization FAD flavin adenine dinucleotide FBDD fragment-based drug discovery FDA United States Food and Drug Administration FDH formaldehyde dehydrogenase FI fluorescence intensity FP fluorescence polarization FRET fluorescence (also: Forster¨ ) resonance energy transfer

GI50 concentration, at which 50% inhibition of cell growth is observed

H3K9me3 histone H3 trimethylated at lysine residue #9 HAT histone acetyltransferase

247 Appendix B. List of Abbreviations

HDAC histone deacetylase HDM histone demethylase HEPES 2-[4-(2-hydroxyethyl)piperazin-1-yl] ethanesulfonic acid HMBC heteronuclear multiple bond correlation HMT histone methyltransferase HPLC high-performance (also: high-pressure) liquid chromatography HSQC heteronuclear single quantum coherence HTS high-throughput screening

IC50 concentration, at which 50% inhibition is observed JARID Jumonji, AT-rich interactive domain JMJC JumonjiC domain JMJC demethylase JumonjiC domain-containing histone demethylase KDM lysine demethylase

Ki inhibition constant

λem emission wavelength

λex excitation wavelength LANCE lanthanide chelate excite LSD1 lysine-specific demethylase 1 MALDI matrix-assisted laser desorption ionization MBG metal-binding group MS mass spectrometry NAD+ oxidized form of nicotinamide adenine dinucleotide NADH reduced form of nicotinamide adenine dinucleotide n. i. no inhibition NMR nuclear magnetic resonance NOG N -oxalyl glycine n. t. not tested o/n overnight PDB Protein Data Bank (www.pdb.org) 2,4-PDCA pyridine-2,4-dicarboxylic acid ppm parts per million PTM post-translational modification RFU relative fluorescence units RNA ribonucleic acid r. t. room temperature SAM S-adenosyl methionine

248 SAR structure-activity relationship s. d. standard deviation SDS-PAGE sodium dodecylsulfate polyacrylamide gel electrophoresis s. e. m. standard error of the mean SGC Structural Genomics Consortium TET ten-eleven translocation TFA triﬂuoroacetic acid TLC thin-layer chromatography TRF time-resolved ﬂuorescence UV ultraviolet light

249

C. List of Compound Synonyms

All chemical compounds that were tested for their inhibition against one of the JumonjiC histone demethylases are numbered consecutively in the text. These tables give a summary of their trivial names (if available) as well as their sources, lab codes, and for commercial compounds their catalog numbers.

Reference Inhibitors

Compound Source Code Compound Source Code NOG 2 Aldrich O9390 IOX1 13a Aldrich SML0067 2,4-PDCA 6 Acros 131860010 JIB-04 31 Ludovica Morera LM62 GSK-J1 23a Aldrich SML0709 Deferoxamine 37 Aldrich D9533 GSK-J4 23b Aldrich SML0701 EDTA 38 Stock

Co-Substrate Analogs (cf. Section 4.1)

Compound Source Code Compound Source Code 39 Ludovica/Soma FKG 41 Alfa Aesar A10831 (R)-5 Cayman Chemical 11605 42 Aldrich Z4277 40 Aldrich 337757 FR-900098 43a Schlitzer FR900098

251 Appendix C. List of Compound Synonyms

Tetrazolyl Hydrazide-Based Inhibitors (cf. Section 4.2)

Compound Source Code Compound Source Code 44a Screening Plate Link IV C5 47h Nicole Rüger NR061 44b Screening Plate Link IV C6 47i Nicole Rüger NR063 44c Screening Plate Link IV C7 47j Nicole Rüger NR064 44d Screening Plate Link IV D2 47k Nicole Rüger NR065 44e Screening Plate Link IV C11 47l Nicole Rüger NR066 44f Link group FW2 47m Nicole Rüger NR067 44g Link group FW3 47n Nicole Rüger NR068 44h Link group FW1 47o Nicole Rüger NR069 44i Screening Plate Link IV C10 47p Nicole Rüger NR070 44j Screening Plate Link IV C9 47q Nicole Rüger NR073 44k Screening Plate Link IV D3, NR001 47r Nicole Rüger NR074 44l Screening Plate Link IV D4 47s Nicole Rüger NR075 44m Screening Plate Link IV D5 47t Nicole Rüger NR089 45 Nicole Rüger NR128 47u Nicole Rüger NR090 46 Nicole Rüger NR016 48 Nicole Rüger NR097 47a Nicole Rüger NR006 49 Nicole Rüger NR098 47b Nicole Rüger NR009 50 Nicole Rüger GF008 47c Nicole Rüger NR014 51a Nicole Rüger NS035 47d Nicole Rüger NR018 51b Nicole Rüger NS022 47e Nicole Rüger NR020 51c Nicole Rüger LD018 47f Nicole Rüger NR059 51d Nicole Rüger LD020 47g Nicole Rüger NR060

252 Tropolone-Based Inhibitors (cf. Section 4.3)

Compound Source Code Compound Source Code 52 Aldrich 469521 53c Brückner group BT3 53a Brückner group BT1 53d Brückner group BT4 53b Brückner group BT2 53e Brückner group DA384

Hydroxamic Acid-Based Inhibitors (cf. Section 4.4)

‘Jung Library’ compounds are substances available in the lab from earlier investigations into HDAC inhibitor development. Michael F¨urstis from the Heinrich group (FAU Erlangen- Nuremberg) and Georg Rennar from the Schlitzer group (Philipps-University Marburg).

