U UNIVERSITY OF CINCINNATI

Date:

I, , hereby submit this original work as part of the requirements for the degree of:

in

It is entitled:

Student Signature:

This work and its defense approved by:

Committee Chair:

Approval of the electronic document:

I have reviewed the Thesis/Dissertation in its final electronic format and certify that it is an accurate copy of the document reviewed and approved by the committee.

Committee Chair signature:

Exploring the PI3K and  binding sites by homology modeling and inhibitors utilizing a 2,6-disubstituted isonicotinic scaffold

A dissertation submitted to the Graduate School of the University of Cincinnati in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

in the Division of Pharmaceutical Sciences of the James L Winkle College of Pharmacy by

Philip T. Cherian M. Pharm. University of Mumbai June 2001

Committee Chair: James J. Knittel, Ph.D.

Abstract

Phosphatidylinositol 3-OH kinases (PI3Ks) are dual specific lipid and protein kinases that catalyze the synthesis of the lipid second messenger Phosphatidylinositol-3,4,5- trisphosphate (PIP3) and influence multiple cellular processes including cell growth, proliferation, survival and motility. PI3Ks are divided into three classes I, II and III and the class I contains four isoforms, namely p110, , and . Of these, the p110 isoform

(PI3K) is an important therapeutic target in cancer as the PIK3CA gene that encodes the p110 catalytic subunit is frequently mutated in a variety of cancers. Though several classes of compounds that inhibit the class I enzymes have been reported, development of inhibitors selective for the PI3Kstill remains a major challenge. In accordance with the ongoing research efforts towards the development of isoform selective inhibitors, we explored the differences between the p110 and  binding sites using a structure-based approach. The study entailed the building of a p110 homology model, development of a novel scaffold that provided the ease of assembly and diversification and designing of focused chemical libraries based on our modeling studies.

Advancement in protein structure prediction methods has simplified the process of obtaining reliable 3D structures of target proteins. Using the p110 protein sequence and

X-ray structure of p110, a homology model of p110 was constructed and refined. This model proved to be in good agreement with the later published X-ray structure of p110

(2rd0). Using this model and literature analysis of PI3K inhibitors, we designed the 2,6- disubstituted isonicotinic scaffold for our study. This scaffold was evaluated for its synthetic feasibility and biological activity by designing, synthesizing and testing an initial set of derivatives. These compounds inhibited the activity of the recombinant

iii

purified PI3Kand  in our in vitro lipid kinase assay and showed inhibition of PI3K- dependent survival of cell lines derived from the hematopoietic FL5.12 cells. The most

potent compound in the series (compound 28) showed potency in the low micromolar

range with 7-fold selectivity for PI3K. The chemistry developed during the synthesis of

the above series provided straightforward access to three chemical libraries. Accordingly,

the three regions of the scaffold were modified in order to explore the hydrogen bonding,

bulk and polarity as predicted by our model. These modifications led to the development

of compound 63 which showed >10-fold selectivity for PI3K vs. PI3K and was more

potent in the cell assay than previous compounds. Based on our docking studies, the

selectivity of this compound can be attributed to its interaction with Arg770 and Trp780

of p110.

Overall we demonstrate the utility of homology modeling and the 2,6-disubstituted

scaffold for exploring the p110 and  binding sites and anticipate that the data generated

during this study may be useful toward the development of more potent and selective

PI3K and  inhibitors.

iv

Acknowledgements

I express my sincere gratitude to my advisor Dr. James J Knittel for his supervision,

support and patience throughout the course of my graduate studies and for his

contribution towards my professional development. I thank my committee members Dr.

David Plas, Dr. Matt Wortman, Dr. Giovanni Pauletti and Dr. Hal Ebetino for their

guidance, constructive criticism and encouragement during the course of this project. In

addition I would like to thank Dr. David Plas for allowing me to use his facilities for the

cell assay, Dr. Matt Wortman for the software and computers for molecular modeling and

Dr. Pauletti for facilities for the kinase assay. I thank Jennifer Barger from the Plas Lab for generating and maintaining the cell lines, designing the cell assays and for her

tremendous help during the assay. I would also like to thank Dr. Namal Warshakoon for his help with the project.

I express my sincere thanks to our lab members Dr. Leonid Koikov and Dr. Eric Hu for

their valuable input and advice during the project and otherwise and Dr. Andrew Ruwe

for his assistance during my graduate studies. I would like to thank the James L Winkle

College of Pharmacy for their support during my graduate studies as well as for the

College of Pharmacy Dean’s Pilot project grant that partially funded this project. I thank

the Genome Research Institute (GRI) for providing the facility and instruments for the

project.

I thank my friends Nirmal, Amit, Sujeet, Moin, Arjun, Rishikesh and Purnima Kulkarni,

Shiv and Bhuvana Vishwanathan, Todd and Sara Porter, Brian Drohan, Paul Tanaka and

Mugove Manjengwa for making my stay in Cincinnati a delightful experience.

vi

I express my sincere thankfulness to my brother and his family for their love and support.

Above all, I express my deepest gratitude to my parents for their ever increasing love, support and encouragement over the years which have made this accomplishment possible.

vii

Table of Contents

Abstract...... iii

Acknowledgements ...... vi

Table of contents ...... viii

List of illustrations...... xi

List of tables ...... xiii

List of abbreviations ...... xiv

I. Introduction

 PI3K family...... 4

 Therapeutic applications of PI3K/Akt pathway...... 16

 PI3K inhibitors...... 20

 Inhibitors and crystal structures...... 26

 Goal, hypothesis and specific aims...... 30

II. Molecular modeling

 Protein structure prediction...... 32

 Modeling programs and servers...... 44

 Limitations of structure prediction methods...... 46

 Homology modeling in drug design and development...... 47

 Results and discussion

- Constructing the p110 homology model ...... 50

- Designing the scaffold ...... 53

 Conclusions...... 58

 Limitations of our modeling ...... 61

viii

III. Evaluating the 2,6-disubstituted isonicotinic scaffold

 Introduction...... 65

 Results and discussion

- Chemistry...... 69

- Biological Evaluation ...... 73

 Conclusions...... 83

 Limitations of our experimental methods...... 83

IV. Structure-activity relationship (SAR) of 2,6-disubstituted isonicotinic derivatives

 Introduction...... 85

 Results and discussion

A. Modifications of Carboxylic acid group ...... 87

- Chemistry...... 88

- Biological evaluation...... 90

B. Substitution of ...... 91

- Chemistry...... 93

- Biological evaluation...... 94

C. Substitution at the 6-methyl position ...... 97

- Chemistry...... 99

- Biological evaluation...... 103

- Docking of 63 in p110 model...... 105

 Conclusions...... 109

V. General discussion and Future directions ...... 111

VI. Experimental...... 114

ix

VII. References...... 136

x

List of Illustrations

Figure # Title/Description Page #

1. PI3K in action 1

2. Relationship between PI species 2

3. The PI3K family of enzymes 7

4. The p110/niSH2-p85 heterodimer X-ray structure 8

5. Kinase domain of p110 9

6. PI3K signaling pathways 13

7. Natural compounds and their derivatives as PI3K inhibitors 21

8. LY294002 and related compounds 23

9. Second generation PI3K inhibitors 24

10. Endogenous PI3K inhibitors 25

11. PI3K inhibitors in clinical trials 26

12. Binding mode of different compounds in the PI3K active site 28

13. Comparison of the Swiss-Prot and PDB databases 33

14. Fold recognition - Unrelated proteins showing similar architecture 35

15. The p110 homology model 53

16. Proposed scaffold for exploring the p110 and  isoforms 54

17. Overlap of test LY294002 with X-ray conformation 55

18. Docking of LY294002 and Isonicotinic acid derivative in p110 56

19. Comparing binding modes of LY294002 and proposed scaffold 56

20. Simple synthetic approaches allow preparation of three libraries 57

21. MM3 calculations for three rotatable bonds of scaffold 57

xi

22. Ramachandran plot for the p110 homology model 60

23. SAR of 2-Morpholinylchromone and X-ray of LY294002 in p110 65

24. SAR of TGX series 66

25. Overlap of PI3K and PI3K reveals important differences 67

26. Initial series of derivatives 68

27. Proposed synthetic scheme for 26-29 69

28. Modified synthetic scheme for 26-29 70

29. Synthesis of 30, 31 71

30. Optimization of the bromination 71

31. Optimization of the morpholine addition 72

32. The Ultra-Glo™ Recombinant Luciferase reaction 73

33. Inhibition curves for L294002, 26, 26, 30 and 31 74

34. FL5.12 cell model system and description of important cell lines 76-77

35. Results of Cell assay 1 81

36. Results of Cell assay 2 82

37. Three regions of the scaffold explored for SAR study 85

38. Amino acid sequences of p110 and p110 aligned with BLAST 87

39. Modifications of the carboxylic acid group 88

40. Synthesis of 39-43 89

41. Percent inhibition of PI3K and  by 39-43 90

42. Important residues in the morpholine pocket 92

43. Substitution of morpholine 93

44. Synthetic scheme for 44-55 94

xii

45. Percent inhibition of PI3K and  by 44-55 95

46. Hydrophilic residues at the entrance of binding pocket of p110 97

47. Cation- interaction between the Arg 770 and Trp780 of p110 98

48. Substitution at the 6-methyl position 99

49. Synthetic scheme for 56-59, 63 and 65 100

50. Synthesis of amino derivatives 60, 61 101

51. Synthesis of 62 102

52. Synthesis of 64 102

53. Percent inhibition of PI3K and  by 56-65 103

54. Docking of 63 in p110 model and p110X-ray 106

55. Results of Cell assay 3 108

56. Important improvement provided by 63 over LY294002 110

57. Summary and Future directions 112

List of Tables

Table # Title Page #

1. Docking scores and IC50 of select compounds on the p110 64

2. Name, chemical structure and IC50 values of 29-31 75

3. Biological activity of 39-43 91

4. Biological activity of 44-55 97

5. Biological activity of 56-65 104

xiii

List of Abbreviations

3-PI 3-Phosphoinositides

AIBN azobisisobutyronitrile

ANOVA analysis of variance

BLAST Basic local alignment search tool

CASP Critical Assessment of Techniques for Protein Structure Prediction DIBAL-H diisobutylaluminum hydride

DMSO dimethyl sulfoxide

DNA-PK DNA-dependent protein kinase

fMLP N-formyl-MetLeuPhe

GA genetic algorithm

GPCRs g-protein coupled receptors

IL-3 Interleukin-3

LC-MS Liquid chromatography – Mass spectrometry

LiAlH4 lithium aluminum hydride

NBS N- bromosuccinamide

NMR nuclear magnetic resonance

PDB protein data bank

PI3K Phosphatidylinositol 3-OH kinase

PtdIns (3,4,5) P3 or PIP3 P hosphatidylinositol -3,4,5 -trisphosphate

RBD Ras binding domain

RMSD or rmsd root mean square deviation

PMF potential of mean force

xiv

SAR structure-activity relationship

SBDD structure-based drug design

RP-HPLC reverse phase high performance liquid chromatography

xv

I: Introduction

Phosphatidylinositol 3-OH Kinases (PI3K) are a family of dual specific lipid and protein

kinases that generate the lipid second messenger 3-phosphoinositides (3-PIs) by

phosphorylating the 3 hydroxyl group of phosphatidylinositol (PtdIns) (Fig. 1). These 3-

PIs interact with the lipid binding domains of various cellular proteins and influence

multiple cellular processes including cell growth, proliferation, survival and motility. At

present, four known species of 3-PIs (Fig. 2) exist of which three are generated by PI3K.

