NOVEL RETINOID ENHANCERS FOR ANTI-
CANCER THERAPIES
A thesis submitted in fulfilment of the degree of
Doctor of Philosophy
By
Christopher R Gardner
Supervisors:
Prof. Naresh Kumar Dr. Belamy Cheung Prof. David StC. Black
School of Chemistry The University of New South Wales Kensington, Australia
July, 2015
PLEASE TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet
Surname or Family name: Gardner
First name: Christopher Other namefs: Robert
Abbreviation for degree as given in the University calendar: PhD
School: Chemistry Faculty: Science
Title: Novel retinoid enhancers for anti-cancer therapies
Abstract 350 words maximum: (PLEASE TYPE)
Retinoids play a crucial role in the treatment of multiple human cancers, particularly neuroblastoma and leukaemia. However, their utility is often hampered by the inherent toxicities associated with the concentrations necessary for therapeutic benefit, as well as innate or acquired resistance to treatment. This thesis focuses on the discovery and development of novel scaffolds that possess cytotoxic activity mediated through the retinoid pathways.
A protocol for docking scaffolds into the binding site of the retinoic acid receptor ~ (RAR~) was developed. These docking methodologies were used to identify novel scaffolds as potential RAR~ ligands, as well as probe the mechanisms behind the observed structure activity relationships (SAR) of the prepared compounds.
A wide range of novel indole-benzothiazole acetamides were synthesized via PyBOP amide coupling of 3-indoleacetic acids and 2-aminobenzothiazoles. These scaffolds were investigated for their cytotoxic activity against neuroblastoma cells. Several analogues were found to have low micromolar IC50 values, demonstrated selectivity towards cells with high RAR~ expression and also were selective for cancerous cells over normal cells. Furthermore, the ability of these compounds to modulate RAR protein expression was demonstrated.
Indole and quinoline-oxadiazoles were synthesized by the dehydrocyclization of bis-hydrazides by 4-toluenesulfonyl chloride. The cytotoxic activity of these scaffolds was investigated with their selectivity towards cells overexpressing different RAR subtypes (a, ~ . y) also investigated.
Benzothiazole-thieno[2,3-c]pyrazole amides were also synthesized by the one-pot, single step alkylation-aldol condensation of methyl thioglycolate and formyl-halopyrazoles. The cytotoxicity and SAR of these scaffolds was also explored.
Declaration relating to disposition of project thesis/dissertation I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).
- ~~~A1¥~t~!(?.ls.b.. : ...... ~.. Vf?J:S?.:PL?...... ·u ·:-.-.:.{ 7 Witness Date The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and ~e uire the approval of the Dean of Graduate Research.
FOR OFFICE USE ONLY Date of completion of requirements for Award:
THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS CERTIFICATE OF ORIGINALITY
I hereby declare that this submission is my own work and that, to the best of my knowledge and belief, it contains no material previously published or written by another person, nor material which to a substantial extent has been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis.
I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.
Signed ......
Date ...... 7.: 0 ~/?. . Pl.?...... ii
ABSTRACT
Retinoids play a crucial role in the treatment of multiple human cancers, particularly neuroblaostoma and leukemia. However, their utility is often hampered by the inherent toxicities assosciated with the concentrations necessary for therapeutic benefit, as well as innate or acquired resistance to treatment. This thesis focuses on the discovery and development of novel scaffolds that possess cytotoxic activity mediated through the retinoid pathways.
A protocol for docking scaffolds into the binding site of the retinoic acid receptor β
(RARβ) was developed. These docking methodologies were used to identif y novel scaffolds as potential RARβ ligands, as well as probe the mechanisms behind the observed structure activity relationships (SAR) of the prepared compounds.
A wide range of novel indole-benzothiazole acetamides were synthesized via PyBOP amide coupling of 3-indoleacetic acids and 2-aminobenzothiazoles. These scaffolds were investigated for their cytotoxic activity against neuroblastoma cells. Several analogues were found to have low micromolar IC 50 values, demonstrated selectivity towards cells with h igh RARβ expression and also were selective for cancerous cells over normal cells. Furthermore, the ability of these compounds to modulate RAR protein expression was demonstrated.
Indole and quinoline-oxadiazoles were synthesized by the dehydrocyclization of bis- hydrazides by 4-toluenesulfonyl chloride. The cytotoxic activity of these scaffolds was investigated with their selectivity towards cells overexpressing different RAR subtypes
(α, β, γ) also investigated. iii
Benzothiazole-thieno[2,3-c]pyrazole amides were also synthesized by the one-pot, single step alkylation-aldol condensation of methyl thioglycolate and formyl- halopyrazoles. The cytotoxicity and SAR of these scaffolds was also explored.
iv
ACKNOWLEDGEMENTS
My first words of thanks are offered to my supervisor Prof. Naresh Kumar. Through your guidance, support, encouragement and input, I have been able to overcome the many obstacles of post-graduate research. You have provided me with the opportunity not only to learn and develop as a scientist, but to develop the necessary skills for a career in research and teaching. But most of all, you have given me the opportunity and support to achieve one of my dreams.
I wish to also offer the same sentiment to my co-supervisors Dr. Belamy Cheung and
Prof. David StC. Black. Thank you Belamy for taking me on when I was just a chemist, with no understanding of the intricacies of molecular biology. You gave me every guidance, support and opportunity, allowing me to gain knowledge and skills I never thought I would. Thank you David for your enthusiastic encouragement, immeasurable knowledge, the stories you shared and our discussions about cricket. You have not only set an example, but challenged and encouraged me to reach it.
Also, I must thank my other supervisors; Prof. Glenn Marshall and A/Prof. Renate
Griffith. Thank you for your enthusiasm Glenn, your kindness and insight. You always reiterated the importance of doing good science, the Nature paper question and the reason we do this at all: to make someone’s lif e better. Thank you Renate for all of your assistance with my compoutational studies and the time you spent training me with the software and troubleshooting the issues I had along the way. Thank you also for guiding and supporting me with this new endeavour.
Thanks also to the staff of the UNSW School of Chemistry, the Analytical Centre and
Children’s Cancer Institute . You have taught, helped and befriended me since the time I v
was a first year undergraduate student, providing a safe and enjoyable environment for me to study and work. Without your efforts to maintain the facilities, supply chemicals and equipment, run samples and the countless other duties you perform, there would not be a School of Chemistry for me to be at, let alone this thesis.
To everyone in the Kumar-Black group, both past and present, thank you for your help and friendship. Thank you all for putting up with my questions (particularly in the early days), my OCD behaviour when it comes to random things, playing music (sometimes loudly) in the lab, my strange (possibly offensive) sense of humour and the random things I say and do. But I must single out a few people who deserve special thanks: Dr.
Kasey Wood for being a friend, mentor and gym buddy; Dr. Renxun Chen for his friendship, advice and always wanting to get lunch, or at least coffee; Dr. Samuel Kutty for always providing help with all aspects of my project, being my friend and a fellow cricket enthusiast; Dr. Kitty Ho for her help, particularly in keeping the lab working, and her friendship; and Dr. Murat Bingul for his much appreciated assistance with chemistry, being my friend, gym buddy and coffee lover.
Thank you also to all the members of the Molecular Carcinogenesis program. At some point or other I have come to you all with questions about what I was doing, or more accurately, what should I be doing, which were always met with kindness, understanding and thorough assistance. Without all your help, there would otherwise not be anywhere near the amount of biology that is in this thesis. Special thanks to
Selina, Owen and Patrick for their continued help and friendship. But of course, the biggest thanks here must go to Jess Koach. Thank you for teaching me almost everything I know about how to do molecular biology, for putting up with my countless questions and constant interruptions, and still being my friend after it all. And a special thank you for always having delicious snacks! vi
To all my friends, thank you so much for helping get me through this long journey.
Thank you to my uni friends Michelle, Robbo, Mark and Matt for all your help over the years, your friendly discussions and all the morning teas, meals and times we went for drinks. Thanks to all the guys from cricket for helping take my mind off things and giving me the chance to take out my frustration by bowling a ball or trying to hit and catch one. To my best friends Branden, Ralph, Jayson and Ruben I must offer very special thanks. You guys have always been there to make me laugh and smile, always offered me help, been there to talk to or just listen and agree with me when I rant, and always given me food , caffeine and alcohol. The times we’ve shared, be they real or in
D&D, have helped keep whatever sanity I ever had and gone a long way to helping me achieve this thesis.
To Talissa, mein schatzen, I offer a most sincere danke schön. Your unwavering love and support have helped me to achieve this dream, for which I will always be grateful.
You have put up with my worried frustration, my sleep deprived confusion, general nonsensical behaviour and periods of extended distraction. Yet through all the times I have been far from my best, you have shown me caring and understanding, always helping me through and loving me more for it. Thank you for always being there for me, for always helping me and for sharing in this journey. I also want to express my deepest gratitude to Wendy and Martin for welcoming me into their home, providing me with many dinners, enjoyable conversations, fond memories, and especially for their words of advice and support.
Finally, I must give the biggest thanks to my family, who have given me more love and support than any one man deserves. Thank you to my grandparents for always taking care of me, being supportive and eternally proud. You have taught me many of my most valuable lessons and stood me in good stead for the many challenges I have faced in my vii
PhD studies. To my sister Candice, thank you for always being ready with a hug, for putting up with my loud music and annoying sense of humour, for making me something to eat and above all for being the best friend and sister anyone could ever want. To mum and dad, I cannot thank you enough for everything you have done for me. You were my first teachers and my first friends. No matter what I have done, or what choices I have made, you have stood proudly by, confident in my ability to make my own way in the world. You have always been the first to offer me help and the first to celebrate my success. You have given me everything, even if it meant you had to go without, and done anything you thought might make my today even just a little bit better than my yesterday. Thanks to you, I not only dared to dream, but was able to achieve it. This thesis is as much a testament to your efforts as it is to mine. So for all that and more, I thank you.
viii
TABLE OF CONTENTS
CERTIFICATE OF ORIGINALITY ...... i
ABSTRACT ...... ii
ACKNOWLEDGEMENTS ...... iv
TABLE OF CONTENTS ...... viii
LIST OF ABBREVIATIONS ...... xiii
PUBLICATIONS AND PRESENTATIONS ...... xvi
CHAPTER 1: INTRODUCTION ...... - 1 -
1.1 Cancer ...... - 1 -
1.1.1 Neuroblastoma ...... - 2 -
1.2 The retinoid signalling pathway ...... - 5 -
1.3 Retinoids and cancer ...... - 8 -
1.3.1 Retinoic acid syndrome ...... - 11 -
1.3.2 Retinoid resistance ...... - 12 -
1.4 Atypical retinoids ...... - 13 -
1.5 Benz-fused heterocycles in medicinal chemistry ...... - 18 -
1.5.1 Indoles as drug scaffolds ...... - 19 -
1.5.2 Benzothiazoles as drug scaffolds ...... - 21 -
1.5.3 Benzimidazoles as drug scaffolds ...... - 23 -
1.5.4 Benzoxazoles as drug scaffolds ...... - 25 - ix
1.5.5 Quinolines in medicinal chemistry ...... - 27 -
1.6 Computer-aided drug design ...... - 31 -
1.6.1 CADD techniques in atypical retinoid design ...... - 31 -
1.7 Thesis aims ...... - 33 -
CHAPTER 2: DEVELOPMENT OF COMPUTATIONAL DOCKING STUDIES FOR
THE IDENTIFICATION OF RAR β LIGANDS ...... - 34 -
2.1 Introduction ...... - 34 -
2.2 Preliminary library screening ...... - 37 -
2.3 The GOLD software package ...... - 44 -
2.4 The RARβ protein ...... - 46 -
2.5 Establishing a docking protocol ...... - 47 -
2.6 Validation of the protocol with the lead compounds as a training set ...... - 51 -
2.7 Conclusions ...... - 55 -
CHAPTER 3: DESIGN, SYNTHESIS AND CHARACTERIZATION OF INDOLE-
BENZOTHIAZOLE ACETAMIDES ...... - 56 -
3.1 Introduction ...... - 56 -
3.2 Identification of indole acetamides as a lead scaffold ...... - 58 -
3.3 Synthesis of indole-benzothiazole amides ...... - 62 -
3.3.1 Synthesis of mixed heterocyclic acetamides ...... - 65 -
3.3.2 Synthesis of indole-benzothiazole amides with modified linkers ...... - 70 -
3.3.3 Synthesis of analogues bearing substituents on indolyl or benzothiazolyl
rings ...... - 74 - x
3.4 Characterization of in vitro activity against neuroblastoma cell lines ...... - 81 -
3.4.1 Screening of selected analogues against normal cells...... - 89 -
3.4.2 Examination of the effect on RAR protein expression ...... - 91 -
3.4.3 Further mechanistic investigations ...... - 95 -
3.5 Conclusions ...... - 97 -
CHAPTER 4: DESIGN, SYNTHESIS AND CHARACTERISATION OF INDOLE
AZOLES ...... - 99 -
4.1 Introduction ...... - 99 -
4.2 Identification of indole azoles as a lead scaffold ...... - 101 -
4.3 Synthesis of indole azoles ...... - 105 -
4.3.1 Synthesis of methylene and ethylene-bridged oxadiazoles ...... - 106 -
4.3.2 Synthesis of analogues with varied linking heterocycles ...... - 116 -
4.3.3 Synthesis of analogues bearing substituents on indolyl or phenyl rings . - 121 -
4.4 Characterization of in vitro activity against neuroblastoma cell lines ...... - 123 -
4.5 Conclusions ...... - 130 -
CHAPTER 5: DESIGN, SYNTHESIS AND CHARACTERIZATION OF
BENZOTHIAZOLE-BASED THIENOPYRAZOLES ...... - 131 -
5.1 Introduction ...... - 131 -
5.2 Identification of benzothiazole-thienopyrazoles as a lead scaffold ...... - 135 -
5.3 Synthesis of benzothiazole-thienopyrazoles ...... - 138 -
5.3.1 Synthesis of starting pyrazolones ...... - 141 -
5.3.2 Attempted synthesis of thienopyrazoles by Sonogashira cross-coupling - 147 - xi
5.3.3 Synthesis of thienopyrazoles from formylpyrazoles ...... - 151 -
5.3.5 Synthesis of furano- and pyrrolopyrazoles ...... - 154 -
5.4 Characterization of in vitro activity against neuroblastoma cell lines ...... - 156 -
5.5 Conclusion ...... - 158 -
CHAPTER 6: 8-ISOPROPYL-4-ARYLQUINOLINES AS RAR α ANTAGONISTS - 159 -
6.1 Introduction ...... - 159 -
6.2 Synthesis of 8-isopropyl-4-arylquinolines ...... - 163 -
6.2.1 Synthesis of quinoline starting materials ...... - 166 -
6.2.2 Synthesis of quinoline oxadiazoles ...... - 173 -
6.2.3 Quinoline-oxadiazoles bearing an o-fluorobenzoic acid ...... - 179 -
6.3.4 Synthesis of quinolines with varied linking regions ...... - 182 -
6.3.5 Synthesis of 4-methoxyquinolines ...... - 186 -
6.4 Characterization of in vitro activity against neuroblastoma cell lines ...... - 189 -
6.5 Conclusions ...... - 191 -
CHAPTER 7: FUTURE DIRECTIONS ...... - 192 -
CHAPTER 8: EXPERIMENTAL ...... - 194 -
8.1 General information ...... - 194 -
8.2 Experimental details ...... - 195 -
8.3 Biological assays ...... - 302 -
8.3.1 Cell culture ...... - 302 -
8.3.2 Cell viability assays ...... - 302 -
8.3.3 Immunoblot analysis ...... - 303 - xii
REFERENCES ...... - 305 -
xiii
LIST OF ABBREVIATIONS
t-RA trans -retinoic acid
13-c-RA 13-cis -retinoic acid
4-TsCl 4-toluenesulfonyl chloride
9-c-RA 9-cis -retinoic acid
AcOH acetic acid
APL acute promyelocytic leukaemia
CADD computer-aided drug design
DCC N,N-dicyclohexylcarbodiimide
DCM dichloromethane
DIPEA N,N-diisopropylethylamine
DMAD dimethyl acetylenedicarboxylate
DMAP N,N -dimethylaminopyridine
DMF N,N-dimethylormamide
DMSO dimethylsulfoxide
EDCI l-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride
EFS event free survival
ESI electrospray ionization
Et ethyl
Et 2O diethylether
Et 3N triethylamine
HDAC histone deacetylase
HOBt 1-hydroxybenzotriazole xiv
HRMS high resolution mass spectrometry
IAA 3-indoleacetic acid
IC inhibitory concentration
IR infrared spectroscopy
J coupling constant
LBD ligand binding domain lit. literature m.p. melting point
Me methyl mL millilitre(s) mmol milli mol
MYCN myelocytomatosis viral related oncogene, neuroblastoma derived
NMR nuclear magnetic resonance
Pr propyl
PyBOP benzotriazol-1-yl-oxytripyrrolidinophosphonium hexafluorophosphate
RAMBA retinoic acid metabolism blocking agents
RAR retinoic acid receptor
RARE retinoic acid response element
RMSD root-mean-square deviation
RXR retinoid X receptor
SAR structure-activity relationship
SEM standard error of the mean
Tf 2O trifluoromethanesulfonic anhydride
TFA trifluoroacetic acid
THF tetrahydrofuran xv
TTNPB 4-[( E)-2-(5,6,7,8-Tetrahydro-5,5,8,8-tetramethyl-2-naphthalenyl)-1-
propenyl]benzoic acid
UV ultraviolet spectroscopy
xvi
PUBLICATIONS AND PRESENTATIONS
A part of this research has been submitted for publication as well as presented at the following conferences:
Publications:
Gardner, C. R. ; Cheung, B. B.; Koach, J.; Black, D. S.; Marshall, G. M.; Kumar, N.