Compound Source Code Compound Source Code 54a Jung Library AW33 59d Michael Fürst MFH240 54b Jung Library AW36 59e Michael Fürst MFH241 54c Jung Library AW37 59f Michael Fürst MFH273 54d Jung Library AW42 59g Michael Fürst MFH278 54e Jung Library J1037 59h Michael Fürst MFH272 54f Jung Library J1038 60a Michael Fürst MFH271 54g Jung Library BA11 60b Michael Fürst MFH316 54h Jung Library BA13 60c Michael Fürst MFH337 54i Jung Library AW2 61 Ludovica Morera LM28 54j Jung Library BA16 62 Ludovica Morera LM77 SAHA 17 Kristina Keller SAHA 63a Ludovica Morera LM75 55 Jung Library M344 63b Ludovica Morera LM97 56 Jung Library D231 63c Ludovica Morera LM105 57a Jung Library SW55 63d Ludovica Morera LM101 58 Jung Library M293 43b Georg Rennar GR51 57b Jung Library FR9 43c Georg Rennar GR50 57c Jung Library FR29 43d Georg Rennar GR78 57d Jung Library FR11 43e Georg Rennar GR52 57e Ludovica Morera LM21 43f Georg Rennar GR81

253 Appendix C. List of Compound Synonyms

57f Jung Library M232 43g Georg Rennar GR115 57g Ludovica Morera LM20 43h Georg Rennar GR116 57h Jung Library SW88 43i Georg Rennar GR111 57i Jung Library SW183 43j Georg Rennar GR80 57j Jung Library SW187 43k Georg Rennar GR55 57k Jung Library SW189 43l Georg Rennar GR57 57l Jung Library SW188 43m Georg Rennar GR114 57m Jung Library ST36 64 Diana Stolfa DS30xx 57n Ludovica Morera LM6 65 Ludovica/Eva-Maria EMH3 59a Michael Fürst MFH235 66 Ludovica/Eva-Maria EMH6 59b Michael Fürst MFH170 67a Ludovica/Nicole NS6 59c Michael Fürst MFH234 67b Ludovica/Nicole NS7

Aminopyrimidylpyridine-Based Inhibitors (cf. Section 4.5)

Compounds denoted as ‘synthesis’ were synthesized as part of this thesis project (cf. Experi- mental Section 7.2.3).

Compound Source Code Compound Source Code 68 Enamine Z1094781780 86y ChemBridge 35208456 69 Enamine Z381006278 86z ChemBridge 59723429 70 Aldrich CDS007376 86aa ChemBridge 53820553 71 Aldrich CDS005118 86ab ChemBridge 33238448 72 ChemBridge 9267223 86ac synthesis MR14 73 ChemBridge 9201099 86ad ChemBridge 74057889 74 ChemBridge 9236294 86ae ChemBridge 80185504 75 ChemDiv 0777-4980 86af ChemBridge 70475535 76 ChemBridge 5185708 86ag ChemBridge 39343335 77 Princeton OSSK 843751 synthesis MR21 78 Princeton OSSL 048224 86ah ChemBridge 74739942 79 Princeton OSSL 401858 86ai ChemBridge 19736132 80 Princeton OSSK 727472 86aj ChemBridge 47278796 81 Princeton OSSL 325359 synthesis MR20

254 82 ChemBridge 88855477 86ak ChemBridge 33362574 83 ChemBridge 72121440 86al ChemBridge 20246832 84 ChemBridge 5673404 86am ChemBridge 87077180 85 Alfa Aesar B22594 86an ChemBridge 74452549 86a ChemBridge 36335349 86ao ChemBridge 95811062 86b ChemBridge 45088700 86ap ChemBridge 33228717 86c ChemBridge 84125275 86aq ChemBridge 98863597 86d ChemBridge 67058636 86ar ChemBridge 62106591 86e ChemBridge 99979990 87 ChemBridge 34137300 86f ChemBridge 31295446 88 ChemBridge 88977885 86g ChemBridge 57121577 89 Enamine Z1688273828 86h ChemBridge 22832293 90 Enamine Z1688284290 86i ChemBridge 15325630 91 Selleckchem S1122 86j ChemBridge 13181507 92 Aldrich T121096 86k ChemBridge 74057889 93b synthesis MR11 86l ChemBridge 13177430 93ac synthesis MR13 86m ChemBridge 59689739 93ag synthesis MR18 86n ChemBridge 36219105 93aj synthesis MR19 86o ChemBridge 37530223 94b synthesis MR10 86p ChemBridge 64775566 94ac synthesis MR15 86q ChemBridge 32397524 95 ChemBridge 28016292 86r ChemBridge 84664460 96 ChemBridge 87269038 86s ChemBridge 77989619 97 ChemBridge 77913308 86t ChemBridge 18436005 98 ChemBridge 56704661 86u ChemBridge 10522335 99 ChemBridge 56162888 86v ChemBridge 90515442 100 ChemBridge 90804355 86w ChemBridge 23828226 101 ChemBridge 15439199 86x ChemBridge 94740409 108ag synthesis MR29

255 Appendix C. List of Compound Synonyms

Clinical Iron Chelators (cf. Section 4.6)

Compounds denoted as ‘synthesis’ were synthesized as part of this thesis project (cf. Experi- mental Section 7.2.2).

Compound Source Code Compound Source Code Deferoxamine 37 Aldrich D9533 109c synthesis MR05 Deferasirox 109a Selleckchem S1712 109d synthesis MR23 synthesis MR03 109e synthesis MR25 109b synthesis MR04 Deferiprone 110 Acros 278740050

256 “Da steh’ ich nun, ich armer Thor! Und bin so klug als wie zuvor.” Faust. Johann W. v. Goethe (1749 – 1832)