These include the PtdIns(3)P, PtdIns(3,4)P2 and PtdIns(3,4,5)P3 (aka PIP3) that are

synthesized from PtdIns, PtdIns(4)P and PtdIns(4,5)P2 respectively. The final PtdIns

(3,5)P2 is generated via phosphorylation of PtdIns(3)P by the PtdIns 5-OH kinase (PI5K)

called p235. Among the various 3-PIs, substantial levels of PtdIns(3)P are found in

resting cells, while the levels of PtdIns(3,4)P2 and PtdIns(3,4,5)P3 are extremely low. In

R2 R1 R2 R1 R = Membrane bound fatty acid tails

O O O O Diacylglycerol O O O O

O O O O

P Phosphodiester P OH Linkage OH O O PI3 Kinase O O 6 6 2 1 OH 2 1 OH OH OH 4 O O 4 OH 3 5 ATP ADP 3 5 OH P OH Inositol head group O O

Figure 1: PI3K in action. Phosphatidylinositols (PtdIns) contain a 1D-myo-inositol phosphate group linked to diacylglycerol. The 2-hydroxyl group of the D-myo-enantiomer is axial while all other hydroxyl groups are equatorial. The esters at R1 and R2 are generally stearate and arachidonate. The PI3Kinases phosphorylate the 3-hydroxyl group. The hydroxyl groups at the 4 and 5 positions are phosphorylated by PtdIns 4-Kinases (PI4K), PtdIns 5-Kinases (PI5K) and a family of Phosphatidylinositol phosphate kinases (PIPKs).

1

spite of its low levels, the PtdIns(3,4,5)P3 has gained attention as production of PIP3 is associated with various cellular process in both normal and pathophysiological states. In response to different cellular stimuli the levels of PtdIns(3,4,5)P3 increase rapidly leading to a momentary elevation of concentration (> 50 fold) which goes back to baseline levels within minutes. Metabolism of the 3-PIs is brought about by various kinases and phosphatases that act on the inositol ring.1, 2

Phosphoinositides (PI)

PtdIns

Monophosphorylation PtdIns(3)P PtdIns(4)P PtdIns(5)P

Bisphosphorylation PtdIns(3,4)P2 PtdIns(3,5)P2 PtdIns(4,5)P2

PtdIns(3,4,5)P Triphosphophorylation 3

Figure 2: Relationship between PIs. Eight species of PIs exist four of which are 3- PIs. Various lipid kinases and phosphatases bring about the conversion between these PIs. Interaction of PIs and cellular proteins

The second messenger phosphoinositides interact with a variety of cellular proteins that contain characteristic binding motifs capable of recognizing the distinct head groups of

PIs. The well known structural motifs include the FYVE domain, the Pleckstrin

Homology (PH) domain, the epsin N-terminal homology domain (ENTH) and the Phox

Homology (PX) domain.3 Other recently discovered binding motifs include the FERM

(band 4.1, ezrin, radixin and moesin), GRAM (glucosyltransferase, Rab-like GTPase

2

activator and myotubularin), BAR (Bin/amphiphysin/Rvs) and the Ca2+ binding C2

domains. 4 - 6

The FYVE domain (FYVE term is derived from the first four proteins that were identified with this domain; Fab1p, YoTB, Vac1p and EEA1) consists of 60 – 80 amino acids that bind with two Zn+2 ions. It consists of eight conserved cysteine residues each of

which can coordinate with a zinc ion. The FYVE domains also contain several conserved

sequences such as the R+HHC+XCG motif surrounding the third and fourth cysteines

(where + is a positively charged and X is any amino acid). It binds only to PtdIns(3)P and

is found in proteins that act at the docking and fusion stages of membrane transport. More

than 30 mammalian proteins have been identified with this domain including the

mammalian early endosomal antigen 1 (EEA1), rabenosyn-5, hepatocyte growth factor-

regulated tyrosine kinase (Hrs), PIKfyve and Fab1p.4, 7

The PH domain consists of about 100 amino acids and is found in several proteins

including protein kinase B (PKB/Akt), phosphoinositides-dependent kinase 1 (PDK1), - actinin and GTP-ases such as dynamin. The PH domains of different proteins show very low sequence identity but they contain a conserved tertiary structure comprising a seven- stranded strongly bent -sheet and a C-terminal -helix that packs against the -sheet.

The PH domain is capable of binding to PtdIns(4,5)P2 and PtdIns(3,4,5)P3 and is found in proteins involved in cellular signaling, cytoskeletal rearrangement and membrane trafficking. 3, 5

The ENTH domain is approx. 140 residues in length and is usually located at the N-

termini of proteins. The ENTH domain is made of 8-10 -helices connected by loops of

varying lengths and forms a compact globular structure. It is found in proteins involved

3

in endocytosis and/or cytoskeletal rearrangement such as epsin 1-3, adaptor protein 180

(AP180) and clathrin assembly lymphoid myeloid leukemia protein (CALM). Proteins

containing this domain bind selectively to PtdIns(4, 5)P2. 3, 5

The PX or Phox (Phagocytic oxidase) domain is approx. 120 residues in length and was

first discovered in the p47phox and p40phox domains of the NADPH oxidase. It consists

of several subdomains and is organized into a flat and compact structure containing three

antiparallel -sheets followed by three -helices. It is found in proteins involved in a number of cellular processes including membrane trafficking, cellular signaling, protein

sorting and activation of T and B cells. The PX domain can interact with a variety of PIs including PtdIns(3)P, PtdIns (3,4)P2 and PtdIns(3,4,5)P3. 3

PI3K family

PI3Ks are part of a large family of PI3K-related kinases or PIKK that share sequence

homology with PI3K. The other members include mTOR (mammalian target of

rapamycin), ATM (ataxia telangiectasia mutated), ATR (ATM and RAD3 related), Smg-

1 and DNA-PK (DNA-dependent protein kinase).139 However, unlike the PI3Ks which

phosphorylate lipids, the other members of the family phosphorylate proteins.

The mammalian PI3K family consists of eight known PI3Ks that are organized into three

classes - I, II and III depending on their structure, regulation and substrate specificity. Of

the three, the class I enzymes is the most studied. The catalytic subunits of most PI3Ks

share homologous regions that include the catalytic domain, the PI kinase (PIK or helical) domain, the C2 phospholipid binding domain (C2) and the RAS binding domain (RBD)

(Fig. 3).

4

Class I

Class I PI3Ks are heterodimeric proteins consisting of a 110 kDa catalytic subunit (p110)

and a smaller adaptor/regulatory subunit. Class I PI3Ks are subdivided into 2 classes IA

and IB based on their structure and mode of activation. The class IA contains three

isoforms namely PI3K,  and . The catalytic subunits p110,  and  are encoded by

three separate genes and share 42-58% sequence identity.1 Each of the three isoforms exist as a complex with any one of the five known adaptor proteins – p85, p55, p50, p85 and p55 that are generated by expression and alternative splicing of three different genes p85, p85 and p55. p55 and p50 are derived from the alternatively spliced mRNAs of the same gene that encodes p85.2 The p110 and  isoforms are

ubiquitously expressed while the p110 expression is restricted mainly to leucocytes.2

The class IA PI3Ks are regulated downstream of receptor tyrosine kinases. Other than the common domains mentioned above, the catalytic unit of class IA enzymes contains an adaptor binding domain (ABD) in the N-terminal region which is responsible for the interaction of p110 with the adaptor proteins.

The adaptor subunits do not possess any intrinsic catalytic activity but contain several modular domains that allow for protein-protein interactions. The term modular refers to domains that can be separated functionally and spatially from the rest of the protein in which they reside. All class IA adaptor subunits have two Src-homolgy2 (SH2) domains, referred to as the N-terminal (nSH2) and C-terminal (cSH2) SH2 domains. Both SH2 domains bind preferentially to the phospho-Tyr-X1-X2-Met motif found in numerous proteins including ones that activate the PI3K signaling e.g. epidermal-growth factor receptor, the tyrosine protein kinase Abl, polyoma virus middle T antigen and insulin-

5

receptor substrate proteins.1 A second methionine or valine at the X1 position increases

binding affinity, particularly for the nSH2. Flanked between the two SH2 domains is the

inter-SH2 (iSH2) domain that is necessary and sufficient for interaction with the N-

terminus of the p110 subunits. All class IA adaptor proteins also contain two or three

proline rich regions. In p85 and , these proline rich regions are separated by a segment

that is homologous to the GTPase-activating proteins for the Rho family of small G

proteins (rho-GAPs) and are followed by the N-terminal src-homology 3(SH3) domain

(Fig. 3). The SH3 and rho-GAPs domain are proposed to have a negative regulatory effect on the catalytic activity of p110 as the p55 and p50 adaptor subunits that lack these domains are better activators of p110 than p85.1

For class IB, the only known isoform is PI3K. It consists of a p110 catalytic subunit

and exists as a complex with a 101kDa adaptor subunit, termed p101 or with another

recently discovered regulatory subunit, called p84 or p87PIKAP.8, 9 PI3K is highly

expressed in cells of haematopoietic origin such as white blood cells, macrophages and

neutrophils while lower levels of PI3K have been found in myocytes, smooth muscle

and endothelia. The p87PIKAP was found to be highly expressed in the heart and is

proposed to play an important role in PI3K related cardiac function.9 The PI3K enzymes are activated by the subunits of heterotrimeric G proteins (G) downstream of G-protein coupled receptors. In vitro, the class I PI3Ks can utilize PtdIns, PtdIns(4)P and PtdIns(4,5)P2 as substrate but the preferred substrate in vivo is the Ptdns (4, 5) P2.1, 2

6

The PI3K family

Class 1A Catalytic Regulatory Regulation Substrate

ABD RBD C2 HELICAL CATALYTIC p110

SH3 GAP nSH2 iSH2 cSH2 p85 p110 ,,  p85, p85, Tyrosine p55, p55, kinases, p50 RAS, PtdIns, Class 1B PtdInd(4)P, G for p110 PtdIns (4,5)P2

ABD RBD C2 HELICAL CATALYTIC p110 p101, p84 or G, RAS p87PIKAP

Class II Tyrosine PtdIns, RBD C2 HELICAL CATALYTIC PX C2 PI3KC2 , ,  kinases, chemokines PtdIns(4)P

Class III

C2 HELICAL CATALYTIC hVps34 p150 Constitutive, ? PtdIns

CATALYTIC HEAT WD40

Figure 3: Members of the PI3K family. The in vivo substrate is denoted in red. RBD = Ras binding domain, ABD = Adaptor binding domain, SH2 = SRc homology 2 domain, SH3 = Src homology 3 domain, GAP = segment homologous to GTPase-activating protein, PX = Phox homology domain.

7

Crystal structure of p110

In Dec 2007, the first X-ray structure of p110/p85 was published (Fig. 4).10 The

structure contains all five domains of p110 and segments of nSH2 (residues 322-430)

and iSH2 (residues 431 -600) of p85

Ras binding domain (RBD)

Kinase domain Helices = red, -sheets = yellow and loops = green

Loop connecting ABD & RBD (White)

Adaptor binding domain Helical domain (ABD)

iSH2 C2 domain

Figure 4: The p110/niSH2-p85 heterodimer. From this view, the crystal has an overall triangular shape with RBD at the tip and iSH2 at the base. ABD and C2 domains interact directly with the iSH2 domain of p85. The C2 domain contains basic residues that might be important for docking to the lipid membrane bilayer. Image created with PyMOL.137 The protein has an overall triangular shape with the RBD forming the tip and the p85

segment forming the base of the triangle. The N-terminus of p110 contains the ABD

(residues 1-108) which is a small globular domain of about 100 residues. It is connected

to another small domain, the RBD (residues 190-291), by means of a long linker

(residues 109-189) made of four helices. The ABD displays a complex set of interactions

between the 1K helix of the kinase domain and helix 1L of the linker between ABD

and RBD. The RBD is linked to the C2 domain (residues 330-524) by a long coil

8

(residues 292 – 329). The C2 domain is a sandwich of two four antiparallel strands.

The 2C5 beta sheet of the C2 domain contains basic residues Lys410, Arg412, Lys413

and Lys416 and is thought to be involved in an interaction with the lipid membrane.

Following the C2 domain is the helical domain (residues 525-696) which contains an all

helical structure. Previously, the helical domain was thought to be the core of the

molecule that binds all the domains together. However in the crystal structure, the suggested role of the helical domain is to bridge the C2 domain and the kinase domain, the only two domains it is in direct contact with.

Figure 5: Kinase domain of C-Lobe p110 . The kinase domain is made of residues 696 – 1068 and is the biggest domain in N-Lobe the structure. It has a typical protein/lipid kinase domain structure made of two lobes. The - helices are in red, - sheets in yellow and loops in green. ATP binds in the cleft between the two lobes. Image created with PyMOL

Catalytic Loop Activation Loop

The kinase domain (Fig. 5) is a typical protein/lipid kinase structure comprising of a two

sub domains or lobes which are separated by means of a cleft. The smaller N-terminal

lobe is made of residues 697 – 851 and contains a five-stranded -sheet and two -

helices similar to that found in other protein kinases. The larger C-lobe consists of

residues 852 – 1068 and is predominantly helical in nature. Residues 933-957 form the

activation loop and 921-928 form the catalytic loop of the p110 catalytic subunit. As in

9

all protein kinases, the ATP binding site is in the cleft region between the two lobes. The

ABD and the C2 domains interact with the iSH2 of the adaptor protein from the top and side respectively and bury a total surface area of 3470A°. None of the other domains

interact with the iSH2 directly.