Bioorg. Med. Chem. 2012, 20 , 6877-6884.
Presentations:
Gardner, C. R. ; Koach, J.; Marshall, G. M.; Black, D. S.; Kumar, N. Cheung, B. B.;
Identification and characterization of novel small molecules which enhance the retinoid anti-cancer signals, The Lorne Cancer Conference, Lorne, Victoria, 9 th - 11 th February,
2012. (Poster presentation)
Gardner, C. R. ; Marshall, G. M.; Black, D. S.; Kumar, N. Cheung, B. B.; Identification and characterization of novel small molecules which enhance the retinoid anti-cancer signals, The EFMC XXIII International Symposium on Medicinal Chemistry, Lisbon,
Portugal, 7 th - 11 th September, 2014. (Poster presentation)
1
CHAPTER 1: INTRODUCTION
1.1 Cancer
Cancer is a heterogeneous group of genetic diseases, resulting from the aberration and alteration of key genetic and molecular pathways. 1 There are more than one hundred types of cancer, with each typically identified by the type of cells in which they arise.
However, different types of cancer share some features, commonly referred to as the hallmarks of cancer, outlined in Figure 1.1 below.2
Figure 1.1: The proposed hallmarks of cancer. 2
As a major cause of sickness and death worldwide, we need not look far to see the devastating impact that cancer has had on society. In Australia alone in 2012, approximately 67,260 and 53,460 new cases of the disease were diagnosed in men and women, respectively. Furthermore, in 2012 an estimated 24,328 males died of cancer, 2
along with 18,516 deaths in females.3 It is estimated that in the year 2020, 80,730-
89,530 men and 63,990-66,100 women will be diagnosed with cancer in Australia.4
Importantly, cancer was estimated to be the largest contributor to the burden of disease
(measured as years of healthy life lost) in Australia in 2010, accounting for approximately 19% of the total burden, with cardiovascular disease accounting for
16%.5 Furthermore, cancer was the 9th and 6 th leading cause of death in Australian men and women, respectively, in 2011. 6
1.1.1 Neuroblastoma
Neuroblastoma is an embryonal tumour of the peripheral sympathetic nervous system, derived from migrating neuroectodermal cells. 7 It is the third most commonly diagnosed children’s cancer, representin g approximately 8-10% of childhood cancer cases. The average age at diagnosis is 1-2 years, with 90% of cases diagnosed before the age of 5, and the disease being very rare in children above the age of 10.
There are two methods for describing the severity and risk associated with a neuroblastoma diagnosis: the International Neuroblastoma Staging System Committee
(INSS) 8 system and the more recent International Neuroblastoma Risk Group Staging
System (INRGSS). 9 Unlike the INSS, the INRGSS uses only the results of imaging tests taken before surgery and does not include surgical results or metastasis to lymph nodes to determine the risk stage (Table 1.1).
3
Table 1.1: Neuroblastoma risk staging systems
System Stage Description INSS 1 Can be completely surgically removed. Distant lymph nodes may/ may not contain cancer, but nearby lymph nodes do not. 2A The tumour is located only in the area it started and cannot be completely removed during surgery. Nearby lymph nodes do not contain cancer. 2B The tumour is located only in the area where it started and may/ may not be completely removed during surgery, but nearby lymph nodes do contain cancer. 3 The tumour cannot be removed with surgery. It has spread to regional lymph nodes or other areas near the tumour, but not to other parts of the body. 4 The original tumour has spread to distant lymph nodes, bones, bone marrow, liver, skin, and/or other organs, except for those listed in stage 4S. 4S The original tumour is located only where it started (as in stage 1, 2A, or 2B), and it has spread only to the skin, liver, and/or bone marrow, in infants younger than one. The spread to the bone marrow is minimal, usually less than 10%. INRGSS L1 The tumour is located only in the area where it started; no risk factors found on imaging scans, such as CT or MRI. L2 The tumour has not spread beyond the area where it started and the nearby tissue; risk factors are found on imaging scans, such as CT or MRI. M The tumour has spread to other parts of the body (except stage MS) MS The tumour has spread to only the skin, liver, and/or bone marrow.
Due to the highly heterogeneous nature of neuroblastoma,10 as well as its unique propensity for spontaneous regression (typically in stage 4S), 11-13 neuroblastoma risk classification and survival rates depend highly upon the age of the patient, 14 the stage at diagnosis,8 tumour histopathology, 15 DNA index (diploid or hyperploid) 16 and the
MYCN amplification status.17-19 Based on these factors, patients are classified with low, intermediate or high risk neuroblastoma (Table 1.2). 20 Low and intermediate-risk neuroblastoma patients have survival rates above 95% over 5-year and 3-year periods 4
respectively,21,22 while high-risk neuroblastoma patients have much lower survival rates of approximately 40% over 5 years.23 Of all neuroblastoma cases diagnosed, 37% are low-risk, 18% are intermediate-risk and 45% are high-risk. 24 Furthermore, many high- risk patients will relapse after treatment, the 5-year overall survival of which is only
20%.25 As a consequence of these factors, neuroblastoma accounts for 15% of all childhood cancer-related deaths. 26
Table 1.2: Description of the neuroblastoma risk groups.
Risk group Description Low • Stage 1 • Stage 2A/B, more than 50% of the tumour surgically removed, except with MYCN amplification • Stage 4S disease, no MYCN amplification, favourable histopathology, and hyperdiploidy Intermediate • Stage 2A/B, no MYCN amplification, < 50% of the tumour removed with surgery • Stage 3, < 18 months, no MYCN amplification • Stage 3, < 18 months, no MYCN amplification, and favourable histopathology • Stage 4, < 12 months, no MYCN amplification. • Stage 4, 12 -18 months, no MYCN amplification, hyperdiploidy, and favourable histology. • Stage 4S, no MYCN amplification, unfavourable histopathol ogy and/or diploidy High • Stage 2/B and MYCN amplification • Stage 3 and MYCN amplification • Stage 3, ≥ 18 months, no MYCN amplification and unfavourable histopathology • Stage 4, < 12 months and MYCN amplification • Stage 4, 12 -18 months with MYCN amplification, and/or diploidy, and/or unfavourable histology • Stage 4, ≥ 18 months • Stage 4S and MYCN ampl ification
Treatment options for neuroblastoma are tailored to the risk group and associated tumour biology, with more advanced tumours requiring more aggressive therapies and 5
larger combinations of therapies. For low-risk tumours, the approach is either surgery, chemotherapy with agents such as carboplatin, cyclophosphamide, doxorubicin, and etoposide, or a combination of surgery and chemotherapy. 22,27 Intermediate-risk neuroblastoma is often treated with a combination of surgery and chemotherapy, with radiotherapy also administered in progressive tumours or those which do not respond to surgery and chemotherapy. 28,29 For those with high-risk neuroblastoma, treatment is stratified into three phases: induction, consolidation and maintenance. The induction phase involves dose-intensive cycles of chemotherapy, alternating between cisplatin, etoposide, vincristine, cyclophosphamide, doxorubicin, and topotecan, followed by surgery. 30,31 Consolidation first attempts to eradicate the minimal residual disease with myeloablative chemotherapy, using combinations of either carboplatin/etoposide/melphalan or busulfan/melphalan, followed by repopulation of the bone marrow through hematopoietic stem cell transplantation (HSCT). 32-35
Maintenance aims to reduce the risk of recurrence through differentiation therapy, administering 13-cis -retinoic acid (isotretinoin), and immunotherapy, with the monoclonal antibody ch14.18 and immune-activating cytokines (GM-CSF and IL-2). 36-
38 Finally, in the case of recurrent neuroblastoma, treatment involves surgery and/or chemotherapy for patients originally classified as low- and medium-risk, while high-risk patients may receive a combination of chemotherapy, surgery, and radiation therapy followed by HSCT. 39,40
1.2 The retinoid signalling pathway
Retinoids are a class of natural and synthetic derivatives of vitamin A 1, also known as retinol.41 This essential vitamin is procured through the ingestion of foods rich in 6
preformed vitamin A, such as dairy products, fish and meat (particularly liver), or provitamin A 2, also known as β-carotene, which is found in vegetables such as sweet potato, spinach and carrot. 42 However, the oxidised metabolites of vitamin A, primarily trans -retinoic acid ( t-RA, tretinoin) 3, which is capable of performing all cellular functions of vitamin A, and 9-cis -retinoic acid (9-c-RA, alitretinoin) 4, are responsible for the activity of this vitamin. These molecules function as regulatory hormones, similar to the thyroid or steroid hormones, and play an essential role in normal embryonic development and in the maintenance of differentiation in later stages of human lives. 43-46 They are critically involved in the development of the heart, embryonal circulation, central nervous system and normal left-to-right cardiac symmetry, as well as the regulation of eye function, bone growth, spermatogenesis and immune responses. 44,47
The retinoid signalling pathway involves nuclear localised, ligand-activated transcription factors known as retinoic acid receptors (RAR). 48 These receptors are classed in three subtypes encoded by separate genes: RARα (NR1B1), RARβ (NR1B2), and RARγ (NR1B3). 48,49 RARs form heterodimers with retinoid X receptors (RXR), also possessing RXRα, RXRβ and RXRγ subtypes, 50-52 in order to associate with 7
retinoic acid response elements (RARE) in the promoter regions of their target genes. 53
Furthermore, there are at least two isoforms for each subtype, generated by differential promoter usage and alternative splicing that differs only in their N-terminal regions.
RARα, RARγ and t he three RXR subtypes each have two isoforms , while RARβ has four isoforms. 54-56 The variable combination of the eight RAR and six RXR subtypes and isoforms allows the possible formation of forty-eight different heterodimers. 44
Additionally, the retinoid receptors also show differential expression across different cell types, as well as exhibiting changing levels of expression throughout the human life cycle from the foetus, through childhood and into adulthood. 57-60 It is noteworthy that during embryonic development, RARα has near ubiquitous expression throughout the embryo , while RARβ and RARγ have much more localised and specific patterns of expression. 61
The RAR-RXR heterodimer is able to bind to the RARE in the absence of a ligand, but preferentially associates with co-repressors such as the nuclear receptor co-repressor (N-
CoR) proteins, and silencer proteins such as the silencing mediator for retinoid and thyroid receptor (SMRT) (Figure 1.2).62 These co-repressors recruit histone deacetylase
(HDAC) and methyl transferase complexes leading to chromatin condensation and sequestration of promoter elements, thereby inhibiting transcriptional activity. 63 Upon binding of a retinoid agonist, the RAR-RXR heterodimer undergoes a conformational change in its C-terminal helix, which allows for coupling to co-activator proteins (CoA) such as members of the steroid receptor co-activator (SCR-1) family, along with histone acetyl and methyl transferases (HAT), and DNA-dependant ATPases. 64,65 The resulting complex then stimulates the decondensation of the chromatin surrounding the promoter regions of target genes, thereby allowing the binding of the thyroid receptor-associated 8
protein (TRAP) and the recruitment of the transcription machinery, including RNA polymerase II and other general transcription factors. 66,67
Figure 1.2: Transcriptional activation and repression of the retinoid pathway. LBD:
ligand-binding domain. DBD: DNA-binding domain.
1.3 Retinoids and cancer
Due to the essential role of the retinoid pathway in many cellular processes, particularly differentiation and apoptosis, it is not surprising that the deregulation of this pathway has been linked to a range of human diseases, including multiple forms of cancer. In 9
several tissues and organs, the loss of a RAR subtype has been associated with tumourigenesis, suggesting that RARs may be considered as tumour suppressors.