This crystal structure is a very valuable tool for understanding the spatial arrangement of

the different domains and mutations in these domains that affect the regulation of the enzyme in various pathophysiological states such as cancer.Mutations that are frequently found in cancers are discussed later in this chapter.

Regulation of class I PI3K activity

The catalytic activity of the PI3Ks is tightly controlled in normal cells. The class IA

PI3Ks is activated by receptor tyrosine kinases (RTKs). Binding of growth factors, cytokines and hormones causes dimerization of the RTKs and leads to autophosphorylation of the tyrosine residue present within the YXXM motif in the kinase domain. Phosphorylation of this residue attracts the p110/p85 enzyme complex and interaction with one of the SH2 domains of the p85 causes an allosteric change within the enzyme leading to its activation. Furthermore, the p110 is also reported to be activated by the G subunits of g-protein coupled receptors (GPCRs).

The class IB enzyme is activated by GPCRs. GPCR agonists such as lysophosphatidic acid (LPA), ATP and the fMLP peptide strongly induce PI3K activity. Binding of these agonists to the heterotrimeric GPCRs causes the release of the G subunit which binds to the p101 adaptor subunits and activates p110.

In addition to the above, the class I PI3Ks are also activated by the small GTPase RAS which upon binding to the RBD induces a conformational change in the C2 domain and

10

in the C-lobe of the catalytic domain. This conformational change increases the membrane association of p110 by exposing the C2 domain and also increases its affinity for PI substrates. Ras activation can be independent of the regulatory subunits or have a synergistic effect to their activation. Another mechanism of PI3K activation involves nuclear hormone receptors such as the estrogen and androgen receptors potentially through the activation of receptor tyrosine kinases.11

Deactivation of the PI3Ks in brought about by phosphatases that dephosphorylate the 3’

and the 5’ position of phosphoinositides. Negative regulation is mainly accomplished by

the action of phosphatase PTEN (phosphate and tensin homology deleted on chromosome

10) which dephosphorylates the D3 position of PIP2 and PIP3, thus counteracting the

activity of PI3Ks. PTEN consists of a C2 domain that allows for its anchoring to the

phospholipid membrane and a phosphatase domain that carries out the

dephosphorylation. PTEN mutations are very frequent in some cancers and these cancers

would be good candidates for PI3K inhibitor therapy. Another phosphatase that

negatively regulates the PI3K pathway is SHIP (SH2-containing inositol phosphatase)

which dephosphorylates the PIP3 at the 5’-position. SHIP enzymes are characterized by

an N-terminal SH2 domain and a C-terminal proline rich region that allow for a range of

protein-protein interactions.12

Biological effects of class I PI3Ks

The PIP3s produced by the PI3Ks interact with a variety of proteins through distinct structural motifs and influence different physiological events such as growth, proliferation, membrane trafficking and survival (Fig. 6).

Growth

11

The PI3K pathway plays a crucial role in controlling cell size. PI3Ks control the activity

of the nutrient sensor mTOR by regulating the activity of PDK and PKB/Akt. Activated

mTOR and PDK phosphorylate p70-S6K and E1-BP1 and promote translation of mRNAs

containing a 5-polypyrimidine tract that ultimately leads to protein and lipid

biosynthesis.13 Over expression of class I PI3K in the fruit fly D.melanogaster led to an

increase in growth and cell size while decreased PI3K signaling led to reverse effects.2

Studies in mice show that disruption of genes encoding the p110 and p110 leads to early-stage lethality during embryonic development.14, 15

Proliferation

The PI3K pathway plays a critical role during proliferation by regulating the S phase entry of the cell in the cell cycle. The cell cycle is highly influenced by the actions of the cyclin dependent kinase (CDK) complexes and its inhibitors. Increased levels of cyclin

D1 are necessary for the transition of the cell from the G1 phase to S phase. The cyclin

D1 levels are regulated by glycogen synthase kinase 3(GSK3), which by phosphorylation of the Thr286 of cyclin D1 facilitates its degradation by proteases.

PKB/Akt inhibits GSK3 by phosphorylation and thus increases cyclin D1 levels. In addition, PKB phosphorylates the forkhead transcription factors (e.g. FOXO4) causing decreased expression of CDK inhibitors such as p27kip and allows for cell cycle

progression.16

Survival

The PI3K/Akt pathway plays an important role in cell survival and protects cells from

programmed cell death or apoptosis. Phosphorylation of Akt results in the inhibition of

proteins involved in apoptosis such as Bad (pro-apoptotic Bcl-2 family member) and

12

RTK GPCRs R2 R1

O O O O

p110 , ,  O O p110

P p101 p85 OH P P O O 6  2 1 OH P  4 P 3 5 SHIP P

PTEN PDK1 PKB/AKT

NF-kB BAD FOXO Rac/Rho MDM2 mTOR GSK3

Apoptosis Actin cytoskeleton Growth, rearrangement, Proliferation, Migration Survival

Figure 6: PI3K signaling pathways. Activation of Class I PI3Ks by receptor tyrosine kinases (RTK) or G-protein coupled receptors (GPCRs) lead to the synthesis of PIP3 that acts as a second messenger for a plethora of proteins. The main effector downstream of PI3K is PKB/Akt that mediates the activation and inhibition of several target proteins resulting in cell growth, proliferation and survival. PTEN (phosphatase and Tensin homology deleted on chromosome 10) and SHIP (SH2-containing inositides 5’- phosphatase) cleave the phosphate group from 3 and 5 position of the inositol ring respectively and abrogate the PI3K-induced response. PKB = protein kinase B, PDK 1 = phosphoinositides dependent kinase, MDM2 = murine double minute oncogene is a negative regulator of the p53 tumor suppressor, mTOR = mammalian target of rapamycin, GSK3= glycogen synthase kinase, NF-B = nuclear transcription factor, Bad = BCL-xl/BCL-2 associated death promoter, FOXO = Forkhead in human rhabdomyosarcoma , Rac/Rho = small GTP-ases. caspase-9. Phosphorylation of the forkhead transcription factors (FOXOs) by PKB/Akt prevents the expression of pro-apoptotic protein Bim. Additionally, activation of NF-B by PKB induces the expression of anti-apoptotic genes such as the inhibitor-of-apoptosis proteins (IAPs). Another key function of PI3K/Akt activation is the induction of glycolysis which is required for Akt-dependent survival.

Migration

13

PI3K signaling promotes cytoskeletal rearrangements and cell motility by activating the

guanosine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs)

for small GTPases of the Rac family. Rac proteins direct the polymerization of peripheral

actin assembly that results in formation of lamellipodia. Among the various Rac-GEFs,

members of the Vav, Sos, Tiam, PIX, SWAP-70 and P-Rex families have been suggested

to be regulated by PI3K in vivo. Evidence suggests that the Rac protein can also activate

the PI3Ks, but the mechanism is not yet clear.17

Class II

The class II PI3Ks are monomeric proteins of higher molecular weight (170 – 210kDa)

due to extensions at both the N- and C- termini (Fig. 3). Extensions at the N-terminus are

more distinct between the different class II isoforms while extensions at the C-terminus

are more conserved. In contrast to the class I enzymes, the class II enzymes lack the

adaptor binding domain but contain a defining C-terminal C2 domain and are often

referred to as PI3KC2. The C2 domain is a Ca 2+ binding motif present in various

proteins and is involved in Ca2+ regulated binding of those proteins to the acidic phospholipids. However for the class II PI3Ks, certain aspartic acid residues necessary for Ca2+ binding are absent and thus the exact function of this domain is not yet known.

All isoforms of class II PI3K enzymes also contain a PX (Phox homology) domain in the

C-terminus. Three isoforms of Class II PI3Ks exist namely, PI3KC2,  and  which are

encoded by three separate genes. PI3KC2 is widely expressed with highest levels in

heart, placenta and ovary. PI3KC2 is highly expressed in the thymus and placenta while

the C is found in the liver and prostrate glands. In contrast to the class I PI3Ks which

are mainly cytosolic, the class II PI3Ks is mostly associated with membrane fractions of

14

cells. Deletion of either the C-terminal C2 domain or the PX domain of PI3KC2 does not affect the localization or the activity of the enzyme and hence their function remains unresolved.18

In vitro the Class II PI3Ks can phosphorylate both PtdIns and PtdIns (4)P and the

preferred in vivo substrate is thought to be PtdIns(4)P; however a definitive classification

of its in vivo products is not yet available. Known stimuli that can activate the class II

PI3Ks include MCP-1 chemokines, cytokines such as leptin and TNF, the glycerol based phospholipid LPA (lysophosphatidic acid), clathrin and insulin. With regards to the physiological function of these enzymes, few reports indicate that the PI3KC2 isoform plays an important role in cell migration while C2 has been associated with insulin signaling.19

Class III

The only member of class III PI3K is a homolog of the yeast vacuolar-protein-sorting

protein 34 (Vps34). Its sole substrate in vitro is PtdIns which is most likely due to the

uncharged amino acids that make-up its substrate recognition site. In comparison the

class I & II PI3Ks contain basic residues and hence can accommodate PtdIns(4)P and

PtdInds(4,5)P as substrates. Nonetheless, the class III PI3K shows considerable

homology with the catalytic units of other PI3Ks. The Vps34 protein has an

uncharacterized ~50 amino acid N-terminal region followed by the C2, helical and kinase

domains. The extreme C-terminal 11 residues of Vps34 are necessary for the lipid kinase

activity. Consistent with this observation, antibodies that target the C-terminus of

mammalian hVps34 are potent inhibitors of this enzyme. The Vps34 enzyme is closely

associated with a protein kinase Vps15 (or p150 in mammals) which is thought to be a

15

Vps34 regulatory subunit. Vps15 contains an N-terminal myristoylated consensus

sequence followed by the protein kinase domain, a central HEAT (Huntington,

elongation factor 3, the PR65/A subunit of protein phosphatase 2A and the lipid kinase

TOR) domain and a series of WD-40 domains at the C-terminal (Fig. 3). The importance

of the Vps15 protein for Vps34 activity is yet to be determined in mammals, though it is

important for Vps34 activity in yeast.20 Vps34 is thought to play an important role in

endocytic trafficking, nutrient sensing in the mTOR pathway, autophagy and also in

GPCR mediated regulation of MAPK pathways. 20

Therapeutic applications of the PI3K/Akt pathway

Cancer

The PI3K/Akt pathway is an important contributor to the processes of cell growth,

proliferation, survival and migration. These processes are also involved in tumorigenesis

and cancer. In fact, aberrant activity of the PI3K/Akt pathway is common in a variety of

cancers.22 The phosphatase PTEN acts as a tumor suppressor by counteracting the

activity of PI3K. Mutation and/or deletions in the PTEN gene accounts for the highest number of PI3K/Akt pathway related human tumors. Human tumor cells with mutated and/or deleted PTEN show elevated levels of PIP3 as well as increased activity of PI3K effector proteins. Somatic mutations or deletions of PTEN have been identified in a variety of human tumors including prostrate, brain, breast, endometrial, cervical, lung and

colorectal cancers. 21 In addition, PTEN germline mutations have been observed in

human autosomal disorders such as Cowdens disease (CD), Lhermitte-Duclos disease

(LDD) and Bannayan-Zonana syndrome (BZS).21

16

Another mechanism of aberrant PI3K activation is by amplification and/or mutation of

the PI3K genes. A recent study reports PI3K gene mutations occurring in a variety of

cancers.22 Interestingly, of all various PI3K isoforms, only PIK3CA, which encodes for the p110 catalytic subunit showed somatic mutations. Mutations in PIK3CA were observed in 25 - 32% of colon, brain and gastric cancers and about 4% of lung cancers. In other studies PIK3CA mutations were observed in 25-40% of breast cancers.22

Fascinatingly, the majority of the mutations are found in the helical domain (47%), and the catalytic domain (33%), while 8-10% of total mutations were found in the ABD and

C2 domains while no mutations were observed in the RDB of p110. Two amino acids that were mutated in the majority of the tumors were Glu545 in the helical domain and

His1047 in the catalytic domain. Expression of recombinant H1047R p110 has shown to posses higher lipid kinase activity than the wild-type. The crystal structure of p110 indicates that His1047 lies in the 12K helix of the C-terminal lobe and is close to the activation loop. Mutation of His1047 is thought to alter the conformation of the activation loop and modify its interaction with the phosphoinositide substrate. Mutations of Glu545 are known to abolish the inhibitory effect of nSH2. Glu545 mutations in the helical domain are proposed to modify the orientation of the nSH2 domain in relation to the helical and kinase domain and also alter the relative positions of the 11K and 12k helices. Somatic mutations in the p85 subunit have been observed in human colon and ovarian cancers which lead to the constitutive activation of the PI3K enzymes.23

Non-selective PI3K inhibitors such as and LY294002 have been shown to

block the proliferation of various cancer cell lines in vitro and in xenograft studies. These

inhibitors also enhance the effects of radiation and cytotoxic agents in various cancer cell

17

lines.24, 25 All these observations promote the PI3K/Akt pathway as a promising target for

the treatment of cancer and PI3K selective inhibitors are being explored as

chemotherapeutic agents.

Cardiovascular Diseases

The PI3K pathway plays an important role in modulating the GPCR mediated 2-

adrenoceptor (AR) signaling.26 The p110 interacts directly with the 2-adrenoceptor

kinase 1(BARK1) via the PIK domain and ultimately leads to termination of -

adrenoceptor signaling. Activation of ARs in the heart lead to the stimulation of

adenylyl cyclase and production of cAMP. A study using cardiac specific PTEN

knockout mice show that mice lacking the PI3K isoform in the heart have increased

levels of cAMP leading to increased myocytic contractility. 27 This study also suggests

that PI3K regulates cardiac function by decreasing the levels of cAMP and phospholamban phosphorylation upon -adrenergic stimulation. Thus inhibitors of the

PI3K isoforms may have therapeutic applications in heart failure where the PI3K is

known to be up regulated.

In another study mice lacking the GPCR activated-PI3K activity were protected from

hypertension induced by the administration of angiotensin II. This study demonstrates

that PI3K is a key transducer of intracellular signals in response to angiotensin II and inhibition of the PI3K might aid the management of hypertension.28 A recent study links

PI3K to artherosclerosis and multiple sclerosis and identifies p110 as a potential

therapeutic target for the management of human artherosclerotic cardiovascular disease.29

Platelet activation involves a three step process comprising of initiation, propagation and perpetuation that ultimately lead to formation and stabilization of blood clots. Platelets

18

contain all class I PI3Ks as well as class II PI3KC2 isoforms.30 Studies show that PI3K

deficient mice have defective platelet aggregation and a decreased thrombotic

tendency.31, 32 Similarly, inhibition of the PI3K isoform leads to a significant defect in

arterial thrombosis.33 Recent studies using PI3K inhibitors wortmannin and LY294002

show that the PI3K/Akt pathway is vital for keeping the platelet aggregates together and

for maintaining the continuous active status of the fibrinogen receptor (Integrin II23)

which is necessary for stable platelet adhesion and aggregation. 34 Additionally activation

of the PI3K isoform was shown to mediate the individual platelet-platelet contacts while

PI3K was deemed necessary for maintaining the integrity of the entire thrombus.34 All these observations suggest that the PI3K and  isoforms are useful targets for antithrombotic therapy.

Inflammation and Immunity

A key step in the progress of inflammation is the migration of leukocytes to the inflamed area in response to chemokines and other chemoattractants. Chemokines and chemotactic peptides such as n-fMLP, C5a, Interleukin-8 (IL-8) bind to GPCRs resulting in their activation and release of the G subunits. In phagocytic cells this triggers a series of events leading to directional cell movement, phagocytosis, degranulation and superoxide production. Leukocytes express all 4 class I PI3Ks but the PI3K and  have gained a lot of recognition with regards to inflammation. An early event in response to chemoattractant stimulation is the polarized activation of the PI3K/Akt pathway at the leading edge of migrating cells. Studies in mice suggest that the PI3K is essential for this event and it also plays an important role in the respiratory burst and motility of neutrophils.35 Mice lacking SHIP-1 show progressive and severe pulmonary

19

inflammation of macrophages, lymphocytes, neutrophils, and eosinophils.36 A recent

study shows that PI3K selective inhibitors are able to reduce the migration of

neutrophils in vivo and also ameliorate CII-induced arthritis.37 All these observations present the PI3Kas a viable target for inflammatory diseases.

The T cells and B cells are vital components of the adaptive immune response and in these cells the PI3K pathway is activated within seconds of antigen-receptor signaling.

Studies involving PI3K “knock-out” and knock-in” mice suggests that the p110 is required for development of marginal zone B cells and peritoneal B cells.38, 39 These studies also suggest a role for p110 in differentiation and/or survival of effector and memory T-cells. The PI3Kis associated with thymocyte development, T cell activation, neutrophil migration and the oxidative burst.40 These observations implicate the PI3K

and  as important targets in autoimmune diseases and graft rejection.

PI3K inhibitors

The PI3K/Akt pathway is the focus of tremendous research in academia and industry

alike and there is interest in developing compounds that selectively inhibit the class I

PI3Ks. Wortmannin and LY294002 are the earliest known PI3K inhibitors and have played a vital role in dissecting the PI3K/Akt pathway. Over the years several chemical classes of PI3K inhibitors have been reported. Most of these are pan inhibitors that show activity on all the four isoforms and only a limited number of isoform selective inhibitors have been reported. As in the case of protein kinases, almost all PI3K inhibitors are competitive ATP inhibitors. The PI3K inhibitors can be roughly divided into two categories - natural and synthetic.

20

Natural compounds and their derivatives

Wortmannin (1), a fungal metabolite isolated from Penicillium wortmannii was initially reported to be a Myosin light chain kinase inhibitor but was later established to be a more

41 potent inhibitor of PI3Ks (IC50 of 4.2nM). Early site-directed mutagenesis studies suggested that wortmannin inhibits PI3K by forming a covalent bond with Lys802 present in the active site of the PI3K enzyme (Fig. 7A). The -amino group of Lys802 attacks the C-20 position of wortmannin and leads to opening of the furan ring and formation of an enamine.42, 43 This hypothesis was later supported by the X-ray structure

Figure 7: Natural compounds and their derivatives as PI3K inhibitors

A O O

H3C O H3C O O CH3 O CH3 H3CO H3CO

CH3 C D CH3 O O A B H H

O O O O

O HO

NH NH 2 Lys802 Lys802 (1) Wortmannin Mechanism of inhibition by wortmannin

B O OH HO OH H3C O O H3C HO H3CO CHO

CH3 O H O O H3C OH H N CH3

H

OH (2) PX866 (3) Liphagal Semisynthetic demethoxyviridin (4) Resveratrol derivative.

21

of p110-wortmannin (1e7u) where a covalent bond is seen between the furan ring and

the equivalent Lys833 (Fig 12B).58 This lysine residue is conserved in many protein and lipid kinases and as a result wortmannin inhibits an array of kinases including mTOR,

DNA-PK (DNA - dependent protein kinase), ATM (ataxia telangiectasia mutated kinase) and PI4K, although at higher concentrations than that needed for PI3K inhibition.43, 56

Consequently, the use of wortmannin as an anticancer agent is hampered by its off-target toxicity and also due to low biological stability.

A structurally similar natural product demethoxyviridin (DMV) isolated from the fungi

Nodulisporium hinneleum has led to the development of biologically stable semi- synthetic analogues such as PX-866 (2) (IC50 = 0.1nM).44 Xanthine derivatives e.g. caffeine and theophylline show PI3K inhibition in the micro to milli molar range.39

Recently, Liphagal (3), a meroterpenoid derived from the marine sponge Aka coralliphaga was reported to be a selective PI3K inhibitor (IC50 = 100nM) with ten fold selectivity over the PI3K isoform.45 Resveratrol (4) (3,5,4’-trihydroxystillbene), a component of red wine has been shown to be a Class IA PI3K inhibitor with IC50 of ~

25M (Fig. 7B) .46

Synthetic compounds

First generation - LY294002 and related compounds

LY294002 (6) was the first synthetic PI3K inhibitor and was based on the natural

compound (5). LY294002 (IC50 = 1.4M) is 500-fold less active than

wortmannin, but has proved to very useful due to its superior chemically stability. (Fig.

8A).47 Nevertheless, similar to wortmannin, LY294002 is not specific for PI3K and

22

inhibits mTOR and DNA-PK downstream of PI3K as well as other protein kinases such as CK2 (casein kinase -2) and Pim-1. 159

Unlike wortmannin, LY294002 is a reversible competitive ATP inhibitor. The crystal structure of LY294002 - PI3K(1e7v) shows the morpholino oxygen forming an H-bond with the backbone amide of Val882 while the chromone ketone H-bonding with Lys833

(Fig. 12C). 63 Interestingly, the binding mode for quercetin is different than that of

LY294002; in quercetin the chromone ketone interacts with Val882 while the 3-OH group on the resorcinol ring forms an H-bond with Lys833.

SAR studies on LY294002 have led to development of compounds such as TGX-126 (7) and TGX-155 (8) in which the chromen-4-one ring is replace by pyrido(1,2-a)pyrimidine-

4-one and 1H-quinolin-4-one ring systems respectively (Fig. 8B). These compounds are

Figure 8: LY294002 and related compounds

O A OH O OH

O N HO O O OH OH

(5) Quercetin (6) LY294002 [Chromen-4-one ring system] B O O H3C N

N N N N HN O H O O

H3C F

(7) TGX – 126 (8) TGX – 155 [Pyrido (1,2-a)pyrimidine-4-one] [1H – Quinolin-4-one]

23

>200-fold selective for the / isoform over the  isoform.48

Second Generation compounds

After the advent of the first generation inhibitors various other scaffolds have been employed for the development of PI3K inhibitors and more focus has been directed

Figure 9: Second Generation PI3K Inhibitors

Isoform selective inhibitors

N S F O S O O

NH F O NH N O O

(9) AS605240 (10) AS604850

Thiazolidinediones (PI3K)

N CH3 O CH O 3 N N Br N CH3 N OCH3 N N N N N O N O S 2 N NH CH N S 3 N N N O H2N CH3 (11) IC87114 (12) PIK-39 (13) PIK-75 Imidazopyridine (PI3K/)

Quinazolinone purines (PI3K)

Pan-PI3K inhibitors (Multi targeted)

O N Cl N CH3 O N N H N HO S N NH O N N O S N H3CO NH OCH3 O N O OH CH3

(14) PI-103 (15) PIK-90 (16) PIK-93 Pyridinylfuranopyrimidine Imidazoquinazoline Phenylthiazole

24

towards the development of isoform selective compounds. A series of thiazolidinediones

(e.g. 9, 10) has been shown to inhibit the PI3Ks with nanomolar potency and show selectivity for the p110. These compounds suppressed joint inflammation and damage in mouse models of rheumatoid arthritis by blocking the PI3K37 A patent application by

ICOS Corporation reported a series of quinazolinone purine based PI3K selective

compounds (e.g. 11, 12) that have been used for studying the role of p110 in chemotaxis

and inflammation response of neutrophils.49 A series of imidazopyridines have been reported recently that show improved potency and selectivity for the p110 (e.g. 13, IC50

= 0.3nM).50, 51 Several other chemical classes have been reported that inhibit multiple

PI3Ks with varied affinity and include the pyridinylfuranopyrimides (e.g. 14), the

imidazoquinazolines (e.g. 15) and the phenylthiazoles (e.g. 16).51, 56

In a recent study, the polyisoprenyl phosphate presqualene diphosphate (17a) was shown to be an endogenous inhibitor (IC50 = 38pmol) of PI3K activity in neutrophils.

Moreover, a structural mimetic of PSDP (17b) blocked human neutrophil activation as

well as mouse lung PI3K activity and inflammation.55

Figure 10: Endogenous PI3K inhibitors.