The classical example of RAR dysfunction is in cases of acute promyelocytic leukaemia
(APL), in which chromosomal rearrangements and point mutations of the RARα gene lead to the formation of a RARα fusion protein , which in 95% of cases is the RARα -
PML (promyelocytic leukemia gene) fusion.68-72 This fusion protein has the ability to homodimerize or oligomerize with or without union to RXRα , thus increasing the recruitment of co-repressor complexes and enhancing the repression of target genes.73-75
Following the observation that t-RA 3 promoted the in vitro granulocytic differentiation of APL cells, 76,77 the first clinical trial of retinoids in the treatment of APL was conducted, in which 23 of 24 patients achieved complete remission. 63 Through subsequent trials, the efficacy of t-RA 3 in remission induction and maintenance in APL was confirmed. When used in combination with chemotherapy, t-RA 3 offered complete remission rates of up to 95% and 5 year EFS rates of up to 74%.78,79
The therapeutic benefit of t-RA 3 in APL is mediated through two mechanisms: restoration of retinoid target gene transcription and induction of RARα -PML degradation. Similar to the mechanism for retinoid induction described in Chapter 1.2, pharmacologic doses of 3 induce a conformational change in the RARα -PML protein that results in the dissociation of HDACs and DNA methyl transferases (DNMT). 73
CoA complexes are then recruited and the transcription of target genes, particularly those involved in granulocyte differentiation, is restored. 80 The degradation of the
RARα -PML oncoprotein is achieved through two non-overlapping, co-operating proteolytic mechanisms. In the first mechanism, treatment with 3 results in the targeting of a cleavage site within the PML α -helix, leaving the RARα moiety intact and capable of resuming transcriptional activity. 81 Alternatively, the ubiquitin/proteasome system 10
(UPS) is activated through the phosphorylation of RARα -PML Ser873 and the binding of the ATP-dependent protease SUG-1 to the RARα transactivation domain. 82,83
In neuroblastoma, the dysfunction of multiple RARs has been observed and identified as a key target in disease therapy. Neuroblastoma tumours typically exhibit high levels of RARα, low levels of RARγ and barely detectable levels of RARβ. 84-87 Similar to
APL, pre-clinical experiments have demonstrated that that neuroblastoma cell lines can often be induced to terminally differentiate on exposure to retinoid compounds in vitro .88,89 In the clinical trial that followed, 13-cis -retinoic acid 5 (13-c-RA, isotretinoin), a synthetic retinoid, was administered to patients following myeloablative consolidation therapy. It was found that the 3-year EFS of patients was increased to
46±6% amongst patients who received 13-c-RA 5, compared to 29±5% for those assigned with no further therapy.34 As a result of such trials, together with a better pharmacokinetic profile with higher peak levels and a longer half-life consistent with superior activity against neuroblastoma compared to t-RA 3,90 13-c-RA has become the standard of care for high-risk neuroblastoma.37
The therapeutic benefit of 13-c-RA 5 in neuroblastoma is believed to be predominantly mediated through transcriptional activation of retinoid target genes associated with differentiation. In particular, it has been suggested that 13-c-RA 5 might enhance the activity of RARβ, whose expression has been linked with growth -inhibitory effects.85
This hypothesis is supported by the observation that the loss of RARβ expression has been associated with tumourigenesis in numerous cancer types, such as those of the lung, breast and cervix.91-94 In such cancers, the RARβ gen e is frequently deleted or the
RARβ promoter silenced due to aberrant DNA methylation or repressive histone 11
modifications. 66,67 Furthermore, the restoration of RARβ expression or RARβ promoter activity was shown to decrease the tumourigenicity of the cancer cells. 95-97
Furthermore, in keratinocytes, the decrease , or absence of expression of the RARγ subtype has been linked to a predisposition to tumours. 98 RARγ was found to be absent in oral keratinocytes from head and neck cancers, as well as being dramatically decreased in skin cancers induced by UV radiation. 99,100
1.3.1 Retinoic acid syndrome
Although retinoid therapy has significantly improved long term survival rates for cancers such as neuroblastoma and APL, their use as single agent therapies has been limited by the significant toxicities associated with their use, including teratogenesis and the induction of retinoic acid syndrome. 101
Retinoic acid syndrome is characterized by unexplained fever, weight gain, elevated white blood cells, respiratory distress, interstitial pulmonary infiltrates, pleural and pericardial effusions, dyspnea, episodic hypotension, and acute renal failure. 102-104 The potentially life-threatening syndrome occurs in approximately 25% of APL cases where t-RA 3 is administered, 105 with up to 5% of cases being fatal. 103,105 In neuroblastoma, following treatment with 13-c-RA 5, the syndrome presents with skin toxicity, renal toxicity, transaminase elevation, gastrointestinal toxicity, infection, hypercalcemia, hematologic toxicity, comprised cheilitis and bone pain. 34,106 It has been shown that 13- c-RA 5 has a lower toxicity profile, with its side-effects typically being low to mild in severity. However, its side-effects are also more common, particularly those affecting the skin, and can occur in up to 80% of patients. 107 12
Due to the possible combinations of the RAR and RXR heterodimers, 44 their differential expression across different cell types, 57-60 as well as the different and specific function of each of these heterodimers, it is believed that these toxicities may be related to the non-selective activation of RAR-RXR functions by the natural ligands. 108
1.3.2 Retinoid resistance
Aside from the potential toxicities associated with the use of retinoid therapy, there have been many examples of intrinsic or acquired retinoid resistance.109 This is not surprising, as cancer cells are highly heterogeneous and exhibit deregulation in multiple cellular signalling pathways, leading to the frequent failure of treatments using chemotherapeutic agents specific to a single biological event or pathway.110
One key mechanism of retinoid resistance in both APL and neuroblastoma is the increased metabolism and subsequent reduction in plasma levels of retinoid compounds to suboptimal doses. 111-113 Sustained treatment with retinoids has been shown to induce a catabolic response through cytochrome P450 enzymes, such as P26, 2C8, 3A7, 4A11,
1B1, 2B6, and 2C9. 114-116 Furthermore, the presence of a RARE in the promoter region of the cytoplasmic retinoic acid binding protein (CRABPII) induces CRABPII following treatment with t-RA 3, leading to RA sequestration and subsequent resistance. 117
There have also been reports of retinoid resistance based on genetic mechanisms. In
APL, the rare translocation of the promyelocytic leukaemia zinc finger (PLZF) gene results in a RARα -PLZF fusion protein that is non-responsive to RA treatment. 118
Similarly, mutations in the ligand-binding domain of RARα in the RARα -PML fusion protein generate acquired retinoid resistance in relapsed cases of APL. 119 In 13
neuroblastoma, the mechanisms for retinoid resistance are largely unknown. Several oncogenic events have been examined, such as the deregulation of the MYC pathway, 120 or the failu re to recruit RXRβ to the RARE. 121
The primary method to circumvent retinoid resistance is through intermittent therapy, whereby the patient is administered treatment in the form of on and off cycles.
However, increasing focus has been directed towards the development of combinatorial treatments of retinoids with other drug types,122 especially HDAC inhibitors (HDACi) such as suberoylanilide hydroxamic acid (SAHA),123 as well as retinoic acid metabolism blocking agents (RAMBA) such as liarozole 6 and talarazole 7.124,125
1.4 Atypical retinoids
In addition to the use of natural retinoids for cancer therapies, there has been a great focus on the development of synthetic retinoids, which have shown the potential to target specific retinoid receptors. Hence, these compounds could not only act as tools to investigate the function of specific RAR or RXR subtypes, but may circumvent the toxic effects associated with pan-RAR activation.108 Other than their reduced levels of toxicity, atypical retinoids may also overcome retinoid resistance by simultaneously acting through non-retinoid pathways, sensitizing cell lines to retinoid treatment, or acting synergistically with other retinoids.126,127 14
Collectively termed atypical retinoids, the synthetic retinoids fall into two classes: retinoid related molecules (RRM), such as acyclic retinoid 8 and 4-
(hydroxyphenyl)retinamide (fenretinide) 9; and polyaromatic derivatives (arotinoids), such as the RARα agonist BMS753 10 and the RARβ/γ agonist a dapalene 11 . Both of these classes of compound follow the same general structure of the natural retinoids, being comprised of three units (highlighted in 8): a bulky hydrophobic region, a linker unit and a polar terminus, typically in the form of a carboxylic acid.
Initial developments in the field of synthetic retinoids were directed at producing molecules capable of binding to the RAR protein, which resulted in the discovery of pan-RAR agonists, similar to t-RA 3. Some pan-RAR agonists of note include 13-cis - retinoic acid (13-c-RA) 5, which is now a standard therapy for high-risk neuroblastoma patients with minimal residual disease (MRD),128 TTNPB 12, which was used to obtain the RARβ crystal structure, 129 EC23 13 and CH55 14.130,131 15
Following these discoveries, detailed structure-activity studies enabled the development of arotinoids that were selective for RAR subtypes. The development of these selective atypical retinoids has been aided by the isolation of a number of RAR and RXR crystal structures. Bourget et al . crystallised the RXRα LBD in 1995, while Renaud et al isolated the RARγ LBD in complex with t-RA 3 later that year. 132,133 In 2004, Germain et al . resolved the cr ystal structure of the RARβ LBD bound to TTNPB 12 , which revealed impo rtant features distinguishing the structure of RARβ from the RARα/γ isotypes. 123
RARα selectivity may be achieved by incorporating hydrogen-bond acceptors, such as amides in the linker region, allowing for strong interactions to the hydrogen-bond donor residue Ser232, which replaces the lipophilic alanine residues in RARβ and RARγ. Thi s feature is exemplified in the structures of the RARα agonists AM580 15 and AM555S
16 .134,135 Selectivity may also be improved through the incorporation of halogens in the non-polar region, as well as through the attachment of fluoro substituents ortho to the carboxylic acid group on the benzene ring in the polar region, such as in AGN193836
17 , the first monospecific RARα agonist. 136 16
Conversely, RARγ selectivity may be increased by the presence of hydrogen -bond donors on or adjacent to the non-polar region, which are able to form weak, non- classical hydrogen bonds to the weakly polar Met272 residue.137 Furthermore, RARγ can tolerate larger hydrophobic groups due to the presence of the smaller Ala397 residue in helix 11. These aspects are featured in the RARγ -selective CD666 18 and
138,139 CD437 19 , respectively.
Most importantly, it was found that the RARβ LBD differs from those of RARα and
RARγ in that it has a larger binding cavity, arising from the smaller Ala225 and Ile263 residues in the hydrophobic region. 129 However, although many ligands possess greater sele ctivity for RARβ over RARα and/or RARγ, complete RARβ mono -specificity is difficult to achieve. For example, the RARβ ligand BMS987 20 is also a R ARα agonist, while AGN193174 21 also acts as a RARα and RARγ antagonist due to its inability to dislodge bound co-repressors from RARα and RARγ. 129,140 Interestingly, AC55649 22 and AC261066 23 are isotype-selective agonists that preferentially target the RARβ 2 isoform. 141 17
Additionally, many arotinoids have been developed as RXR-selective ligands by mimicking the bent conformation of 9-c-RA 4. Strategies to achieve this include design of conformationally-restricted analogues of 4 such as AGN194204 24 ,142 the incorporation of ortho -substituents on the hydrophobic unit of RAR agonists giving structures such as 3-Me TTNPB 25 ,140 incorporating rigid single atom linkers such as in bexarotene 26 ,143 or incorporating benzo-fused ring systems such as in HX600 27 .144
A number of atypical retinoids have received FDA approval for the treatment of various conditions. Etretinate 28a received approval for the treatment of psoriasis, but was subsequently replaced by its metabolite acitretin 28b due to its toxic side effects.
Tazarotene 29 is approved for treatment of psoriasis, acne, and sun damaged skin, while 18
adapalene 11 is approved for use in the treatment of acne, but is also used off-label to treat keratosis pilaris. Additionally, bexarotene 26 is approved as a treatment for cutaneous T cell lymphoma (CTCL). 145
1.5 Benz-fused heterocycles in medicinal chemistry
Medicinal chemistry utilises the talents of people from multiple disciplines to discover and develop the next generation of drugs for treating disease. A critical component of this is the development of structure-activity relationships (SAR) and the identification of structures that can reliably act as a pharmacophore in drug development. One such type of pharmacophore is the heterocyclic ring system. Found in approximately half of all therapeutic agents, heterocycles are a key pharmacophore in medicinal chemistry.
Benz fused heterocyclic systems such as indole 30, benzothiazole 31, benzimidazole 32, benzoxazoleǦ 33 and quinolone 34 have been intensely studied for their interesting pharmacological activities. Based on their prevalence in nature and in synthetic drugs, benz-fused heterocycles can be considered as important scaffolds for the construction of molecules with a vast array of biological activities, as will be discussed further in the following sections of this chapter.
19
1.5.1 Indoles as drug scaffolds
From the amino acid tryptophan 35 to complex alkaloids such as vinblastine 36a and vincristine 36b , the indole motif is an essential component of many biologically active compounds. The spectrum of activity possessed by indole-based compounds includes antitubercular, 146 antimicrobial, 147 antifungal, 148 anti-inflammatory, 149 antihypertensive, 150 antiHIV, 151 antioxidant, 152 antidepressant, 153 tranquilizing and anticonvulsant properties. 154 Examples of drugs possessing some of these activities include etodolac 37 (antiarthritis),155 sumatriptan 38 (antimigraine), pindolol 39
(antihypertensive) and indolmycin 40 (antibiotic).156,157
A number of compounds containing an indole scaffold have also been reported to possess anticancer activity. A series of tetracyclic indoles was developed by Hong et al ., with 41 displaying the highest in vitro activity against human nasopharyngeal carcinoma (HONE-1) and gastric adenocarcinoma (NUGC-3) cell lines. 158 Garcia et al . 20
synthesized pyrrolo[2,3-e] indole derivatives, with 42 displaying in vitro activity against the PC3 human prostate and K562 erythroleukemia cell lines. 159 Also, Queiroz and co- workers developed 3-(dibenzothien-4-yl)indole 43, which showed potent activity in the
MCF-7 breast adenocarcinoma, NCI-H460 non-small cell lung cancer and SF-268 CNS cancer cell lines. 160
Furthermore, the indole scaffold has been utilized in the design of atypical retinoids.
Gernet et al . developed a number of specific RXR modulators based on ( E)-3-(3-(3,5- di-tert -butylphenyl)-1H-indol-5-yl)but-2-enoic acids 44 and ( E)-3-(4-(2-butoxy-3,5-di- tert -butylphenyl)-1H-indol-2-yl)but-2-enoic acid 45.161 These compounds showed improved binding to RXRα compared to their trienoic counterparts.