OEt H H O O O O P OEt O P O P O N H H OH OH H P OEt CH3 CH3 O OEt

(17a) PSDP (17b) PSDP mimetic

PI3K inhibitors in clinical trials

Currently, a few PI3K inhibitors are in clinical trials. These include the demethoxyviridin

analog PX-886 (2) for treatment of advanced solid tumors, the pyridopyrimidinones

25

XL147, XL647 and XL765 (Exelixis) for treatment of breast cancer, mixed gliomas and

non-small-cell lung cancer and the theinylpyrimidine GDC-0941 (18, Genentech-

Piramed) for Non-Hogdkin’s lymphoma and solid cancers.57, 61 Recently, the combination

of PI3K and mTOR inhibitors was reported to be very efficacious for the treatment of

PI3K/Akt related cancers.58-60 Accordingly a novel strategy in PI3K/Akt pathway inhibition is to design multi-targeted kinase inhibitors that have optimal selectivity profile. Two such compounds BEZ235 (19) and BGT226 (structure unavailable) developed by Novartis are dual PI3K/mTOR inhibitors and currently in Phase I/II clinical trials for treatment of solid tumors, breast cancer and Cowden’s disease. 57

Figure 11: PI3K inhibitors in clinical trials.

R4 R5 CH3 R6 N N H3C R2 X N N O O N R1 N N XL series by Exelixis- Phase I CH3

O N (19) BEZ235- Phase I/II N

S N N NH N N O N S H3C O (18)GDC-0941 – Phase I

Inhibitors and crystal structures

Following the publication of the first X-ray structure of the porcine p110 in the year

2000, 17 crystal structures involving various inhibitors have been deposited in the PDB

26

(as of January 2009). These structures provide valuable information about the active site of PI3Ks. The binding features of the natural substrate i.e. ATP and three inhibitors are discussed below.

The porcine and human PI3K have 95.3% overall sequence identity and complete identity within the ATP binding site. The ATP binding site is present at the cleft between the C-lobe and N-lobe of the catalytic domain. The C-lobe forms part of the ATP binding site as well as the binding site for the phospholipid substrate. The N1 nitrogen of ATP forms a hydrogen bond with Val882 while the amine at the 6-position interacts with the backbone carboxylic group of Glu880 (Fig. 12A). The triphosphate group forms a network of H-bonds involving water molecules and residues Lys833, Asp964, Ser806,

Asn 836 and Asn951. The ribose sugar faces the outside of the binding pocket and is in close vicinity of Trp812, Ile831 and Met804 on one side and Thr887 and Met953 on the other. The region opposite the ribose sugar (i.e. interior of the pocket) is similar to the hydrophobic region 1 commonly observed in protein kinases and is bordered by Tyr867 and Ile879. 62

With regards to the different p110-inhibitor complexes, the most common hydrogen bonding is between the inhibitor and backbone amide of Val882; 13 of 17 structures show this bond. Furthermore, 8 of 17 inhibitors also show hydrogen bonds with the - amino group of Lys833. None of the inhibitors extend as far as the  and  phosphates of

ATP within the active site or form as many hydrogen bonds as ATP.

Wortmannin binds deep in the pocket and causes a fairly large conformational rearrangement (Fig. 12B). The D ring is located in the same position as the N1 of the adenine. The ketone group of the D ring interacts with the backbone amide of Val882

27

while the furan ring forms a covalent bond with Lys833. The ketone group of the A ring

interacts with Ser806 while the ketone group of ring B interacts with the backbone amide

of Asp964 and the hydroxyl of Tyr867. The hydroxyl group on the B ring generated as a

result of the furan ring opening (Fig. 7A) interacts with the acid group of Asp964. The

C10 methyl group occupies a pocket created by residues Ile831, Met804 and Pro810

while the C13 methyl group occupies a pocket created by Ile831, Trp812, Ile881 and

Glu880. The methyl ether at C1 extends towards a pocket created by residues Ile963,

Asn951, Asp964, Asp950 and Lys807. The acetoxy group at the C11 position lies in

A] ATP/PI3K B] Wortmannin/PI3K Thr887 Thr886 Asp950 Asp964 Asn951 Met953 Tyr867

Asp964 Ser806 Tyr867 Val882

Val882 Lys833

Ser806 Ile879 Met804 Glu880 Glu880 Lys833 Trp812

Ile881 Met804

Ile831

C] LY294002/PI3K D] PIK – 39/PI3K

Val882

Val882

Met804 Lys833 Glu880 Glu880

Met804 Lys833

Trp812 Trp812

Figure 12: Binding mode of different compounds in the PI3Kactive site. Most inhibitors form hydrogen bonds with Val882 and Lys833. Met804 undergoes a conformational change and flips by more than 90° to accommodate the quinazolinone portion of PIK-39 thus providing selectivity for the PI3K isoform.

28

same region as the ribose of ATP and is in close vicinity of Met804, Trp812 and Ile 881

on one side and Met953 on the other. Modification of the acetoxy group to a more

hydrophobic butryl group has shown to increase affinity for the PI3Ks possibly by

hydrophobic interactions. 63

LY294002 makes fewer contacts than wortmannin with the ATP binding site since it has only two H-bond acceptors (Fig. 12C). The morpholino ether oxygen lies in the same

area as N1 of adenine and interacts with the backbone amide of Val882. The ketone group of the chromone portion forms putative hydrogen bond with -amino of Lys833.

The phenyl group at the 8 position is positioned in the same region as the ribose sugar but extends further towards the exterior of the pocket and is in close vicinity to Met804 and

Trp812 on one side and Met953 on the other. The phenyl group at the 8-position increases the affinity for the enzyme in comparison to a compound without a phenyl

group in this position of the molecule. 63

PIK-39 (12) is a quinazolinone purine that shows high affinity and selectivity for p110.

Reported IC50 for PIK-39 is in the mid nanomolar range for p110, while it is ~100 fold higher for  and  isoforms. PIK-39 shows no activity against p110 as well as other members of the PIK family at concentrations up to 100M. The crystal structure of this inhibitor reveals a very unique binding mode for PIK-39 (Fig. 12D). 51 The purine ring of

the molecule lies in the interior of the binding pocket and makes contacts with the

backbone of Val882 and possibly with Glu880. The rest of the molecule is nearly

perpendicular to purine ring and extends towards the entrance of the ATP binding pocket.

In order to accommodate the molecule, the Met804 undergoes a conformational change

and flips almost at right angles thus aligning itself parallel to the length of the imidazole

29

ring. Even though the Methionine residue is conserved in all four class I PI3Ks, it is part of a highly flexible loop akin to the glycine rich loop or G-loop of protein kinases that plays an important role in proper alignment of the ATP molecule. Furthermore, the amino

acids at the entrance of the ATP binding pocket in the different isoforms have low

sequence identity and such a shift in this region would induce dissimilar conformational

change that might be affecting the binding of PIK-39. Hence, though the exact reason for

this selectivity is not yet known, the differential sequence and conformational plasticity

within the region might be an essential feature to consider while designing isoform

selective inhibitors.

Goal, hypothesis and specific aims

Inhibition of the class I Phosphatidylinositol-3 Kinases (p110, ,  and ) provides

opportunities for treatment of inflammation, autoimmune and cardiovascular diseases and

cancer. The PIK3CA gene that encodes the p110catalytic subunit is frequently mutated

in a variety of cancers and the PI3K is an important therapeutic target in cancer. Though

several classes of compounds that inhibit the class I PI3Ks have been reported,

development of inhibitors selective for PI3K still remains a major challenge. Although

the availability of various p110 X-ray structures has improved understanding of binding modes of different inhibitors, those structures do not effectively facilitate rational design

of p110 selective inhibitors. More importantly, the p110 X-ray structure was only

recently described after we built our own model.10 Understanding the features that

distinguish the p110 from the other isoforms is essential in the development of PI3K

selective inhibitors. Therefore, the overall goal of this project was to explore differences

30

between the p110 and  binding sites. The main hypothesis underlying this research was

Homology modeling of p110 facilitates development of PI3K inhibitors with potential isoform selectivity.

The hypothesis was tested using the following specific aims –

I. Building a p110 homology model and designing a new scaffold.

A p110 homology model was generated from the p110 X-ray structure using Modeller.

This model facilitated design of a novel 2,6-disubstituted isonicotinic scaffold for exploring preferential p110 and  binding (Chapter II).

II. Evaluating the scaffold.

The synthetic feasibility of this scaffold was evaluated by designing and synthesizing an initial set of derivatives. The biological activity of these compounds was determined on recombinant purified PI3K and  using an in vitro lipid kinase assay. In addition, the potency of these derivatives was evaluated using a PI3K dependent cell survival assay

(Chapter III).

III. Structure-activity relationship (SAR) of the scaffold.

Based on docking studies using the p110 homology model, three regions of the scaffold were modified and differences in efficacy between p110 and  were explored. (Chapter

IV).

31

II: Molecular Modeling

Protein Structure Prediction

The 3D structure of a protein is a valuable tool in biochemistry and drug design. It

provides insight regarding the structure of the protein, its function, the binding of ligands

and other proteins and mutations that can affect its activity. Experimental methods for

determining 3D structure include X-ray crystallography, NMR spectroscopy and electron

microscopy (EM, especially for large macromolecules, mol. Wt > 500KDa). As of

December 2008, the PDB contained 47315 X-ray structures, 7640 structures determined

by NMR, 211 by EM and 105 by other techniques such as electron/neutron diffraction,

solution scattering and IR spectroscopy.64 Despite tremendous advancement in these

fields, crystallization of proteins is still a major challenge as is the NMR determination of

macromolecules. 3D structures solved by EM are usually of low resolution and as such

this technique is in its infancy stage. Furthermore, owing to the human and other genome

sequencing projects, the number of sequenced proteins is growing at an enormous rate as

compared to the number of available 3D structures (Fig. 13). As of December 2008 the

UniProt knowledgebase/Swiss-Prot reported a total of 405506 protein sequences while

the UniProt/TrEMBL database (collection of protein sequences derived from computer

translation of genetic information from the EMBL nucleotide database) reported ~7

million protein sequences.65 Accordingly, the prediction of 3D structure using

computational techniques has gained a lot of interest. Literature data involving

computational models and methods for 3D structure prediction have grown exponentially

over the years as have the number of programs and online servers used for this purpose.

The Protein Structure Prediction Center (part of NIH) organizes a biennial blind protein

32

structure prediction contest - the Critical Assessment of Techniques for Protein Structure

Prediction (CASP) to assist the advancement of protein structure prediction methods and

to provide an in-depth analysis of progress in this field.66

Figure 13: Analysis of the Swiss-Prot vs. the PDB databases show a large sequence-structure gap. Graph generated using data from the two databases

Traditionally, computational techniques employed for protein structure prediction are

divided into three categories (i) ab initio (free modeling) (ii) fold-recognition and (iii)

homology (comparative) modeling.67 While homology modeling relies on the availability

of a known template for building the 3D structure, ab initio is a template free technique.

Fold recognition lies somewhere in between and utilizes a template that has little or no obvious sequence relation to the target.

(i) Ab initio method

The ab initio methods predict the 3D structure of the protein entirely on the basis of the amino acid sequence. In 1962, Anfinsen and Haber demonstrated that even after denaturation a protein can attain the same, stable and native conformation just by its reintroduction into the original environment and without the use of any genetic

33

mechanisms.68 This led to Anfinsen’s thermodynamic hypothesis of protein folding

which states that under given environmental conditions the amino acid sequence of the

protein alone contains adequate information for finding its native conformation. Thus the

native structure of a protein is the one for which the free energy achieves a global

minimum. Based on this principle, ab initio methods consider all the energetics involved

in the process of folding and then find the structure with the lowest free energy.

Nevertheless, the vast number of possible conformations that can be adopted by the

protein is one of the main problems for this method as is the validation of the predicted

3D structure. A combination of powerful algorithms and experimentally accurate models

are necessary to overcome these challenges.