Yoshimura et al . have also developed a number of RAR antagonists based on 4-
(7,7,10,10-tetramethyl-4,5,7,8,9,10-hexahydro-naphtho[2,3-g]indol-3-yl)benzoic acids
46, 4-(4,5-dihydro-1H-benzo[g]indol-3-yl)benzoic acids 47 and 1,1,4,4,-tetramethyl-
2,3,4,6,7,12-hexahydro-1H-naphtho[2,3-a]carbazole-9-carboxylic acids 48 .162 These compounds were found to inhibit t-RA binding in HL-60 cells at sub-micromolar concentrations. 21
Additionally, the group of Gurkan-Alp et al . synthesized a small series of ( E)-3-(1 H- indol-3-yl)-1-(5,5,8,8-tetramethyl-5,6,7,8-tetrahydronaphthalen-2-yl)prop-2-en-1-one derivatives. The unsubstituted derivative 49 was found to possess IC 50 values ranging from 3.92 µM to less than 0.01 µM against a range of tumour cell lines. Furthermore, 49 was also shown to influence the levels of RXRα and RXRγ expression in multiple cell lines. 163
1.5.2 Benzothiazoles as drug scaffolds
Benzothiazoles are another medically important scaffold occurring frequently in natural products, such as in luciferin 50 and violatinctamine 51 .164,165 Benzothiazole-based molecules possess a range of biological properties, including antimicrobial, 166 anticonvulsant, 167 antidiabetic, 168 antitubercular, 169 antiviral, 170 anti-inflammatory and antioxidant activities.171,172 Not surprisingly, benzothiazoles also form the basis of pharmaceuticals with various properties, including riluzole 52 (neuroprotective) and phortress 53 (antitumour).173,174 22
Many benzothiazoles, both naturally occurring and designed, have been investigated as potential antitumour agents. The natural product dercitin 54 and its analogues 55, 56a-c, isolated from marine sponges, are examples of benzothiazoles with in vitro and in vivo anti-tumour activity. 175,176 These compounds were shown to inhibit the proliferation of
P388 murine leukaemia, HeLa cervical cancer and MONO-MAC 6 acute monocytic leukemia cells in vitro .177
Yoshida et al. synthesized a series of 2,6-dichloro-N-[2-(cyclopropanecarbonyl-amino) benzothiazol-6-yl]benzamides 57 that were found to possess antitumor activity against the WI-38 VA-13 (2RA) lung cancer cell line.178 Additionally, Besson et al. developed new benzothiazole-2-carbonitrile derivatives, of which 58 displayed the greatest activity against L1210 murine leukaemia cells. 179 Furthermore, Caleta et al. synthesized styrylbenzothiazoles such as 59 and 60 that showed antiproliferative activity against multiple cancer cell lines. 180 23
However, despite their many biological activities, atypical retinoids based on the benzothiazole nucleus have not yet been discovered. The RARβ 2 isotype-selective, thiazole-based 23 and RAMBA 7 do offer some precedent for the investigation of the benzothiazole scaffold in this thesis.
1.5.3 Benzimidazoles as drug scaffolds
Benzimidazole is another important scaffold in medicinal chemistry. However, unlike the simple imidazole motif, benzimidazoles occur quite rarely in natural products, with only a few examples such as kealiiquinone 61 being known.181 In organic synthesis, benzimidazoles are used for the preparation of analogues of natural alkaloids, such as the kealiiquinone regioisomer 62 and analogues of makaluvamines 63 based on imidazoquinoxalinone 64.182,183 24
Despite their relatively rare occurrence in nature, benzimidazoles offer a vast array of biological activities, including antibacterial, 184 antifungal, 185 anthelmintic, 186 antiretroviral, 187 anti-inflammatory, 188 antihepatitic and antihypertensive properties.189,190 Consequently, benzimidazoles form the basis of many pharmaceuticals, such as astemizole 65 (antihistamine) and albendazole 66 (antimicrobial). 191,192
Benzimidazoles have also been widely explored for their anticancer activities, and have shown inhibitory activity against a number of cancer-related pathways. 2-
Dimethylamino-4,5,6,7-tetrabromo-1H-benzimidazole 67 has the lowest-ever reported
Ki value for a CK2 protein kinase inhibitor at 40 nM, while 3-benzimidazol-2-yl-1H- indazoles 68a and 68b are potent inhibitors of receptor tyrosine kinase (RTK). 193,194 25
Benzimidazoles have also been examined as potential scaffolds for atypical retinoids.
Eyrolles and co-workers synthesized 4-(5,6,7,8-tetrahydro-5,5,8,8- tetramethylnaphth[2,3-d]imidazol-2-y1)benzoic acids such as 69 , which were shown to induce differentiation in HL-60 human promyelocytic leukaemia cells through binding to RARs.195 Furthermore, Buyukbingol et al . synthesized a number of 2-(5,5,8,8- tetramethyl-5,6,7,8-tetrahydronaphthalen-2-yl)-1H-benzo[d]imidazole-5-carboxylic acids 70a -70c, of which the methyl ester derivative 70b showed the most potent anti- proliferative activity in HL-60 cells. 196,197 Additionally, Hirotaka et al . prepared a series of 2-substituted 1-(3,5,5,8,8-pentamethyl-5,6,7,8-tetrahydronaphthalen-2-yl)- benzo[d]imidazole-5-carboxylic acids 71 that demonstrated RXR-modulating activity. 198
1.5.4 Benzoxazoles as drug scaffolds
The benzoxazole is another useful heterocycle in medicinal chemistry. Similar to benzimidazole, it is found infrequently in natural products, being restricted to few 26
examples such as pseudopteroxazole 72 and seco-pseudopteroxazole 73 ,199 ileabethoxazole 74 ,200 or degradation products of cyclic hydroxamic acids in dried corn tissue, such as 6,7-dimethoxy-2-benzoxazolinone 75 .201
Nevertheless, benzoxazoles have been shown to possess various kinds of biological activities including anti-inflammatory, 202 antitubercular, 203 antileishmanial, 204 anti-
HIV, 205 antimicrobial, 206 antifungal, 207 anti-insomnia and anti-Alzheimer properties.208,209 Several examples of benzoxazole-based pharmaceuticals exist, including calcimycin 76 (antibiotic) and flunoxaprofen 77 (anti-inflammatory). 210,211
Benzoxazoles have also been studied for their anticancer activities. Hou et al . developed the hydroxamic acid 78 as a lead HDACi, which possessed potent cytotoxic activity
212 against 56 tumour cell lines with IC 50 values ranging between 10-540 nM.
Additionally, Easmon et al . prepared several series of 2-benzoxazolyl hydrazones, with the pyrimidine-pyridine derivatives 79 and the methyl-pyrimidine derivative 80 displaying antiproliferative activity against 60 cancer cell lines with IC50 values between 12-447 nM. 213 27
In their study to develop potent RXRα agonists, the group of Haffner synthesized the tetrahydrobenzoxazole-based propenoic acid 81. However, 81 was one of the least potent of the 16 compounds tested in the study.
1.5.5 Quinolines in medicinal chemistry
Quinolines are another important scaffold in medicinal chemistry, appearing in numerous alkaloids such as quinine 82 and skimmianine 83 .214 Quinoline based molecules possess many important biological properties including immunosuppressive, 215 antihelmintic, 216 antibacterial, 217 antifungal, 218 anti- inflammatory and antiviral activities. 219,220 Furthermore, quinolone has been incorporated into many pharmaceuticals, including chloroquine 84 (antimalarial) 221 and pitavastatin 85 (cholesterol). 222 28
In addition to these biological activities, quinolines have also been studied for their anticancer properties.223 The group of Koh developed diarlyamides and diarylureas with potent antiproliferative activity against 10 melanoma cell lines. Of the compounds tested, the 2,3-dihydrobenzo[b][1,4]dioxine analogues 86a and 86b were found to be
224 most effective, possessing IC 50 values of 27-460 nM over six of the cell lines.
Furthermore, Croisy-Delcey et al . synthesized diphenyl quinolones 87a and 87b with cytotoxic activity against the L1210 murine leukemic and MCF-7 human mammary carcinoma cell lines. 225
Quinolines have also been intensely investigated in the field of retinoid related molecules for many years. An in vitro screening of the effects of polycyclic aromatic hydrocarbons against the P19 mouse teratocarcinoma cell line identified several 29
quinolone scaffolds as able to inhibit t-RA mediated activity by 30-60%.226
Furthermore, the appendage of a quinolone residue to the dihydro-dimethylnaphthalene core, commonly found in synthetic retinoids, resulted in the generation of BMS614
(88 ), a selective RARα inverse agonist. 227,228
Eyrolles and his group investigated the differentiation activity of multiple classes of quinolone derivatives 89 -92 against the HL-60 human promyelocytic leukaemia cell line. They found that of the quinolones, only 89a displayed activity, with an ED 50 of 2.0
µM. Interestingly, all of the quinoline derivatives not only displayed activity, the ED 50 values were at least an order of magnitude lower than the quinolones. The strongest
229 performers were 90d and 90e, with ED 50 values of 46 and 20 nM, respectively.
Additionally, Beard et al . developed a series of N-aryl tetrahydroquinolines 93 -96 and studied their retinoid agonistic/antagonist properties. Notably, there was little difference 30
in activity between the alkynyl (93 ) and alkenyl (94) derivatives, both demonstrating high affinity for each of the RAR receptors, with K d values ranging between 3-13 and 5-
15 nM respectively. Changing to quinolone 95 had little effect on selectivity relative to
93 , but increased the K d values to 17-59 nM. Interestingly, utilization of the amide linker (96 ) was found to confer extreme selectivity for RARα (K d 27 nM) over RARβ
230 (K d >1000 nM) and RARγ (K d >1000 nM).
Kikuchi et al . designed pyrrole linked quinolines such as 97a-c with potent and selective RARα antagonistic activity. 231 It was found that methoxy derivative 97b had decreased RARα selectivity relative to 97a, as well as inducing RARβ transactivation.
Additionally, the incorporation of a phenyl group in 97c showed high levels of RARα selectivity, potent antagonistic activity against differentiation and no transactivation of other RAR receptors. 231
31
1.6 Computer-aided drug design
The advent of computer-aided drug design (CADD) techniques has revolutionised drug discovery, as it has allowed for the faster, less expensive and therefore more efficient discovery of bioactive compounds.232 CADD methodologies are frequently applied during the identification and optimisation process of lead compounds, in parallel with experimental techniques. 233-235
CADD methods may be applied to almost any drug design scenario, regardless of whether the three-dimensional structures of a protein target are available or not. 236
Where this structural information is absent, ligand-based approaches such as pharmacophore modelling and quantitative structure-activity relationship (QSAR) methods can be used to predict the activities of compounds that have not been tested experimentally. 233 Where the structure of a known target or a homologous protein is available, CADD methods can be used to predict ligand binding modes (docking) or protein structures (homology modelling), aiding the de novo design of novel compounds. 236 The use of computational techniques allows researchers to focus synthetic efforts, prioritise compounds for biological testing, or visualise possible molecular processes.
1.6.1 CADD techniques in atypical retinoid design
The most common role of computational techniques in retinoid design is in attempts to explain the activity of synthesized compounds, with many groups using computational docking methods to describe the potential binding modes of their ligands to the target
RAR(s). 163,237-239 However, the use of docking in preliminary screening studies or as a predictive technology has had relatively low application to date. 32
Silva and co-workers performed virtual screening of approximately 350,000 molecules from the Available Chemical Directory (ACD) database in order to identify new ligands for RAR isotypes. Using density functional calculations, natural bond orbital charges and multiple docking procedures, they identified three scaffolds 98 -100 as potential
RAR ligands. Additionally, by comparison with the binding modes of other RAR ligands, they proposed that 98 could be a pan-agonist, while 99 and 100 could be
RARγ -selective, as they have an increased number of H-bonds and hydrophobic interactions relative to the control ligands. 240,241
In a similar study, Stebbins et al . used a combination of scoring functions (FlexX,
Screening, Chemscore and DJ score) in a virtual screening campaign of a 50,000- compound subset of the Chem Bridge library to identify novel RXRα inhibitors 101-103.
The 200 top scoring compounds were first verified by competitive ligand binding studies, and subsequent in vitro transcriptional activation assays and receptor expression inhibition assays confirmed the RXRα inhibitory activity of 101-103.242
33
1.7 Thesis aims
The increasing frequency of cases of intrinsic or acquired retinoid resistance, 109 as well as the significant toxicities associated with conventional retinoid use, 101 highlights the need for the further development of synthetic retinoids that act through novel and selective mechanisms. Considering the limited chemical space that has been explored in the development of these drugs to date, the aim of this thesis is to develop novel atypical retinoids based on scaffolds identified by CADD methods. The new heterocyclic scaffolds that were identified over the course of the studies then became the subject of focus, due to their promising RAR activities.
Chapter 2 describes the identification of novel scaffolds through the in silico screening of a virtual library against the crystal structure of the RARβ -LBD, the subsequent in vitro validation of these scaffolds and the further development of predictive computational docking methods based on these results.
Chapter 3 explores the design, synthesis and in vitro evaluation of indole- benzothiazole acetamides capable of modulating RAR expression.
Chapter 4 covers the design, synthesis and in vitro evaluation of indole azoles as novel chemotherapeutics.
Chapter 5 details the design, synthesis and in vitro evaluation of benzothiazole- thieno[2,3-c]pyrazole carboxamides as RAR modulators.
Chapter 6 investigates the design, synthesis and in vitro evaluation of 8- isopropylquinolines as selective RARα antagonists or dual -action RARα antagonists/
RARβ agonists.
34
CHAPTER 2: DEVELOPMENT OF COMPUTATIONAL DOCKING
STUDIES FOR THE IDENTIFICATION OF RARβ LIGANDS
2.1 Introduction
The development of a new drug is a very complex process, requiring the collaboration of experts from a range of fields, employing many tools to identify lead candidates for further development. 243 This critical process is therefore very costly in terms of both money; costing in excess of $2.5 billion per drug, and time; typically taking 2-8 years for clinical trials and FDA approval.244 The introduction of computational-based methodology, capable of pre-screening candidate compounds to predict biological activity of compounds prior to their synthesis, has assisted medicinal chemists in streamlining this lengthy and arduous task. With computers rapidly increasing in processing power, the amount of time necessary to perform such preliminary analysis may be reduced, while providing high quality output, thereby reducing the number of unsuccessful hits, speeding up optimization of potential lead compounds and saving both time and money that would otherwise be spent on synthesis and biological assays. 245
Virtual screening is typically implemented in one of two ways: ligand-based or structure-based methods. Ligand-based methods analyse sets of structures with a common receptor target in order to identify common features that may be responsible for the ligand binding and subsequent target activation. The nature and 3-dimensional organisation of these features may then be mapped to generate a pharmacophore, against which novel ligands may be screened to identify ligands that will potentially interact with the target. Additionally, the mapping of such features can also be used in 35
the generation of QSAR models, whereby the mapped features can be correlated to the experimentally determined binding of similar ligands to the target receptor. The main advantages of using ligand-based screening methods is that they do not require any structural information about the target receptor, making them applicable to almost any screening scenario, and that they are a relatively quick and computationally inexpensive method, requiring a fraction of the time and computational cost of many structure-based methods. By comparison, structure-based methods rely on having access to the three- dimensional structure of the target receptor, either through NMR studies, X-ray crystallography or homology modelling of an analogous receptor. These methods then allow the docking of scaffolds or fragments, either as part of an existing or novel library, to an identified or proposed binding site and the fit to be determined by scoring functions based on various levels of theory depending on the method selected. These scoring functions can then be used to compare proposed structures to natural ligands or known binders, allowing for the prioritization of ligands for synthesis and further evaluation. Furthermore, such methods allow researchers to visualize their scaffolds within the binding pocket, allowing them to identify important features such as specific interactions between the ligand and receptor, amino acids that may potentially form interactions with functional groups, additional cavities that may be filled by appending groups and conformational restrictions that may confer more favourable binding modes to the ligand.
With the rapid growth of X-ray crystallography and NMR spectroscopy in generating quality 3D-models of protein structures, the in silico screening of compound collections, such as the National Cancer Institute (NCI) database and commercial compounds databases, represents an increasing trend in academic and industrial research towards the fast and efficient generation of focused compound libraries for specific protein 36
targets. 246 These virtual screening, as well as other molecular modelling techniques, have been widely applied in the pharmaceutical industry, contributing to the development of drugs such as Capoten, 247 Trusopt, 248 Viracept, and Relenza, 249,250 as well as anti-cancer agents Gleevec 104 and Tarceva 105.251,252
A number of different software packages have been developed to perform such virtual screens, relying on different levels of theory (molecular dynamics, molecular mechanics). Typically, such programs consist of two fundamental components: sampling and scoring. Sampling involves the generation of supposed ligand conformations near a protein binding site, and can be further subdivided into two aspects; ligand sampling and protein flexibility, while scoring is the prediction of the binding tightness for individual ligand conformations with a physical or empirical energy function. 253 However, such programs are still limited in their capacity as the output results depend heavily on the quality of the 3D-protein structure, the specific algorithms used to obtain the results and the power of available computational resources. Examples of common protein-ligand docking software packages are detailed below (Table 2.1). 254
37
Table 2.1: Common docking tools for protein-ligand virtual screening. 254
Method Ligand Sampling Scoring function DOCK incremental build force field or contact score FlexX incremental build empirical score Slide conformational ensembles empirical score FRED conformational ensembles Gaussian or empirical score GOLD genetic algorithm empirical score Glide exhaustive search empirical score Autodock genetic algorithm force field Ligandfit Monte Carlo empirical score ICM Pseudo-Brownian mixed force field & & local minimization empirical score QXP Monte Carlo force field
2.2 Preliminary library screening
In order to efficiently identify novel scaffolds as potential ligands for RARβ, an in silico screening involving computational docking methods and energy calculations was conducted through Quantum Pharmaceuticals. Small molecules from the 1,000,000 compound Enamine Ltd library were used, with all compounds conforming to
Lipinski’s rule of five. 255 3-D structural information of the RARβ -LBD, obtained from
X-ray crystallography, was accessed via the protein data bank (PDB ID: 1XAP). 129 This virtual screen returned a series of 26 compounds proposed to have suitable binding affini ties to RARβ, shown in table 2.2 below.
38
Table 2.2: The structures and predicted binding affinities of the proposed ligands.
Ligand Structure Predicted
Kd (M)
T0500 -6280 2.1 x 10-07 (L1)
T5302411 7.8 x 10-07 (L2)
T5321905 4.3 x 10-08 (L3)
T5347410 5.1 x 10-07 (L4)
T5392521 7.7 x 10-07 (L5)
T5446398 6.2 x 10-07 (L6)
T5449679 2.1 x 10-07 (L7)
39
T5460017 6.8 x 10-07 (L8)
T5481961 1.7 x 10-07 (L9)
T5497721 1.9 x 10-08 (L10)
T5514418 2.9 x 10-07 (L11)
T5578882 3.7 x 10-07 (L12)
T5990076 2.1 x 10-08 (L13)
T5628388 5.8 x 10-08 (L14)
T5682216 1.3 x 10-07 (L15)
40
T5693687 6.6 x 10-07 (L16)
T5729414 6.6 x 10-08 (L17)
T5793743 2.1 x 10-07 (L18)
T5822440 2.7 x 10-07 (L19)
T5950497 5.9 x 10-07 (L20)
T5999398 7.9 x 10-07 (L21)
T6025681 8.6 x 10-07 (L22)
T6028969 2.1 x 10-07 (L23)
T6030571 2.9 x 10-07 (L24)
41
T6033194 3.8 x 10-07 (L25)
T6043628 2.8 x 10-07 (L26)
In order to determine if these compounds possessed any cytotoxic activity, they were screened against BE(2)-C neuroblastoma cells, stably transfected with the MEP-4
85 expression vector to express either the empty vector (E.V.) or the RARβ 2 gene. The ligands were tested at a fixed concentration of 5 M over 96 h, after which the proportion of viable cells was determined by Alamar blue assay (n = 2) relative to solvent only (DMSO) control. Following a comparison of the effect in the two cell lines, it was observed that of these 26 compounds, 3 appeared to be more effective in the E.V. cell line (L11, L12 and L21), 2 had indistinguishable activity (L1 and L26) and the remaining 21 were more effective in the RARβ cell line (Figure 2.1).
Empty Vector RAR b 1 2 0
1 0 0
8 0
6 0
4 0
Cell Viability2 0 (% )
0 7 D L 1 L 2 L 3 L 4 L 5 L 6 L 7 L 8 L 9 L 1 0 L 1 1 L 1 2 L 1 3 L 1 4 L 1 5 L 1 6 L 1 L 1 8 L 1 9 L 2 0 L 2 1 L 2 2 L 2 3 L 2 4 L 2 5 L 2 6
Figure 2.1: Cell viability (± SEM) of transfected BE(2)-C cells following 96 h
treatment with 5 M of ligand. 42
Of the 21 ligands that displayed increased cytotoxic activity towards the RARβ overexpressed cell line, it was found that 8 displayed cytotoxic activity with a differential of more than 20% between the RARβ overexpressed and E.V. cell line , indicating that these compounds may indeed exert their activity through the RAR β pathway (Figure 2.2). If we consider these 8 compounds as the leads identified by these studies, we can calculate that the virtual screening provided a hit rate of 31% for compounds displaying cytotoxicity to both cell lines, with at least 20% increased efficacy towards the RARβ overexpressed ce ll line. However, of these 8 ligands only
L10 displays a more marked differential between the two cell lines (>30%), therefore giving a hit rate of only 4% for highly potent and selective compounds.
Empty Vector RAR b 2
1 2 0
1 0 0
8 0
6 0
4 0
Cell Viability2 0 (% )
0 D L7 L9 L10 L13 L14 L15 L19 L23
Figure 2.2: Cell viability (± SEM) of BE(2)-C cells following 96 h treatment with 5 M
of top 8 ligands.
In order to determine if the predicted binding affinity (K d) of these 8 hit ligands could be correlated to their in vitro activity, a comparison was made between the percentage of viable cells following treatment and the K d value (Figure 2.3). Unfortunately, following a Pearson style analysis, it was observed that there was no significant linear correlation (r = 0.73, p = 0.10) between the in vitro activity of the ligands and the affinity predicted by the manufacturer. Furthermore, it was observed that of the five 43
-7 ligands with K d values less than 1.0 x 10 (L3, L10, L13, L14 and L17), two of these,
L3 and L17, did not show sufficient activity to pass the initial validation. Furthermore, this comparison is limited in that the Kd values are only predicted and not experimental, relying solely on the computational methods utilised by the company, as well as only covering a small range, being 2.51 x 10 -7 M for the top 8 hits and 8.41 x 10 -7 M for the entire 26 ligands.
1 0 0
8 0
6 0
4 0
2 0 Cell Viability (% )
0 0 1 .0 0 1 0 -7 2 .0 0 1 0 -7 3 .0 0 1 0 -7 K d (M)
Figure 2.3: Comparison of predicted binding affinity (K d) and observed in vitro
activity.
Considering the structures of these 8 lead molecules (Figure 2.4), it is apparent that due to the large structural differences, comparisons of other physical properties, such as cLog P, would not be able to further enhance an understanding of these compounds ’ activity. Furthermore, none of these compounds bears a carboxylic acid derivative, which is one of the key features of every retinoid derivative reported in the literature thus far. The closest functionality to this is the tetrazole present in L14, which is a common isostere for the carboxylate group. If these compounds, or subsequent analogues, were found to bind to an RAR, they would be the first non-carboxylate derivatives found to do so. However, the available information is unable to explain how these compounds would facilitate such binding and activation of the receptor, as no data 44
regarding the orientations of the ligands, or their interactions with the RARβ -LBD residues was provided. Such information is necessary not only to rationalize these results, but to assist in the development of analogues that are more potent in terms of their cytotoxicity, as well as their selectivity for t he RARβ cell line. Consequently, the development of in-house computational docking methods was investigated as an alternative. For the current work, the commercial protein-ligand docking engine GOLD was accessed through the Discovery Studio software.
Figure 2.4: Structures of the 8 lead molecules
2.3 The GOLD software package
The GOLD software package (Cambridge Crystallographic Data Centre, CCDC) is a popular docking program that employs a genetic algorithm to determine the best 45
binding mode of a ligand. 256 Such genetic algorithms mimic the process of evolution by following the principle of the survival of the fittest.257 A population of possible protein- ligand complex conformations (chromosomes) is initially created randomly and the fitness of each individual is evaluated based on a scoring function. Subsequently, random individuals (biased towards high fitness) are modified to provide new docking solutions through operations including mutations (small modifications to the protein- ligand conformation) and cross-overs (combination of two protein-ligand conformations). The efficiency of this process can be improved through the use of an island model, where the population is divided into sub-populations that develop independently, but are able to migrate to exchange genes. The use of an island model was found to improve the efficiency of the algorithm. 256 New docking solutions
(children) then replace the least-fitting members of the population niche they belong to, which is defined as a cluster of ligand poses where all H-bond donors and acceptors are within a 1.0 Å root-mean-square distance (RMSD). Repetition of this process then progresses the population towards the optimal solution.
The Goldscore fitness function consists of four components: protein-ligand hydrogen bond (external H-bond) and van der Waals (vdW) energies (external vdW), as well as a ligand-internal vdW (internal vdW) and torsional strain energies (internal torsion). 256
When flexible side-chains are used, a protein conformation score is added. The fitness score is the negative sum of the component scores. Weighting factors of the individual components can be applied. By default, the external vdW score is multiplied by 1.375, an empirical value which encourages hydrophobic protein-ligand contacts.
46
2.4 The RARβ protein
The crystal structure of the RARβ bound to the pan -RAR agonist TTNPB 13 was solved by Germain et al (PDB ID: 1XAP, Figure 2.5), enabling a much more efficient approach to developing computational screening methods towards selective RAR ligands by eliminating the need to construct homology models and develop pharmacophores. 129 The resolution obtained was 2.10 Ǻ, with an R-value of 0.312 and
R-free of 0.253, indicating an accurate representation of the crystal structure, which is necessary to construct accurate computational models.
Figure 2.5: The RARβ – TTNPB complex and the structure of TTNPB 13 . 47
The crystallized protein is truncated from the N-terminus, incorporating residues
Ala177 – Asn409, which make up part of the hinge domain and the entire ligand binding domain. Inspection of the ligand binding reveals two key polar residues involved in binding the deprotonated TTNPB 13 , namely Ser280 and Arg269.
Furthermore, a series of predominantly non-polar amino acids were identified as important residues, namely Phe221, Ala225, Leu259, Leu262, Ile266, Phe295, Arg387,
Val388 and Met406 (Figure 2.6).
Figure 2.6: Reported interactions for the RARβ LBD -TTNPB complex. 129
2.5 Establishing a docking protocol
All structures (protein and ligands) were constructed/edited with DS visualizer
(Discovery studio modelling, Accelrys Software Inc.). The protein was prepared by removing the TTNPB ligand and bound water molecules. Hydrogen atoms were added to the protein and the structure processed using the CHARMm forcefield. Ligand structures were prepared in a similar fashion. 48
The appropriate binding site was determined by detecting cavities, using a site opening of 5 Ǻ and grid resolution of 0.5 Ǻ, which identified an appropriate cavity of size 3138 point units. Based on the observed interactions in the RAR LBD-TTNPB complex, a subset of amino acids was selected to be considered as flexible during the docking protocol, namely Phe221, Ala225, Leu259, Leu262, Ile266, Phe295, Arg387, Val388 and Met406.
The TTNPB ligand was docked into the defined binding site using the Goldscore function, calculating 10 dockings, permitting flipping of ring corners, amide bonds, planar and pyramidal nitrogens and protonated carboxylic acids, and allowing intramolecular hydrogen bonds. The docking process resulted in different poses with the relevant docking scores used for comparison of the different binding poses obtained. It was observed that in this docking (purple), the predicted pose was quite different from that observed in the published complex (yellow). Not only was there significant rotation along multiple bonds, exemplified by heavy atom RMSD values of 0.43 – 2.44 Ǻ (0.95
Ǻ average) , the hydrogen bonding was only observed to one of the carboxyl oxygen atoms, not both as was observed previously (Figure 2.7). 49
Figure 2.7: Preliminary docking of TTNPB (green) results in significant
conformational changes and reduced H-bonds compared to the complex derived from
X-ray crystallography (yellow).
In order to procure a more accurate docking protocol, TTNPB was re-docked using the same protocol with a modified set of flexible amino acids, replacing Ala225 with
Arg269 and Ser280, the pair of residues responsible for the hydrogen bonding between
TTNPB and the protein (Figure 2.8). This set of flexible residues gave much stronger correlation to the literature conformation with highly conserved torsion angles and hydrogen bonding (Table 2.3), as well as RMSD values of 0.23 – 0.72 Ǻ (0.39 Ǻ average) and an average Goldscore of 72.65. 50
Figure 2.8: Modified docking of TTNPB (green) shows consistent binding with the
complex derived from X-ray crystallography (yellow).
Table 2.3: Comparison of torsion angles between the docked TTNPB and X-ray derived structure.
Pose θ1 θ2 Literature 34.65° -158.79° Experimental 34.91° -138.20°
The nature of TTNPB itself was further explored by repeating the docking with the protonated carboxyl group. This returned a greatly diminished hydrogen bonding profile, as well as vastly different conformation with RMSD values of 0.73 – 2.60 Ǻ
(1.86 Ǻ average) and a maximal Goldscore of 50.17, indicating that the deprotonated form is indeed the best ligand for comparison (Figure 2.9). 51
Figure 2.9: Docking of protonated TTNPB (purple) shows significantly altered binding
and reduced H-bonding compared to the complex derived from X-ray crystallography
(yellow).