Ab initio methods can be broadly categorized as (a) ones that start from a random/open

conformation and simulate the folding process or minimize conformational energy (b)

segment-based approaches and (c) a combination of the two.69 Literature on these

approaches are well documented and some of the well known methods include

TASSER70, ASTRO-FOLD71 and TOUCHSTONE72. One of the more successful ab

initio prediction programs is the ROSETTA algorithm which is a fragment/segment

based approach that narrows the vast conformational searching space by local structure

predictions and builds the 3D protein structure by assembling the local structures of these

fragments. At CASP6 (2004), ROSETTA was able to predict the structure of a 70 residue

alpha-beta protein from Thermus thermophilus with C-RMSD of 1.6 Å compared to the

later derived crystal structure.73 Thus, although the ab initio approach is the most difficult

of the protein structure prediction methods mentioned above it is arguably the most promising since it does not depend on typical knowledge-based assumptions.

34

(ii) Fold-recognition

Through evolution the structure of the protein is more conserved than its sequence i.e. proteins with rather different sequences can fold into homologues structures (Fig. 14). It is estimated than the total number of folds found in nature would only be about 2000 and

74 A the number of unique folds is very low. The SCOP (Structural classification of proteins)

database reported a total of only 1082 different

folds among all existing 3D structures of B experimentally determined proteins as late as Sept

2007.75 Hence the probability of a newly

discovered protein adopting one of the already

Figure 14: Pairs of unrelated proteins discovered folds is very high. In fold-recognition, showing similar architecture. A] Cytochrome B562 (1qpu, left) involved in electron transport and the FKBP- the amino acid sequence of the target protein is rapamycin binding domain (FRB) of the FKBP-rapamycin associated protein of the compared directly to all known folds/structures phosphatidyl kinase related kinase family. B] CheY protein (1ehc, left) involved in bacterial flagellum motion and the and assigned a statistical score which identifies oncogene ras p21 (1wq1, right). These proteins show overall similar topology the possible structure adopted by the target though the secondary structural elements have different sizes and extra peripheral protein. elements. Fold-recognition methods are broadly classified into three categories; (a) Conventional sequence comparison or Profile-based alignment (b) Threading and (c) Structure based alignment.76 As it is often difficult to identify homologus proteins that have similar structure but share little sequence similarity; conventional sequence comparison makes use of multiple sequences from the same protein families to improve the sensitivity of homology detection and to improve the quality of sequence alignment. One successful

35

program using this method is PSI-BLAST (Position Specific Iterative Basic local alignment search tool) which generates a Position Specific Scoring Matrix (PSSM) or

profile by assigning a substitution score for each position in the multiple sequence

alignment.77 Highly conserved positions receive a higher score while the less conserved

ones receive a lower score. This profile is then used to execute a second BLAST search

by performing a sequence-profile alignment and the results are used for refining the

matrix and so forth. Programs based on this approach include FFAS78, COMPASS79 and

COACH80.

In threading a query protein is fitted directly onto the backbone coordinates of a known

protein structure and using knowledge based scoring functions the most compatible

sequence-structure relationship is determined. One of the earliest known programs based

on this method was THREADER.81 As traditional threading methods do not use sequence

information they have the potential to identify structurally similar proteins with no

evolutionary link, nevertheless, the success rate with such methods is very low. Some

recent threading techniques do incorporate sequence information in their calculations and

evaluate the protein structure-sequence compatibility using a statistical potential model

which represents the preference of two types of amino acids to be at a certain spatial

distance. Programs based on this approach include GenTHREADER82, PROSPECT83 and

RAPTOR84.

Structure based alignment can be considered to be an extension of the threading method

which incorporates specific information about the 3D structure of proteins such as secondary structure, the accessibility of amino acids, etc. This approach generates a 3D- profile that evaluates the frequency of a particular amino acid appearing in a particular

36

structure environment. Programs based on this approach include 3DPSSM85 and

FUGUE86.

(iii) Homology modeling

Homology modeling builds 3D structures of target proteins based on experimentally

derived structures of homologues proteins. This method is based on the concept that

evolutionary related proteins have similar conformations and hence the experimental

structure of a protein can serve as a starting model for other members of its evolutionary

family. Of the available protein prediction methods homology modeling is the best and

the most used method primarily because the quality of the model based on reasonably

close evolutionary relationships has been shown to be more accurate on average than

those produced by the other techniques. Secondly the reliability of the final model

depends on the quality of the available experimental data which helps one decide whether

a sufficiently accurate model can be obtained for a particular study. A traditional

comparative modeling procedure involves the following steps: 87, 88

1] Identify proteins of known structure that are evolutionarily related to the target protein,

2] Identify the structurally conserved regions (SCRs) and structurally variable regions

(SVRs), 3] construct a reliable alignment between the target protein and the reference

protein for the SCRs 4] construct the SCR for the target protein using the coordinates

from the template structure, 5] construct the SVRs 6] model the positions of the side-

chains of the target, 7] optimize the final 3D structure using energy minimizations and

molecular dynamics.

Detection of reference proteins and sequence alignment: Identifying a reference protein

by searching and comparing thousands of protein sequences from data bases can be very

37

difficult and time consuming. Hence, heuristic methods such as FASTA89, BLAST90 and

PSI-BLAST91 have been employed that do not always necessarily find the global optimal solution but very rarely miss a sequence with a significant match.

The sequence alignment step is of utmost importance for comparative modeling since these methods not only identify similar sequences but also detect the correspondence between the amino acids of the structurally known protein and those of the target protein that form the basis of transferring the coordinates of the reference protein to the target protein. Sequence alignment algorithms can be either global or local. A global alignment

contains the entire sequence of the protein while the local alignment focuses on regions

of the sequences with the highest similarities. One of the well known global alignment

algorithms is the one developed by Needleman and Wunsch.92 It guarantees the best

alignment for the two sequences and is used by programs such as ALIGN, BESTFIT and

GAP.93 However one drawback of global alignment is that it is slow and hence, local

alignment methods were generated to overcome this problem. Widely used programs that

implement local alignment strategies include FASTA and BLAST. In comparison to the

above methods that perform only pair-wise sequence alignment recent methods such as

ClustalW94, ClustalX95 and MAXHOM96 carry out multiple sequence comparisons.

In order to assess the probability of a particular amino acid substitution being acceptable

in the aligned sequences a scoring function is used that is expressed in the form of

substitution matrices. These matrices can be based, e.g., on chemical similarity between

the amino acids or the minimum number of base substitutions needed to transform a

triplet coding for one amino acid into the code for the other. The most commonly used

matrices are PAM (percent accepted mutation)97 and BLOSUM98 (blocks substitution).

38

Identification and construction of the conserved core regions: Homology modeling relies

on the fact that proteins belonging to the same family have practically identical regions in the 3D structures that serve as framework for assignment of the atomic coordinates. A multiple sequence alignment of the protein family if available makes it easier to identify the conserved core. Even more advantageous is the availability of more than one crystal structure of related proteins whereby structural superimposition can help identify regions of the protein that are conserved as well as regions that are more prone to change their structure through evolution. Generally, recognizing the conserved regions of the proteins is done by superimposition of the proteins by means of least-square fitting methods. As a first approximation, the structures can be superimposed by least-square fitting of the C atoms and subsequently optimized using corresponding conserved secondary structural elements e.g. -helices, -turns etc that occupy the same conformational space throughout the entire protein family.87

Construction of the structurally variable regions(SVRs): The SVRs are customarily called

“loops” irrespective of whether they consists of large regions that have undergone local

refolding during evolution or relatively short regions where insertions and deletions

occur. Modeling of the loops is a very challenging task which is further complicated by

insertions and deletions in the sequence due to differences in the number of amino acids.

Segments of equivalent length in the homologues protein having similar amino acid

character are a good guide for modeling these loops as they are known to have similar conformations. In the absence of a corresponding loop, two approaches can be used for

their modeling – a] loop search method where coordinates can be obtained from peptide

segments found in other proteins and that fit into the model’s spatial environment or b]

39

the loop can be generated de novo. The loop search method scans possible peptide

segments in the PDB based on specified geometry input such as distance and coordinates

and retrieves a set of loops that best fit the criteria. These fragments can be evaluated on

the basis of quality of fit to the residues containing the loop region, by examining the

sequence identity between the original loop and the retrieved loop or by evaluating the

steric interactions and energy criteria. If this approach does not provide satisfactory

results, one can try to create the conformation of the loop de novo by using algorithms that randomly generate numerical values for all the backbone dihedral angles of the peptide chain. However, due to the complexity of the calculations this technique can only

be used for loops smaller than seven residues.87, 88

Side-chain modeling: After generation of the backbone, the next step is addition of the

side-chains. Side-chains interact with each other and the energy contribution of their

interactions is important for stabilization of the 3D conformation of the protein. Finding the correct combination of these side-chain conformations is a much more complex problem than predicting the backbone conformation. As conserved residues in homologues proteins tend to adopt similar conformations, their side-chain angles are usually taken from the template. Another resource for this purpose is the rotamer library.

The conformational energy of the side-chains is influenced by their 3D environment and

amino acids are known to prefer specific side-chain conformations. Their frequencies are

recorded in tables called rotamer libraries. The rotamers are chosen based on the protein

sequence and the protein backbone conformation using a defined energy function and

search strategy. Energy functions used for this purpose include simple steric energy

functions combined with the log probabilities of backbone-dependent rotamer

40

populations as well as molecular mechanics potential energy functions with complex

solvation free energy terms. 88

One of the widely used methods for modeling the side-chains is SCWRL.99 This method

initially places the side-chains on the backbone using the backbone dependent rotamer

probabilities and scans for steric interactions with the modeled main chain other than that

of the amino acid itself. In case of steric clashes with the main chain the next best

conformation from the rotamer library is chosen and so on. Side-chains that clash with

other side-chains according to the steric energy function are termed as “active residues”

and are clustered. The active residues are then subjected to an optimization procedure without involving the inactive residues as these already exist in their minimum energy

conformation. The optimization procedure in SCWRL and other side-chain modeling

procedures are based on a Dead End Elimination (DEE) strategy that iteratively

eliminates rotamers that are incompatible with the global energy minimum

conformation.100

Other than the traditional model building procedure mentioned above there are two

approaches that do not strictly follow this classic strategy. One is applied by the widely

used program “MODELLER” which builds the model on the basis of spatial

constraints.101 It computes a set of knowledge-based distance and dihedral angle

probability distributions for the final model and then builds the model to satisfy these

requirements. In another approach, rather than refining the model at each step, several

optimal and suboptimal models are funneled into the subsequent step and the final model

is selected on the basis of the analysis of several resulting atomic structures.

41

Once the model is built a refinement step is typically desirable. Regions where the SCRs and SVRs are connected usually suffer from high steric strains while several side-chains adopt conformations that result in bad van der Waals contacts. A local refinement procedure involving only the bad residues should be performed as optimization of all side-chains simultaneously would probably destroy the important H-bonds and the overall 3D structure. Energy minimization and /or molecular dynamics are important methods for exploring the local region of conformational space and may yield a better model. Common force fields for protein modeling include AMBER, GROMOS, CVFF and CHARMM.87 Energy minimization is a molecular mechanics based approach used only for removing the unfavorable contacts and fixing the H-bonds lengths in a model. It finds only the local minima on the potential energy surface but not necessarily the global energy minimum. It can be divided into first derivatives techniques such as steepest descent, conjugate gradient and Powell and second derivative techniques such as the

Newton-Raphson and related algorithms. While energy minimization is used to find the local minimum conformation, molecular dynamics simulation is used to detect the energetically most favored 3D structure of a protein. Molecular dynamics simulation samples the conformational space of a protein by integrating the classical equations of motions over a period of time. They collect snapshots of the simulation at regular time intervals (called trajectory) and explore large fractions of the protein conformational space while avoiding energetically unreasonable conformations. Despite this progress, refinement of the homology model is not easy and developing methods that can consistently improve over the starting model is still a major challenge as noted by the

CASP experiments.88

42

Validation of the homology model

Problems in comparative modeling commonly arise due to errors in fold assignment and

alignment (especially if sequence identity is less than 30%), distortions and shifts in the

core segments and loops, and errors in side-chain packing. Hence, after building and

refining the protein model, the next step is to assess the quality of the model. The quality

of the model can be defined in terms of its correctness and accuracy.102 The correctness

of the model is dependent on the quality of the sequence alignment and regions with

wrong sequence alignment will have corresponding wrong spatial arrangements. The

accuracy of the model depends on the deviation between the templates structure used and

the experimental structure of the protein (determined at a later time). Furthermore, error

in structure prediction is also influenced by limitations of the modeling technique as well

as the experimental procedures used for determining the structure. Methods for model

validation include evaluation of its stereochemical accuracy, the packing quality and fold

reliability.87

Stereochemical accuracy is determined by assessing parameters such as bond lengths,

bond angles, torsion angles and correctness of amino acid chirality. The Ramachandran

plot which defines the distribution of the main chain  (Phi) and  (Psi) torsion angles is

an important indicator of the quality of the model. Programs developed to evaluate the

model’s stereochemical quality include PROCHECK103, WHATCHECK104 and

VADAR105.