2.6 Validation of the protocol with the lead compounds as a training set
In order to further validate the docking protocol, it was decided that a comparison of the generated Goldscores and biological activity of the original hit compounds would be the next logical step. Therefore, the eight biologically active hit compounds, along with the two least biologically active compounds, were docked into the RARβ LBD using the established protocol with modified flexible amino acid subset, calculating 100 poses.
Table 2.4 below details the distribution of the poses determined by their similarity within a given RMSD value (Å), the total number of clusters (No.) and the number of poses in the largest cluster. It also describes three potentially useful Goldscore values for analysing the poses; highest individual score of a pose, highest score of the largest cluster, and the average score of the largest cluster.
52
Table 2.4: Analysis of hit ligand docking results.
Ligand Clustering Goldscore Poses in Highest Cluster Cluster (Å) No. largest overall highest average T5449679 (L7) 1.98 25 11 71.39 63.70 54.46 T5481961 (L9) 1.99 24 22 58.37 54.89 45.34 T5497721 (L10) 1.99 14 19 70.23 63.78 52.89 T5990076 (L13) 2.00 27 17 66.58 66.58 56.19 T5628388 (L14) 2.00 25 15 63.69 57.97 46.80 T5682216 (L15) 2.00 22 17 67.19 67.19 52.61 T5822440 (L19) 1.89 35 12 61.53 54.47 47.69 T6028969 (L23) 2.00 14 37 58.82 58.82 48.09 T6025681 (L22) 1.99 37 9 64.41 53.79 43.88 T6030571 (L24) 2.03 15 49 55.05 55.05 48.26
It is possible to rank the ligands in three ways based on different analysis of the docking outputs. The first would be to rank them based on the overall highest score achieved, the second would be to rank them based on the highest score for the largest cluster, and the third would be to rank them by the average score of the largest cluster. Table 2.5 below details the rankings of each of the compounds based on the biological data and the three computational methods.
53
Table 2.5: Various ranking methodologies for the hit ligands.
Ligand Ranking Biological Highest Cluster Cluster activity overall highest average T5497721 (L10) 1 2 4 3 T5481961 (L9) 2 9 8 9 T5628388 (L14) 3 6 6 8 T5990076 (L13) 4 4 2 1 T6028969 (L23) 5 8 5 6 T5682216 (L15) 6 3 1 4 T5449679 (L7) 7 1 3 2 T5822440 (L19) 8 7 9 7 T6030571 (L24) 9 10 7 5 T6025681 (L22) 10 5 10 10
Initially, it was assumed that ranking the ligands based on the average score of the largest cluster would most closely reflect the ranking of the biological activity. This assumption was based on the thought that the average of the poses in the largest population would be a better representation of the actual state of the ligand, compared to any single orientation represented by a single score. However, upon comparing the difference between the rankings based on in vitro activity to any of the computational methods, it was found that ranking based on the highest score of the largest pose was marginally better than the other two methods (Table 2.6). On average, this method ranks the ligands closer to the correct rank, having a mean of 2.6 places difference, a median difference of 2.5, being at most 6 ranks off target and having 50% of rankings differing by ≤ 2. Considering this, the ranking method based on highest Goldscore of the largest cluster was selected as the most appropriate method for further investigations in this study. Interestingly, upon a Spearman style analysis of the ranking methods versus the biological activity, it was observed that there was no significant correlation for any 54
method (Table 2.6). However, ranking the ligands based on the highest score of their largest cluster did return the highest r s value (0.37) and lowest p value (0.30), further supporting this as the strongest of the three ranking methods.
Table 2.6: Differences between predicted and observed rankings.
Ligand Absolute Difference Highest Cluster Cluster overall highest average T5449679 (L7) 6 4 5 T5481961 (L9) 7 6 7 T5497721 (L10) 1 3 2 T5990076 (L13) 0 2 3 T5628388 (L14) 3 3 5 T5682216 (L15) 3 5 2 T5822440 (L19) 1 1 1 T6028969 (L23) 3 0 1 T6025681 (L22) 5 0 0 T6030571 (L24) 1 2 4 Mean 3 2.6 3 Median 3 2.5 2.5 Range 7 6 7
rs 0.15 0.37 0.19 p 0.68 0.30 0.61
For the analysis of docking results for novel scaffolds, a cut-off Goldscore of 56.00 for the highest score of the largest cluster has been proposed to differentiate between potentially active and inactive scaffolds (Figure 2.10). This value was selected because it excludes both biologically inactive ligands (L22 and L24), while including six of the eight active ligands in the appropriate region. This cut-off does however exclude L9 and
L19, despite L9 being the second most biologically active compound. Alternatively, a higher cut-off of 60.00 could be employed; however this would then exclude L14 and 55
L23 from the active region, which were the third and fifth most biologically active compounds, respectively.
7 0 Active region
6 0
G o ld5 s c o6 r e
Inactive region 5 0 L7 L9 L10 L13 L14 L15 L19 L23 L22 L24
L ig a n d
Figure 2.10: Proposed separation of active versus inactive Goldscores.
2.7 Conclusions
A comparison of the biological activity and predicted RARβ affinity of the initial 26 ligands demonstrated no significant correlation. A computational docking protocol utilizing the GOLD software package and RARβ crystal data was developed and subsequently employed to generate docking scores for the eight most active and two least active of the ligands. The closest matching between the observed biological data and Goldscore was obtained through comparison of the highest score of the largest cluster. A cut-off Goldscore of 56.00 was proposed to differentiate between potentially active and inactive novel scaffolds in the following chapters.
56
CHAPTER 3: DESIGN, SYNTHESIS AND CHARACTERIZATION
OF INDOLE-BENZOTHIAZOLE ACETAMIDES
3.1 Introduction
Amides are a common feature of many natural and synthetic compounds. In biology, the amide bond forms the structural link between the amino acids that form proteins. The delocalization of the nitrogen lone pair electrons gives the amide bond partial double- bond character between the nitrogen and carbonyl (Figure 3.1), making amides less basic than amines and also less susceptible to hydrolysis than esters. Furthermore, this partial double-bond character results in a planar configuration around the nitrogen atom, giving rise to cis - and trans -amide geometries, with the trans isomer being the dominant form. This biased geometry, as well as the formation of H-bonds, confers structure to proteins and polypeptides.
Figure 3.1: Delocalization of nitrogen lone pair electrons in amides.
Amide derivatives of 3-indoleacetic acid (IAA) 106 , a naturally occurring plant auxin, are an important class of biologically active molecules. 258 Cysmethynil 107 and its analogues display antitumor activity,259 while N-(pyridin-4-yl)-(indol-3-yl)acetamides
108 are antiallergic agents and 2 (1 H indol 3 yl) N phenylacetamides 109 possess antioxidant activity. 260,261 Ǧ Ǧ Ǧ Ǧ Ǧ Ǧ 57
Derivatives of 2-aminobenzothiazole 110 have also been intensely studied as a particularly important subset of benzothiazoles, with the combination of 110 with other heterocycles being a common approach in the design of new drug-like molecules. A number of amide derivatives with anti-cancer activity include the 1,3,4-thiadiazole-2- thione conjugate 111, and benzothiazolyl urea 112.262,263
The development of indole and benzothiazole conjugates has generated scaffolds with remarkable biological activities. Piperazine linked systems such as 113 have shown selective agonistic activity towards dopamine D2 receptors, with useful antioxidant and anti-Parkinson’s disease activity. 264 Also, the benzyl linked 3-indoleacetamide 114 inhibits macrophage nitric oxide (NO) production by suppression of the iNOS protein and mRNA expression, while also suppressing the expression of COX-2 through NF-κB inactivation. 265 58
Considering the important activities of 3-indoleacetamides, 2-aminobenzothiazoles and their conjugates, this chapter is concerned with the development of amide based indole- benzothiazole hybrids as novel scaffolds with anti-tumour activities.
3.2 Identification of indole acetamides as a lead scaffold
An analysis of the 26 proposed ligands revealed that amide linkages were a common feature, being present in 15 of the ligands, as well as in 4 of the top 8 structures. Benz- fused heterocycles were also a common feature, being present in 10 of the 26 ligands and 3 of the top 8 structures. Interestingly, the most common types of benz-fused heterocycles in the 26 ligands were derivatives of 2-aminobenzimidazoles or 2- aminobenzothiazoles, which were also the moieties present in L10 and L15, the 1 st and
4th most active compounds respectively (Figure 3.2).
Figure 3.2: Lead structures showing amide linkages (blue) and benz-fused heterocycles
(red). 59
Examination of the docking results for L10 and L15 revealed that they displayed a similar orientation within the RARβ -LBD (Figure 3.3). It was observed that both ligands directed the benz-fused heterocyclic moiety towards the Arg269 and Ser280 residues. Furthermore, each ligand displayed several favourable interactions with the receptor residues. In the case of L10, the benzimidazole NH was observed to donate an
H-bond to the backbone carbonyl of the Leu262 residue, while the imidazole and pyrazole rings formed a π -π and π -σ interaction with the Phe221 and Leu259 residues res pectively. Additionally a π -cation interaction was also observed between the benzimidazole phenyl ring and the Arg265 residue. In the case of L15, the benzothiazole heterocycle was observed to form π -σ interactions with the Leu262 residue, while the termin al phenyl ring formed a π -cation interaction with the Arg387 residue. Furthermore, the benzimidazole NH also formed an intramolecular H-bond with the phenylacetyl carbonyl group, generating a 7-membered ring. This result suggests that the aromatic interactions are the dominant contributors to these binding modes.
Figure 3.3: Overlapped docking poses of L10 (green) and L15 (yellow) showing
similar orientations within the RARβ -LBD. 60
In order to determine appropriately-sized starting structures, indole amide scaffolds with linkers of various lengths (Table 3.1) were docked into the RARβ binding pocket following the method detailed previously (Chapter 2). 3-Indolealkylamides were selected as they possessed both the aromatic phenyl moiety of L15, as well as the heterocyclic feature of L10, and also because the 3-indolealkylamides are well established as bioactive moieties in the literature, as discussed previously in Chapters
1.5.1 and 3.1. Additionally, the use of varied alkyl chain lengths may allow the compounds to orientate themselves in a similar manner to L10 or L15, depending upon the length. Furthemore, substitution of the benzothiazole moiety for benzimidazole was also examined.
Table 3.1: Goldscores of the initially proposed scaffolds.
Ligand Goldscore Rank
66.72 2
63.73 3
66.81 1
58.11 4
Analysis of the docking results suggest that the combination of a 3-indoleacetamide moiety with either a benzothiazole (115), or a benzimidazole ( 116), would give ligands 61
with similar affinity for the RARβ -LBD, having docking scores of 66.72 and 63.73, respectively. This agrees with the prevalence of these heterocycles in the initial 26 ligands, with four occurrences of a benzothiazole and one occurrence of a benzimidazole. Gratifyingly, these scores were highly similar to those achieved by the lead compounds L10 and L15, which had docking scores of 63.78 and 67.19 respectively. It was also observed that the binding modes of 115 and 116 were quite similar, with both analogues directing the indole ring towards the hydrophobic RA Rβ cavity, and with the benzothiazole or benzimidazole moiety directed towards the
Arg269 and Ser280 residues (Figure 3.4). Additionally, the docking also suggests that the propionamide 117 is similar to acetamide 115 in terms of its binding orientation and binding affinity (Goldscore 66.81). However, carboxamide 118 showed a marked reduction in its Goldscore (58.11), suggesting that analogues of this size are incapable of filling the large binding cavity of RARβ.
Figure 3.4: Overlapped docking poses of 115 (green), 116 (yellow) and 117 (blue)
showing similar orientation s within the RARβ -LBD.
62
3.3 Synthesis of indole-benzothiazole amides
Coupling reactions are one of the most commonly used methodologies for the formation of amide bonds. They possess many advantageous properties over other amide forming reactions, such as their high rate of reaction , their ability to achieve near quantitative yield, the use of milder reaction conditions, the reduced incidence of substrate racemization, a lack of side reactions and the formation of easily removable side products. 266 l-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDCI) is a popular coupling reagent that has been used in the synthesis of both indole and benzothiazole based amides. The Wu group treated a DCM mixture of indole-acid 119 and cyclohexylamine 120 with EDCI and 4-(dimethylamino)pyridine (DMAP), giving the anti-obesity indole-amide 121 in 50% yield (Scheme 3.1). 267
Scheme 3.1: Reagents and conditions: a) EDCI, DMAP, DCM.
Similarly, Kamal et al . treated a mixture of pyrazolo[1,5-a]pyrimidines 122, EDCI and
1-hydroxybenzotriazole (HOBt) with 2-aminobenzothiazoles 123 at 0 °C, followed by stirring at r.t. to give benzothiazole-amides 124 in yields of 72-85% (Scheme 3.2). 268
Scheme 3.2: Reagents and conditions: a) EDCI, HOBt, DCM, 0 °C - r.t. 8 h. 63
N,N-Dicyclohexylcarbodiimide (DCC) is another commonly used coupling reagent that has also been used in the synthesis of indole and benzothiazole amides. Duflos et al . heated a mixture of indole acid 125 and 4-aminopyridine 126 with DCC at reflux in
THF for 48 h. Following work-up, the N-pyridinyl-indole-3-(alkyl)carboxamides 127 were isolated in 58-66% yield (Scheme 3.3).269
Scheme 3.3: Reagents and conditions: a) DCC, THF, reflux, 48 h.
Similarly, Esteves et al . added 2-aminobenzothiazoles 128 to a DMF mixture of aspartic acid ester 129, DCC and HOBt at 5 °C. After 24 h at r.t., amides 130 were isolated in yields of 58 –91% (Scheme 3.4).270
Scheme 3.4: Reagents and conditions: a) DCC, HOBt, DMF, 0 °C – r.t., 24 h.
Aside from carbodiimide-based coupling reagents, the use of (benzotriazol-1- yloxy)tripyrrolidinophosphonium hexafluorophosphate (PyBOP) has also seen increased use throughout the literature. PyBOP has several advantages over earlier generation coupling reagents, such as the generation of a non-carcinogenic tripyrrolidinophosphate by-product, compared to the carcinogenic tris(dimethylamino)phosphate by-product produced by the use of the earlier
(benzotriazol-1-yloxy)tris (dimethylamino)phosphonium hexafluorophosphate (BOP) reagent, as well as a simpler work-up procedure and the ability to tolerate more difficult 64
coupling substrates. 271 Thompson et al . coupled 106 with anilines 131 by treating them with PyBOP in the presence of N,N-diisopropylethylamine (DIPEA) in chloroform
(Scheme 3.5). After 18 h, indolyl-phenylacetamides 132 were isolated in yields of 36-
73%.
Scheme 3.5: Reagents and conditions: a) PyBOP, DIPEA, CHCl 3, 18 h.