The interior packing of a globular protein is a major contributor of its stability and hence

the packing quality of the model can be used to estimate its accuracy. In one approach, a

model is checked for bad van Der Walls contacts by measuring all interatomic distances

43

and comparing to knowledge-based values. Programs such as PROCHECK and

WHATCHECK can be used for this purpose. Another approach evaluates the secondary

structural elements of the model since these are evolutionarily highly conserved regions

in homologues proteins. Programs applied to this end include DSSP106 and STRIDE107.

Fold reliability approaches are based on observations that proteins with homologues amino acid sequences have similar folds. Hence the overall 3D structure of the model and its template should be similar. Verify-3D is an algorithm that tests the compatibility between the 3D structure and the amino acid sequence by converting the 3D into a 1D representation known as a 3D profile.108 The 3D profile is basically an evaluation of how

well an amino acid fits in its environment e.g. hydrophobicity, area of side chain buried

in protein, secondary structure to which the amino acid belongs etc. Alternatively, the

program known as ProSA evaluates the fold reliability by using knowledge-based force

field methods that are based on compilation of potentials of mean forces from a data base

of known 3D protein structures.109

Modeling programs and servers

Over the years the number of 3D structure prediction programs has grown rapidly as evident from literature and also from the CASP experiment, wherein the number of participating groups has increased from 35 in 1994 (CASP1) to 234 in 2008 (CASP8). A few well known homology modeling programs include Modeller, ICM-HOMOLOGY110,

Discovery Studio111, SYBYL112 and WHAT IF113. Earlier building of a 3D structure of a

target protein would require in-house software but with advancement in computer

technology, a 3D model can now be generated free of charge by means of an automated

44

server and requires only the submission of target protein sequence by the user. Widely

used 3D structure prediction servers include MODWEB114, I-TASSER115, SWISS

MODEL116, EsyPred3D117 and PHYRE118. Recent years have also witnessed the advent

of Meta-servers. A meta-server utilizes other autonomous servers for generating 3D

models and employs automated consensus methods to deliver the best prediction. Well

known meta-servers include 3D-JURY119, 3D-SHOTGUN120 and ROBETTA121. The

performance of various servers and meta-serves are constantly evaluated by experiments

such as LiveBench, CAFASP (Critical Assessment of Fully Automated Structure

Prediction) and EVA.66

MODWEB: 101, 114

MODWEB is a widely used automated comparative modeling server that employs

MODPIPE122, a fully automated software pipeline for building the 3D model of the target

protein. The pipeline consists of two sections; the first determines the sequence-structure match between the target and the template and the second calculates the model.

Sequence-structure matches are established by aligning the PSI-BLAST profile of the

target sequence to the best templates obtained from the PDB as well as using the

complimentary approach IMPALA (Integrating Matrix Profiles and Local Alignments)

that scans the target sequence against a database of the template profiles.123 Significant

alignments covering different regions of the target sequence are used for the modeling.

The model is calculated using the comparative modeling program MODELLER which

builds the model by satisfaction of spatial restraints followed by optimization using a

combination of energy minimization and molecular dynamics. The spatial restraints

include (a) homology-based restrictions obtained from the alignment of the target

45

sequence with the template sequences, e.g., placing restraints on distances and dihedral

angles (b) stereochemical restraints such as bond lengths and bond angle preferences

obtained using the CHARMM-22 molecular mechanics forcefield (c) knowledge-based statistical preferences for dihedral angles and non-bonded interatomic distances and (d) optional manual curative restraints based on several factors including rules of secondary structure packing, knowledge from NMR and other spectroscopic techniques and general know-how of the modeling procedure.

The search for the best model in MODWEB can be tweaked using various user parameters including a more stringent assessment of useful sequence-structure relationships and controlling the conformational sampling space for a given sequence– structure alignment. The sequence–structure relationships are not prejudged at the fold- recognition stage but are validated only after construction and evaluation of the models which enable a thorough exploration of fold assignments, sequence–structure alignments and conformations, with the aim of finding the model with the best evaluation score.

MODWEB accepts the protein sequences in FASTA format for predicting the 3D structure of a target. Alternatively, MODWEB also accepts a protein structure as an input and calculates models for all its identifiable sequence homologs in the non redundant

SWISS-PROT protein sequence database.

Limitations of structure prediction methods

1] The quality of the model depends on the quality of the template used and the model

cannot be better than the template determined by experimental methods. As the model is

an extrapolation of the template, problems in the template will be carried on to the model.

46

Problems in the template can arise from various factors including high mobility regions

of the protein, improper connection between the secondary structural elements and data

lost during refinement of the crystal structure.124, 125 Furthermore, insertions and deletions

in the amino acid sequence will further affect the 3D structure. 2] The model will vary

according to the structure prediction program used for its construction and even the top

two programs in the field will not generate identical models. Thus, though the models

constructed by these programs may be the best there still remains the question as to

which model would be the better one. 3] The model obtained may not represent the native

conformation of the protein but rather a native-like conformation. Comparing the model

to the experimental structure (if available) would be ideal, but again, the experimental

structure itself is influenced by multiple factors including solvents, temperature, presence

of co-factors and structural elements (e.g. tagged proteins) added to improve the stability

of the crystals. 4] H2O molecules are an important components of the protein structure

that may be left out while generating the homology model and affect the quality of the

final model (e.g. during energy minimization). Additionally, water molecules in the

protein active site play important roles in the binding of substrates and inhibitors which

may affect the applicability of the model e.g. in drug discovery.

Homology modeling in drug design and development

Drug discovery is a difficult, time-consuming and expensive process that has warranted

the development of high through-put screening (HTS) and structure-based approaches for improving its effectiveness. HTS is a random approach that involves testing of vast compound libraries against the pharmaceutical target in a relative short time. However

47

this approach is burdened by factors such as cost and limitations in terms of exploring the chemical space. Structure-based drug design (SBDD) on the other hand is a rational approach that uses the 3D structure of the target for selection and optimization of lead candidates and is more cost effective with higher success rates than HTS. Nevertheless, a major limitation of this approach is the availability of the target structure. In the year

2000, a structural genomics project was started that aims at experimentally determining at least one 3D structure for every protein family and approximately 3000 structures have been solved so far.126 Information from these representative structures can be employed to predict protein structures of pharmaceutical interests and homology modeling is expected to play a crucial role in structure-based drug design. Homology modeling can be helpful at various stages of the drug discovery process as described below.127

Assessment of target: An important step in drug discovery is identification and evaluation of the target for pharmaceutical intervention. Of the several thousand proteins (>20K) found in the human proteome, only 600-1500 are estimated to be of pharmaceutical importance based on their involvement in disease states and their ability to interact with drug-like molecules.128 Structure based studies can aid the prioritization of protein targets by evaluating various factors that influence the druggability of the target, e.g., by identifying the ligand binding sites in the target and its feasibility towards development of modulating agents, assessing possible side effects due to active site similarity with other proteins, the pharmacogenetic and mutation profile of the target, suitability of the target for animal model studies, presence and potential use of allosteric binding sites, etc.

Additionally, structural informatics can also be used at later stages such as addressing the issue of drug resistance, e.g., the drug resistance of certain patient populations against the

48

chronic myelogenous leukemia (CML) drug Gleevec (Novartis) was explained using

structural information that pinpointed a specific mutation within the Abl kinase binding

domain.129

Lead discovery: Homology models are routinely used for docking studies that enable the

virtual screening of a large library of compounds against the target binding site. Several

programs have been developed for this purpose and evaluate the docking on various

parameters such as van Der Waals interactions, hydrophobicity, electrostatic potential

etc. Discovery of lead molecules using homology models has been well documented in

the literature, e.g., a homology model of cyclin-depenedent kinase 4 (CDK4) based on the X-ray structure of CDK2 was used in the discovery of the diarylurea class of

compounds as potent and selective CDK4 inhibitors130 and a homology model of DNA methyltransferase 1 (DNMT1) based on multiple template proteins led to the design of

N4-fluoroacetyl-5-azacytidine as highly potent DNMT1 inhibitors.131

Lead optimization: Following the discovery of a lead molecule, structure-based methods

can be used to optimize parameters such as potency, selectivity, ADME profile, safety-

toxicity and resistance issues. Homology models of important cytochrome (CYP) p450

enzymes have been used to explain and predict the probable sites of metabolic attack in a

variety of CYP substrates that may facilitate the optimization of a lead with respect to

ADME parameters. Several drugs developed using structure-based approaches are on the

market or in different stages of clinical trials, e.g., captopril (ACE inhibitor, hypertension,

BMS), aluvarin (HIV protease inhibitor, anti-HIV, Abbott) and doramapimod (p38 MAP

kinase inhibitor, Phase II/rheumatoid arthritis, Boehringer Ingelheim). 132

49

Prediction of animal model suitability: It is often observed that results from human in

vitro studies cannot be demonstrated in animal models. The binding sites of the target

protein in humans and animals may vary just enough to give completely different results.

Homology models of the target protein from different species can be used for evaluation

and hence for the selection of the correct animal model e.g. comparing the binding sites

of human estrogen receptor  (hER) to the homology model of the rat, murine and

bovine receptor showed an important amino acid difference in the bovine receptor that

resulted in lower activity for certain ER ligands in bovine models.133

In addition to the above, homology models are used for understanding the binding

characteristics of ligands, designing in vitro test assays and for predicting toxicity and

drug-drug interactions. Implementation of structural aspects during and after drug

development is becoming a routine practice in the pharmaceutical industry. Generating

homology models is relatively easy and quick and can considerably reduce attrition rates

in the drug discovery process.

Results and discussion

Constructing the p110 homology model

As part of the first specific aim, a p110 homology model was built to evaluate the binding mode of the proposed scaffold in p110 and to analyze structural differences between the p110 and  binding sites that complemented the structure-activity relationship (SAR) study of the scaffold. The X-ray structures of PI3K obtained in

complex with various inhibitors are available in the PDB and served as a guide for

generating our p110 model. The PI3K and  isoforms contain a relatively conserved

50

catalytic subunit (p110) and different regulatory subunits. The amino acid sequences of

the two catalytic units were obtained from Swiss-prot and consisted of 1068 residues for p110 (Acc. number P42336, pg. 58) and 1102 residues for p110Acc. number

P48736). A BLASTP search shows 35% sequence identity and 53% sequence similarity

for the two isoforms.

Generating the p110 model. The p110 amino acid sequence was uploaded onto the

Mobweb server in FASTA format. The model was generated with Modeller 6v2 using the high resolution X-ray structure (2Å) of p110-wortmannin (1e7u) as the template.

Output from the run (pg. 58) shows a model score of 1 which indicates a good model

(>0.7). The model was evaluated and corrected for Asn/Gln/His flips with

MolProbity.134

Model refinement by molecular dynamics simulation. The model was refined with

YASARA dynamics.135 A simulation cell was built around the model with a 10Å cutoff

for the electrostatic forces, which were calculated using the Ewald method. The pKa

values of the ionizable groups in the protein were predicted and assigned protonation

states based on pH 7.4. The cell was filled with water and the AMBER99 electrostatic

potential was evaluated at all water molecules; the one with the lowest or highest

potential was turned into sodium and chloride counterion until the cell was neutral. A

short steepest descent minimization of all atoms removed severe bumps followed by

simulated annealing minimizations at 298K. Velocities were scaled down every 10 steps

for a total time of 5 ps in 500 steps. A start-up simulation was then run for 5 ps, using a

multiple time step of 1 fs for intramolecular and 2 fs for intermolecular forces, with all

back-bone protein atoms fixed. Simulated annealing minimizations were started at 298 K,

51

and velocities were scaled down every 10 steps for a total time of 5 ps in 500 steps.

Molecular dynamics simulations were run with the AMBER99 force field at 298 K and

0.9% NaCl in the simulation cell for 500 ps to refine the model.