Similarly, Zhang et al . coupled acids 133 with 110 using PyBOP in the presence of
DIPEA in a DMF solution, giving amides 134 in yields of 65-87% (Scheme3.6).272
Scheme 3.6: Reagents and conditions: a) PyBOP, DIPEA, DMF .
Furthermore, previous work in our group has demonstrated that PyBOP is an effective and efficient reagent for the coupling of IAA 106 and 2-aminobenzothiazole 110, generating acetamide 115 in 53% yield after only 2 h stirring in the presence of DIPEA in DCM (Scheme 3.7).273 Consequently, PyBOP was selected as the primary means for the generation of indole-benzothiazole acetamides in this work.
65
Scheme 3.7: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h .
3.3.1 Synthesis of mixed heterocyclic acetamides
As the desired benzothiazole-based acetamide 115 was already at hand, it was decided to prepare a series of acetamides with various combinations of heterocycles to determine the SAR for this class of compounds. Firstly, the necessity of the indole heterocycle was investigated by substituting it with a naphthalene moiety. Replacement of the indolyl pyrrole ring with a benzene ring generates a more lipophilic molecule, which may increase bindin g affinity for the hydrophobic RARβ binding cavity.
Additionally, the lack of a hydrogen bond donor may have additional consequences for binding. Finally, by utilizing both the 1-naphthyl and 2-naphthyl acetic acids, the importance of the naphthalene moiet y’s directionality may also be explored. Hence, 2- aminobenzothiazole 110 was added to naphthaleneacetic acids 135a-b and PyBOP in
DCM, and after stirring for for 2h give acetamides 136a-b in yields of 37 and 41% respectively (Scheme 3.8).
Scheme 3.8: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h . 66
As an additional modification to the acid portion of the molecule, 3,4-
(methylenedioxy)phenylacetic acid 137 was also considered. This moiety has the heterocycle directed away from the amide linkage, as well as having two H-bond acceptors, as opposed to the H-bond donor of the indole. Therefore, 110 was coupled to
137 using PyBOP, generating amide 138 in a yield of 46% (Scheme 3.9).
Scheme 3.9: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h .
The identity of 136a, 136b and 138 were confirmed primarily through spectroscopic analysis. The 1H NMR spectra displayed consistent aromatic patterns with respect to the benzothiazole ring, while each aromatic acid displayed the expected splitting patterns
1 and integrations. As an example, the H NMR (d 6-acetone) spectrum of 136a showed the benzothiazole peaks as two doublets of doublets at δ 7.90 and 7.71 ( J = 0.7, 7.9 Hz) corresponding to H4 and H7, and two doublet of doublet of doublets at δ 7.42 and 7.29
(J = 1.3, 7.9, 7.9 Hz) corresponding to H6 and H5. The naphthyl CH 2 was identified as a singlet at δ 4.47, while the naphthyl aromatic peaks were identified as two doublets at δ
7.88 (J = 8.2 Hz, H4′) a nd 7.62 ( J = 6.7 Hz, H7′), three doublet of doublets at δ 8.19 (J
= 0.6, 7.4 Hz, H8′), 7.94 ( J = 2.0, 7.4 Hz, H5′) and 7.49 ( J = 8.3, 8.3 Hz, H3′ and H6′) and a doublet of doublet of doublets at δ 7.55 (J = 1.7, 7.2, 7.2 Hz, H7′).
It was also desired to investigate the SAR of the amine portion of the molecule. As benzimidazoles and benzoxazoles were also present in the initial 26 ligands, these heterocycles were targeted first. Not only would this allow for comparisons back to the lead structures, but it would also investigate the potential importance of hydrogen- bonding by substitution of the sulfur atom with the donor-capable NH group of 67
benzimidazole, or the acceptor-capable oxygen atom of benzoxazole. Furthermore, the resultant changes to polarity may also have an effect on solubility and efficacy of the compound. Therefore, 106 was reacted with 2-aminobenzimidazole 139a or 2- aminobenzoxazole 139b according to the PyBOP procedure developed earlier (Scheme
3.10). In the case of 139b, acetamide 140 was isolated in 62% yield. However, the reaction with 139a did not produce acetamide 116, even after extended durations up to
48 h, possibly due to the low solubility of 139a in DCM. The use of acetonitrile or DMF as solvents also did not afford the desired 116. However, the use of EDCI and HOBt in place of PyBOP generated 116 in a yield of 37% (Scheme 3.10).
Scheme 3.10: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h ; b) EDCI, HOBt,
DMF, 24 h.
The 1H NMR spectra of 116 and 140 both displayed consistent patterns for the indolyl moiety, with different splitting patterns for their benz-fused heterocycles as a result of the electronic effects of their varied heteroatoms. Singlet peaks were observed at δ
1 11.94, 11.63, 10.90 and 3.86 in the H NMR (d 6-DMSO) spectrum of 116, assigned as the benzimidazolyl NH, indolyl NH, amide NH and CH 2 protons respectively. Doublets were observed at δ 7.65 (J = 8.1 Hz), 7.36 ( J = 8.1 Hz) and 7.30 ( J = 2.2 Hz), assigned as H4, H7 and H2 respectively. A doublet of doublet of doublets at δ 6.99 (J = 1.2, 7.2,
7.2 Hz) was assigned as H5, while multiplets at δ 7.41 and 7.05 were assigned as
1 H4′/H7′ and H6/H5′/H6′ respectively. In the H NMR (d 6-DMSO) spectrum of 140 68
singlet resonances were observed at δ 11.80, 10.96 and 3.92, corresponding to the indolyl NH, amide NH and CH 2 protons, respectively. Doublet of doublet of doublets were observed at δ 7.36 (J = 1.2, 1.2, 8.0 Hz), 7.08 ( J = 1.2, 8.0, 8.0 Hz) and 6.99 ( J =
1.2, 8.0, 8.0 Hz) corresponding to H7, H6 and H5 respectively. Multiplet resonances at
δ 7.61-7.55 and 7.30-7.27 were assigned as H4/H4′/H7′ and H5′/H6′ respectively.
Quinoline was also selected as another heterocycle to investigate in the context of the
SAR of these amides. As detailed in Chapter 1.5.5, quinolines possess many important biological properties and have also been explored as retinoid modulators. Subsequently,
106 and 8-aminoquinoline 141 were coupled using PyBOP to generate acetamide 142 in
52% yield (Scheme 3.11).
Scheme 3.11: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h.
1 The H NMR (d 6-acetone) spectrum of 142 displayed characteristic singlet resonances at δ 11.01, 10.16 and 4.01, assigned as the indolyl NH, amide NH and CH 2 protons respectively (Figure 3.5). Doublets were observed at δ 7.63 (J = 7.1 Hz), 7.53 ( J = 3.1
Hz), 7.44 ( J = 2.4 Hz) and 7.40 ( J = 8.1 Hz), assigned as H4, H6′, H2 and H7, respectively. Additionally, doublet of doublet resonances were observed at δ 8.73 (J =
1.7, 4.2 Hz), 8.65 ( J = 1.5, 7.8 Hz), 8.34 ( J = 1.7, 8.3 Hz), 7.60 ( J = 1.5, 7.8 Hz) and
7.57 (J = 1.7, 1.7 Hz), assigned as H4′, H7′, H2′, H5′ and H3′ respectively. Doublet of doublet of doublets were also observed at δ 7.09 (J = 2.5, 7.5, 7.5 Hz) and 6.99 ( J = 1.1,
7.5, 7.5 Hz), assigned as H6 and H5 respectively. 69
1 Figure 3.5: H NMR spectrum of 142 in d 6-acetone.
Finally, 3-indoleacetic acid 106 was coupled to 2-aminothiazole 143 to give acetamide
144 in 62% yield (Scheme 3.12). It was thought that by removing the benzene ring, it would be possible to confirm the importance of steric effects and possible π-interactions in this region of the molecule.
Scheme 3.12: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h.
70
3.3.2 Synthesis of indole-benzothiazole amides with modified linkers
Further development of the SAR was next targeted towards the linking region of the amides. The preliminary docking studies suggested that increasing the length of the linker by one methylene unit does not affect the predicted affinity of the molecule, as was shown by the similar Goldscores of acetamide 115 (66.72) and propionamide 117
(66.81), allowing different orientations of the ligand within the receptor to be explored.
Additionally, it was also predicted that reducing the length of the linker would dramatically reduce affinity, demonstrated by the relatively low Goldscore of carboxamide 118 (58.11).
As it was predicted to have the most similar binding affinity to acetamide 115, propionamide 117 was first targeted. PyBOP coupling of 3-indolepropionic acid 145 to
110 gave propionamide 117 in a yield of 73% (Scheme 3.13). The 1H NMR (d6-
DMSO) spectrum of propionamide 117 differed from that of acetamide 115 primarily in the alkyl region. Two triplets were observed at δ 3.07 and 2.87 ( J = 7.1 Hz) corresponding to the two CH 2 protons, compared to the singlet in the spectrum of 115, which had only the one CH 2 group. Additionally, the indolyl NH proton was observed to give a resonance slightly upfield at δ 12.36, relative to δ 12.54 for 115.
Scheme 3.13: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h.
As the longer linker was predicted to have slightly higher affinity relative to the acetamide, further analogues with this length were considered attractive targets for synthesis. The replacement of the indolyl methylene group with a sulphur atom was considered a promising option as such linkages were observed throughout many of the 71
initial 26 ligands. Furthermore, the subtle modifications to the hydrophobicity and size of the linker may also have further effects on the affinity of these molecules to the
RA Rβ binding site. In order to achieve this, the method of Levkovskaya was followed, 274 whereby indole 30 was reacted with thiourea in the presence of potassium iodide and iodine in THF over 3 days to generate the intermediate thiouroniumindole iodide 146. Treatment of this intermediate with potassium hydroxide then gave the corresponding thiol 147 in 61% yield. This was then converted to the required (3- indolesulfonyl)acetic acid 148 in 54% yield by treatment with chloroacetic acid in the presence of potassium hydroxide. Finally, PyBOP coupling to 110 gave the desired amide 149 in 87% yield (Scheme 3.14).
Scheme 3.14: Reagents and conditions: a) thiourea, I 2, KI, MeOH/H 2O, 3 days; b) i. 2
M NaOH, 12 h, ii. 2 M HCl; c) chloroacetic acid, KOH, EtOH, 12 h; d) PyBOP,
DIPEA, DCM, 2 h.
As an alternative to extending the length of the acyl chain, the separation of the benzothiazole heterocycle and amide group was considered as another point of modification. It was thought that incorporating a hydrazide moiety, rather than amide, may also confer a more rigid structure through the possible generation of an intramolecular hydrogen bond, as well as again modifying the hydrophobicity of the 72
ligand. However, when 106 was coupled to 2-hydrazinobenzothiazole 150 using PyBOP conditions, the expected hydrazide 151a was not the isolated product, nor was the tautomeric 151b (Scheme 3.15).
Scheme 3.15: Reagents and conditions: a) PyBOP, DIPEA, DCM, 18 h.
1 The H NMR (d 6-DMSO) spectrum of the product showed the expected aromatic peaks for both the indole and benzothiazole moieties (Figure 3.6). The indolyl NH and CH 2 protons appeared as singlets at δ 10.37 and 4.35 respectively, while H4 and H7 gave doublets ( J = 7.9 Hz) at δ 7.55 and 7.36 respectively. The doublet of doublet resonances
(J = 7.9, 7.9 Hz) corresponding to H5 and H6 were observed at δ 6.99 and 7.08 respectively, while H2 overlapped with the be nzothiazole H6′ at δ 7.34-7.29. The remaining benzothiazole protons were observed as doublet of doublets (J = 0.6, 7.9 Hz) at δ 7.95 (H4′) and 7.82 (H7′), with H5′ givng a doublet of doublet of doublets ( J = 1.3,
7.3, 7.3 Hz) at δ 7.46. There was also an additional singlet observed at δ 5.84, integrating for 2 protons, which could not be accounted for by structures 151a or 151b.
A D 2O exchange was also conducted, which resulted in the suppression of this signal in
1 the H NMR (d 6-DMSO + D 2O) spectrum confirming that the protons at δ 5.84 are exchangeable. Additionally, the HRMS (+ESI) identified the [M+H]+ peak at m/z
323.0964, corresponding to the formula of 151a-b: C17 H15 N4OS (required m/z 73
323.0961). Therefore, the structure was assigned as N-(benzo[d]thiazol-2-yl)-2-(1 H- indol-3-yl)acetohydrazide 152, which was isolated in a yield of 89%.
1 Figure 3.6: Comparative H NMR (d 6-DMSO: blue, d 6-DMSO + D 2O: red) spectra of
the unexpected structure 152.
Although unexpected, structures similar to 152 are not entirely unprecedented in the literature. The group of Adam et al . reported hydrazide 153, achieved via a rearrangement of the corresponding hydrazono penicillanate, 275 while Haviv et al . prepared oxazole 154 utilizing a DCC mediated coupling. 276 The benzoxazole 154 is of particular interest as it was also prepared through the use of a peptide coupling reagent, which typically is expected to give the N′-hydrazides ( 151a-b).
Figure 3.7 : Structures of reported hydrazides similar to that of 152. 74
Finally, the synthesis of a carboxamide was targeted in order to determine what effects a reduction in linker length would have on the cytotoxicity of these scaffolds. In order to achieve this, indole-2-carboxylic acid 155 and 110 were coupled under PyBOP conditions to generate carboxamide 156 in 83% yield (Scheme 3.16).
Scheme 3.16: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h .
3.3.3 Synthesis of analogues bearing substituents on indolyl or benzothiazolyl rings
In drug design, substitution about a heterocyclic ring system can have dramatic effects on the biological activity of a molecule. This may be through modulating the hydrophobicity of the compound and increasing its membrane permeability, blocking metabolic sites and increasing half-life, or generating stronger or different interactions with a binding site to improve the efficacy and/or improve the toxicity profile of the compound.
Figure 3.8: Lead structures containing substituted benzothiazoles.
Interestingly, of the four occurrences of the benzothiazole moiety in the initial twenty six ligands, two of these had appended substituents, both of which were methyl groups
(Figure 3.8). In the case of L11, the substituent was located at the benzothiazole C6 position, while in L15 the substituent was located at the benzothiazole C4 position. As 75
the molecular modelling predicted that the benzothiazole moiety would be oriented towards the Arg269 and Ser280 residues responsible for binding to t-RA 3, substitution about this ring, particularly at C6, was considered a crucial aspect of the SAR.
Consequently, 106 was coupled to a range of substituted 2-aminobenzothiazoles 157a-l following the standard PyBOP coupling discussed earlier in this chapter, giving amides
158a-l in good yields of 41-71% (Scheme 3.17).
Scheme 3.17: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h.
The 1H NMR spectral data of amides 158a-l all displayed consistent signals for the indolyl portion, differing only in the splitting patterns and chemical shifts of the substituted benzothiazole protons. The 6-methoxy derivative 158e had representative 1H
NMR (d 6-acetone) spectral data for analogues 158a-g (Figure 3.9). Three singlets were observed at δ 10.93, 10.25 and 3.84, corresponding to the indolyl NH, amide NH and methoxy protons, respectively. Two doublets were observed at δ 7.46 (J = 2.5 Hz) and
4.04 (J = 0.8 Hz), corresponding to H7′ and the CH 2 protons, respectively. Resonances assigned to H7, H4′ and H5′ respectively, were observed as doublet of doublets at δ 7.68
(J = 0.6, 7.9 Hz), 7.56 ( J = 0.4, 8.9 Hz) and 6.99 ( J = 0.6, 8.9 Hz). Two doublet of 76
doublet of doublets resonances were observed at δ 7.12 (J = 1.2, 7.0, 7.0 Hz) and 7.04 ( J
= 1.2, 7.9, 7.9 Hz) assigned as H6 and H5, while H4 was observed as a multiplet at δ
7.41.
1 Figure 3.9: H-NMR (d 6-acetone) spectrum of 158e.
1 Similarly, the H-NMR (d 6-acetone) spectral data of 155k was representative of analogues 158i-k (Figure 3.10). Singlet resonances were observed at δ 12.66, 10.98,
3.91 and 3.89, assigned as the indolyl NH, amide NH, methoxy and CH 2 protons, respectively. One doublet signal was observed at δ 7.31 (J = 2.4 Hz) corresponding to
H2, while two doublet of doublets were observed at δ 7.49 (J = 0.9, 8.0 Hz) and 7.23 ( J
= 8.0, 8.0 Hz), assigned as H7′ and H6′ respectively. The remaining protons ga ve doublet of doublet of doublets at δ 7.60 (J = 1.0, 1.0, 8.0 Hz), 7.36 ( J = 1.0, 1.0, 8.0 Hz),
7.08 (J = 1.0, 8.0, 8.0 Hz) and 6.99 (( J = 1.0, 8.0, 8.0 Hz), assigned as H4, H7, H6 and
H5/H5′ respectively. 77
1 Figure 3.10: H-NMR (d 6-acetone) spectrum of 158k.
The next consideration was to determine if the addition of substituents on the indolyl ring would have any impact upon the SAR. There are many reactions for preparing 3- substituted indoles, such as the Bartoli, Fukuyama, Larock and Madelung methods. 277-
280 However, these methods typically produce indoles that also bear substituents at the
2-position. As it was first desired to produce indoles bearing a single substituent, other methods were investigated.
Initial attempts aimed at producing substituted 3-indoleacetic acids via alkylation of the reactive indole 3-position were unsuccessful. Following the method of Kumar et al. ,281 a methanolic solution 5-methoxyindole 159 was heated at reflux with ethyl chloroacetate in the presence of potassium carbonate for 8 h. However, this did not yield the desired
5-methoxyindole-3-acetic acid 160, returning only unreacted starting material (Scheme
3.18). 78
Scheme 3.18: Reagents and conditions: a) ethyl chloroacetate, MeOH, reflux, 8 h.
The Wolff-Kishner-type reduction of 3-indoleglyoxylic acids reported by Petit et al. was next investigated as a means of generating the desired 3-indoleacetic acids. 282
Treatment of 159 with oxalyl chloride in anhydrous ether at 0 °C, followed by quenching with a saturated solution of aqueous sodium bicarbonate produced glyoxylic acid 161 in a high yield of 91%. However, heating 161 at 150 °C with hydrazine and sodium methoxide in 2-methoxyethanol did not generate the desired 162 (Scheme 3.19).
Scheme 3.19: Reagents and conditions: a) i. oxalyl chloride, Et 2O, 0 °C, 2 h. ii. aq.
NaHCO 3; b) NH 2NH 2, NaOMe, 2-methoxyethanol, 150 °C, 1 h.
Demopoulos reported a catalytic reduction of the glyoxylate esters, which was next investigated. Indole 159 was first treated with monoethyl oxalyl chloride in ether to generate ester 163 in a yield of 69%. The ester 163 was then treated with sodium hypophosphite in the presence of 10% palladium on charcoal and heated at reflux in 1:1 dioxane/water for 48 h. However, this again did not yield the desired ester 160 (Scheme
3.20). 79
Scheme 3.20: Reagents and conditions: a) monoethyl oxalyl chloride, Et 2O, 0 °C, 4 h;
b) NaHPO 4, Pd/C, dioxane/H 2O (1:1), 48 h.
As the required indole acetic acids could not be produced by direct alkylation or reduction of glyoxylates, attention was subsequently turned to the Fischer indole synthesis. Accordingly, phenyl hydrazine 164 and ethyl levulinate 165 were heated at reflux in glacial acetic acid in the presence of sodium acetate for 3 h. The solvent was removed and the crude intermediate 166 heated at reflux in 4 M HCl/dioxane for 15 h to give ester 167 in 25% yield. Hydrolysis of ester 167 with ethanolic sodium hydroxide, followed by acidification with 2 M HCl then gave acid 168 in 90% yield (Scheme 3.21).
Scheme 3.21: Reagents and conditions: a) AcONa, AcOH, reflux, 3 h; b) 4 M
HCl/dioxane, reflux, 15 h; c) i. NaOH, EtOH, 1 h, ii. 2 M HCl. 80
Upon attempting to utilize this procedure with substituted hydrazines, only a thick black residue was obtained, from which no product could be isolated by column chromatography. In order to overcome this, a modified procedure was developed in which hydrazines 169a-b and levulinic acid 170 were heated at reflux in a 1:1 mixture of conc. HCl and toluene for 3 h, generating acids 171a-b in yields of 34 and 67%, respectively (Scheme 3.22).
Scheme 3.22: Reagents and conditions: a) HCl:toluene (1:1), reflux, 3 h.
With the required substituted indoles now in hand, the corresponding acetamides could be prepared. Subsequently, indole acids 168 and 171a-b were coupled to 110 using
PyBOP in the presence of DIPEA in DCM to give acetamides 172a-c in yields of 39-
1 51% (Scheme 3.23). The H NMR (d 6-DMSO) of 172c was representative for these compounds, displaying five singlet resonances at δ 12.51, 10.74, 3.83, 2.38 and 2.34, corresponding to the indolyl NH, amide NH, CH 2, C2 methyl and C5 methyl protons, respectively. Two doublets were observed at 7.31 ( J = 1.3 Hz) and 7.13 ( J = 8.2 Hz), assigned as H4 and H7, respectively, with the doublet of doublets at δ 6.81 ( J = 1.3, 8.2
Hz) assigned as H6. The four doublet of doublet of doublets at δ 7.94 (J = 0.6, 1.2, 7.2
Hz), 7.74 ( J = 0.6, 1.2, 8.0 Hz), 7.43 ( J = 1.3, 7.2, 7.2 Hz) and 7.28 ( J = 1.2, 8.0, 8.0
Hz) were assigned as H4′, H7′, H6′ and H5′ respectively. 81
Scheme 3.23: Reagents and conditions: a) PyBOP, DIPEA, DCM, 2 h.
3.4 Characterization of in vitro activity against neuroblastoma cell lines
In order to determine the cytotoxic activity of the synthesized indole-benzothiazole amides towards neuroblastoma cells in vitro , the compounds were screened against the
BE(2)-C neuroblastoma cell lines stably transfected with the MEP4-empty vector (E.V.) and MEP4-RARβ constructs, which were utilized in the preliminary library screen. The cells were treated with compound at concentrations ranging from 0.1-20 µM over 72 h and the percentage of viable cells was measured by the Alamar blue assay. The treatment was conducted in 6 replicate wells and the experiment repeated three times (n
= 3). Dose response curves were generated and the IC 50 values were determined using the GraphPad Prism 6 software package (Table 3.2).
82
Table 3.2: Cytotoxicity of indole-benzothiazole amides towards empty vector and
RARβ stably transfected BE(2)-C neuroblastoma cells following 72 h exposure.
Compound IC 50 ( M) E.V. RARβ 115 1.5 0.9 136a 33.8 23.5 136b 10.1 8.4 138 >50 31.1 140 >50 >50 116 >50 >50 142 >50 26.2 144 >50 >50 117 15.8 10.8 149 30.5 18.3 152 >50 >50 156 >50 >50 158a 16.1 14.1 158b 23.5 25.7 158c 20.1 23.5 158d 16.5 15.4 158e 37.1 34.2 158f 42.1 41.7 158g 9.0 9.9 158h 6.5 4.1 158i 3.6 2.1 158j 7.5 5.1 158k 14.2 8.3 158l 12.1 10.7 172a 5.4 12.5 172b 3.2 5.0 172c 5.8 5.8
Overall, the cytotoxicity of these compounds was observed to be relatively strong, with the majority of compounds possessing either a low IC 50 value, selectivity towards the
RARβ overexpressing cell s, or both. It was observed that indole-benzothiazole amide
115 possessed the lowest IC 50 values of 1.5 µM for the E.V. and 0.9 µM for the RARβ 83
overexpressing cells. This was a promising result as it showed selectivity towards the
RARβ overexpressing cells (1.7 times) at such a low concentration, similar to what had previously been shown in the SH-SY5Y neuroblastoma cell line where it had an IC 50 value of 1.1 µM. 273
It was also interesting to see that changing either of these heterocycles resulted in a reduction of activity. Replacement of the indolyl moiety in 115 with naphthalene in
136a-b resulted in decreased activity towards both cell lines, with 136a having IC 50 values of 33.8 and 23.5 µM for the E.V. and RARβ overexpressing cells respectively, and 135b having IC 50 values of 10.1 and 8.4 µM for the E.V. and RARβ overexpressing cells respectively. Furthermore, while both 151a and 151b had similar levels of selectivity towards the RARβ cell line (1.4 and 1.2 times, respectively), the 2-napthyl derivative 136b was much more potent (3-fold) than the corresponding 1-napthyl derivative 136a. This suggests that the orientation of the heterocycle is important for both activity and selectivity. Additionally, the methylenedioxy derivative 138 displayed reduced activity compared to 115, possessing IC 50 values of >50 and 31.1 µM for the
E.V. and RARβ overexpressing cells, respectively . Furthermore, upon the change from a phenyl ring in 136b to the methylenedioxy ring in 138, the activity in both cell lines is reduced, suggesting that the presence of a second aromatic ring is favoured to a 5- membered methylenedioxy ring. Substitution of the benzothiazole sulphur atom in 115 with either an NH group ( 116) or oxygen atom ( 140) resulted in a loss of activity, with both 116 and 150 possessing IC 50 values above 50 µM in both cell lines. This is quite an interesting result considering the similar docking scores of 115 and 116 (66.72 and
63.73 respectively). Replacement of the benzothiazole in 115 with quinoline in 142 resulted in decreased cytotoxicity, but increa sed selectivity towards the RARβ cell line
(IC 50 : >50 and 26.2 µM). Additionally, replacing the benzothiazole of 115 with a 84
thiazole in 144 resulted in a loss of cytotoxicity, suggesting that the size of the heterocycle and the ability to form hydrophobic interactions is highly important in this region of the molecule.
With regard to the amide linker, it was observed that increasing the length of the linker by one methylene unit (as in 117) reduced the cytotoxicity relative to acetamide 115, but retained similar selectivity towards the RARβ cell line (IC 50 : 15.8 and 10.8 µM, 1.5 times selectivity). This is despite the near identical Goldscores of 115 and 117 (66.72 and 66.81 respectively). Furthermore, replacement of the additional methylene with a sulfur atom in 149 increased the selectivity towards the RARβ cell line, relative to 117, albeit at the cost of cytotoxicity (IC 50 : 30.5 and 18.3 µM, 1.7 times selectivity).
Additionally, changing the amide of 115 to a hydrazide in 152 resulted in a loss of cytotoxic activity, with 152 possessing IC 50 values above 50 µM in both the E.V. and
RARβ overexpressing cell line s. Similarly, reducing the length of the linker (as in carboxamide 156) also resulted in IC 50 values above 50 µM in both cell lines.
Overall, these results show that the combination of the indole and benzothiazole heterocycles give the most potent scaffold (as in 115). Changes to the type of heterocycle, or the positioning of the amide substituent around the heterocycle result in a loss of cytotoxicity, but may increase the selectivity towards the RARβ overexpressing cell line. Bulkier heterocycles are preferred, particularly those with more hydrophobic character as opposed to the incorporation of electronegative atoms capable of participating in hydrogen bonds. Lengthening the linking region reduces cytotoxicity whilst maintaining selectivity towards the RARβ overexpressing cell line over the E.V. cell line. However, reducing the length resulted in a complete loss of activity. Taken together, these data suggest that both the length of the linker and the orientation of the heterocyclic moieties are important for cytotoxicity. 85
Upon investigation of the biological activity of the benzothiazole substituted analogues, it was observed that 6ɂ -substituted analogues 158a-g generally showed large reductions in activity. The effect also seemed proportional to the size of the substituent, with larger substituents generally resulting in less active compounds. Considering the halogenated analogues 158a-c, the 6-fluoro derivative 158a showed some selectivity towards the
RARβ cell line, with IC 50 values of 16.1 and 14.1 M in the E.V. and RARβ cell lines respectively. Increasing the size of the halogen to chlorine in 158b or bromine in 158c was observed to further reduce the cytotoxic activity, with 158b having IC 50 values of
23.5 and 25.7 M against the E.V. and RARβ overexpressing cell lines, and 158c having IC 50 values of 20.1 and 23.5 µM against the E.V. and RARβ overexpressing cell lines, respectively. Notably, the incorporation of halogens at the benzothiazole C6 led to inversion of the cell line selectivity, with the compounds being more effective in the
E.V. cell line.
Similar trends were observed with the alkyl derivatives 158d-g. Methyl derivative 158d showed similar levels of cytotoxicity and selectivity to fluoro derivative 158a, with IC 50 values of 16.5 and 15.4 M in the E.V. and RARβ cell lines respectively. Subsequent ly increasing the size of this substituent to methoxy ( 158e, IC 50 : 37.1 and 34.2 µM) and ethoxy ( 158f, IC 50 : 42.1 and 41.7 µM) resulted in further decreases to cytotoxicity with no significant impact to selectivity. Interestingly, the large trifluoromethyl group of
158g improved cytotoxicity relative to 158a and 158d, possessing IC 50 values of 9.0 and
9.9 M against the E.V. and RARβ cell lines respectively. Furthermore, thi s analogue was also observed to have inverted cell line selectivity, similar to 158a-c.
The positioning of the substituent about the benzothiazole ring was also found to be an important factor. The 4,6-difluoro analogue 158h was observed to have increased cytotoxicity compared to the 6-fluoro derivative 158a, with IC 50 values of 6.5 and 4.1 86