Evaluating the model. The quality of the model was evaluated by means of its

stereochemical accuracy. The Ramachandran plot was generated with PROCHECK

(Fig. 22, pg. 59).136 The plot shows 84% of the residues in the CORE, 13 % in the

ALLOWED, 1.7% in GENEROUS regions while only 1.3% (i.e.11 residues) in the

disallowed region. Of the 11 residues only 2 are present in the catalytic domain – Lys724 which is far away from the binding site and Arg852 which is located at the entrance of the binding pocket (border of 5Å radius) with the side-chain facing the solvent. Having

84% of the residues in the CORE region is equivalent to an X-ray with ~2.5Å resolution.73 Overall the quality of the model was reasonable for our study.

The p110 model: The model contains four domains – the Ras binding domain (RBD, residues 190-291), the C2 domain (330-524), helical domain (525-696) and the catalytic domain (697-1062) (Fig. 15A). The model lacks the adaptor binding domain (ABD, residues 1-106) akin to the template p110 X-ray (residues 1-143) as PI3K is stable and

active even in the absence of an adaptor protein. Aligning the model with 1e7u using

PyMol137 shows rmsd of 1.67 Å (Fig. 15B).

Comparison of the p110 model vs. X-ray structure (2rd0): In December 2007, the first

X-ray structure of PI3K with a 3Å resolution was published. The protein contained the catalytic p110 subunit and a part of the p85 regulatory subunit. Aligning the model to the 2rd0 structure shows rmsd of 2.92 while deleting sections of 2rd0 that were not present in the model i.e. p85 subunit and ABD, improves the rmsd to 2.4 (Fig. 15C).

52

When a 5Å radius around LY294002 docked in the ATP binding site of the p110 model

(pg. 54) was compared to the corresponding region in the X-ray structure, an rmsd of

0.942 was obtained. Therefore, the homology model was in excellent agreement with the

X-ray structure and corroborates the reliability and utility of the model for our study.

A B

C

Figure 15: A. PI3K 3D model. The four domains are colored as follows: RBD - blue, C2 - brown, helical – aquamarine and catalytic - red. B. Overlap of the p110 model (yellow) with the p110 X-ray (1e7u, blue). Wortmannin is shown as spheres, carbon atoms in cyan and oxygen in red. C. Overlap of p110 model (yellow) with the X-ray structure 2rd0 (magenta).

Designing the scaffold

In order to explore the two binding sites we developed a scaffold that provided easy

assembly and rapid generation of chemical libraries. Our scaffold was based on the

53

morpholinylchromone LY294002 which is a widely used non-selective PI3K inhibitor for studying PI3K related cellular mechanisms. Its X-ray structure with p110 shows H- bonds of C=O with Lys833, the morpholine oxygen with backbone amide of Val882 and phenyl at the entrance of the binding pocket facing the solvent. This molecule is compact

and has only two rotatable bonds. Though the rigid structure of LY294002 permits

defining the binding mode it does not provide enough conformational freedom for exploring the subtle differences between the two binding sites. Analyzing the literature on morpholine containing PI3K inhibitors suggest that the chromone core functions as a pharmacological spacer and its replacement is permissible provided the H-bond requirements are satisfied. 47, 48, 51, 56, 138, 139 Moreover, the phenyl group can be further

H2N Lys833 H O OH O O C H C 3 N

O N N N N N O HN O O V882 X H N O

LY294002 TGX-126 2,6-disubstitued isonicotinc scaffold Figure 16: Proposed scaffold for exploring the p110 and p110 binding sites.

away from the core and have increased flexibility as seen in structural analogs such as

TGX126.48 In accordance with these observations, we selected the 2, 6-disubstituted isonicotinic acid for our study (Fig. 16). Next we evaluated the scaffold with molecular modeling. Docking studies were performed with CAChe v6 & v7 (Fujitsu).140

Validation of Docking: The method was validated by docking LY294002 in the 1e7v

(LY294002-p110X-ray structure. LY294002 was built separately in CAChe workspace and minimized using augmented MM3 force field (labeled LY2, grey). A 5Å radius around the LY294002 (yellow) of 1e7v X-ray structure was set up as the binding site.

54

1 Next LY2 was place d in the binding site by

means of atoms labeled 1, 2, 3 & 4 as points

for superimposition. LY2 was docked using 3 4 2 the potential of mean force (PMF) scoring

function with flexible ligand-flexible protein

option. PMF is a knowledge-based approach

that exploits structural information of known Figure 17: Overlap of built LY294002 (grey) with the X-ray conformation (yellow). protein-ligand complexes extracted from the

Brookhaven Protein Data Bank and converts it into distance-dependent Helmholtz free interaction energies of protein-ligand atom pairs. Settings used for the docking - population size 50, crossover rate 0.8, elitism 7, maximum generation 3000, Mutation rate 0.3, convergence 1.0. The docked pose of LY2 matched the X-ray structure with

RMSD 0.095 (Fig. 17).

Docking the LY294002 in the p110 homology model: The binding site of the p110 homology model was generated by aligning the sequences of the two isoforms and matching the residues along the 5Å binding site radius of the p110. In the absence of a ligand for superimposition, LY294002 was externally docked into the p110 binding site.

The docking settings were same as before. Docking of LY294002 in our p110model revealed binding mode similar to p110 as expected owing to the similarity of the two binding sites (Fig. 18A, compare with Fig. 12C). 

Docking the 2, 6-disubstituted isonicotinic acid derivative: A derivative based on the proposed scaffold was generated with the 4-F-2-MePhNH2 at the 6-position based on reported literature.48 This derivative was later synthesized and is denoted as compound 28

55

A B Val851(882) Val851(882) Lys802(833) Lys802(833)

Lys776(808)

Lys776(808) Trp780(812) Trp780(812)

Figure 18: Docking pose of LY294002 (A) and isonicotinic acid derivative, compound 28 (B) in the p110 binding site viewed from the same angle. Numbering in parentheses is that of p110. in chapter III. The generated compound 28 was minimized using the augmented MM3

forcefield and placed in the p110 binding site using the 4 points of superimposition as mentioned previously; the carbon of the COOH was superimposed onto the ketone (atom

1, Fig. 17). The docking settings were same as before. Docking showed the carboxylic acid of 28 interacting with the side chain amine of Lys802 possibly through a salt-bridge and the morpholine oxygen forming an H-bond with the backbone amide of Val851. These groups are located in the interior of the binding pocket while the

4-F-2-CH3Ph is at the entrance facing the solvent

(Fig. 18B). The docked pose is comparable to that obtained for LY294002 (Fig. 19). Compound 28 Figure 19: Comparing binding showed a counterclockwise shift of approx. 1.5Å with modes of LY294002 and 28.

respect to the H-bond atoms of LY294002. The phenyl groups of both molecules are

along the same plane but the 4-F-2CH3Ph group of 28 extends further outside the pocket

than the phenyl of LY294002.

Salient features of the scaffold: The proposed scaffold can be readily synthesized from

2-chloro-6-methyl isonicotinic acid. Bromination of the methyl differentiates the

56

Figure 20: Simple synthetic approaches allow preparation of three libraries. reactivity at the 2 and 6-positions towards nucleophilic attack thus allowing sequential attachment of various chemical groups at these positions (Fig. 20). The carbonyl of

LY294002 and TGX-126 occupies the phosphate binding area of ATP but does not

extend as far as the triphosphate group. The COOH in our scaffold extends one carbon

further from the aromatic core and can be easily modified for probing this region.

Substitution of the chlorine will allow assessment of the area surrounding the

morpholine. Docking studies place the phenyl group at the 6-position towards the

entrance of the binding pocket. This region of the enzyme is quite variable between the

two isoforms and provides opportunity for differentiation.141 The presence of three

A B C

Figure 21: A. Three rotatable bonds present between the phenyl and pyridine rings. B. Depiction of the area that can be scanned by the phenyl group of 28 due to the flexibility of the bonds. This conformational search was performed in vacuum using the MM3 forcefield, 300 steps, 25 steps/rotatable bond (number of steps restricted for clarity purposes). C. Minimum conformations with the three rotatable bonds were generated using CAChe (conflex/MM3). Blue part of the curve represents favorable region and shows ~45 conformers within 1Kcal/mol of the lowest energy conformation indicated by the black dot.

57

rotatable bonds makes this portion of the scaffold highly flexible. A conformational search with Macromodel 9.6 (Schrödinger Suite 2008)142 shows a wide area that can be

explored by this group (Fig. 21). As such this group might involve potential hydrophobic

interactions within the enzyme and contribute towards the activity and/or selectivity.

Conclusions:

Advancement in the field of protein structure prediction has greatly facilitated the process

of generating reliable 3D models of target proteins. At the same time, a structure based

approach has become an integral part of drug development with tremendous application

at every stage – from discovery to post market analysis (e.g. understanding resistance,

drug-drug interactions). In this study we successfully generated a reliable p110

homology model and used it in designing the novel 2,6-disubstituted isonicotinic scaffold

for exploring the PI3K and  binding sites.

58

Protein sequence of p110. (Swiss-prot Accession # P42336)

1 mpprpssgel wgihlmppri lvecllpngm ivtleclrea tlitikhelf kearkyplhq 61 llqdessyif vsvtqeaere effdetrrlc dlrlfqpflk viepvgnree kilnreigfa 121 igmpvcefdm vkdpevqdfr rnilnvckea vdlrdlnsph sramyvyppn vesspelpkh 181 iynkldkgqi ivviwvivsp nndkqkytlk inhdcvpeqv iaeairkktr smllsseqlk 241 lcvleyqgky ilkvcgcdey flekyplsqy kyirscimlg rmpnlmlmak eslysqlpmd 301 cftmpsysrr istatpymng etstkslwvi nsalrikilc atyvnvnird idkiyvrtgi 361 yhggeplcdn vntqrvpcsn prwnewlnyd iyipdlpraa rlclsicsvk grkgakeehc 421 plawgninlf dytdtlvsgk malnlwpvph gledllnpig vtgsnpnket pclelefdwf 481 ssvvkfpdms vieehanwsv sreagfsysh aglsnrlard nelrendkeq lkaistrdpl 541 seiteqekdf lwshrhycvt ipeilpklll svkwnsrdev aqmyclvkdw ppikpeqame 601 lldcnypdpm vrgfavrcle kyltddklsq yliqlvqvlk yeqyldnllv rfllkkaltn 661 qrighfffwh lksemhnktv sqrfgllles ycracgmylk hlnrqveame klinltdilk 721 qekkdetqkv qmkflveqmr rpdfmdalqg flsplnpahq lgnlrleecr imssakrplw 781 lnwenpdims ellfqnneii fkngddlrqd mltlqiirim eniwqnqgld lrmlpygcls 841 igdcvgliev vrnshtimqi qckgglkgal qfnshtlhqw lkdknkgeiy daaidlftrs 901 cagycvatfi lgigdrhnsn imvkddgqlf hidfghfldh kkkkfgykre rvpfvltqdf 961 liviskgaqe ctktreferf qemcykayla irqhanlfin lfsmmlgsgm pelqsfddia 1021 yirktlaldk teqealeyfm kqmndahhgg wttkmdwifh tikqhaln

Modweb output

HEADER ModWeb Comparative Modeling Server: modweb TARGET pi3ka_3 TEMPLT 1e7uA REMARK 1 Length of target sequence : 1068 REMARK 1 Region of target sequence modeled : 107 to 1062 REMARK 1 REMARK 2 Template Structure used : 1e7uA REMARK 2 Length of template structure : 872 REMARK 2 Region of template structure used for modeling : 144 to 1091 REMARK 2 REMARK 3 E-value of the alignment : 0 REMARK 3 Sequence identity of the alignment : 37 REMARK 3 Quality of the model (model score) : 1.00

59

Figure 22: Ramachandran plot for the p110 homology model 60

Limitations of our modeling and suggestions for improvement

Molecular modeling served as an important tool for this project that allowed for the generation of the p110 homology model, designing of the 2,6-disubstituted isonicotinic scaffold, selection of the derivatives and evaluating the binding modes of the derivatives.

Nevertheless, during the course of the project, we realized some of the limitations of our techniques and software and these are discussed in this section.

With regards to homology modeling, the p110 model was generated using MODWEB which is a widely used protein modeling server and provides reliable models as seen in this chapter. However, the model might not be the best that could be generated and for this reason, use of a metaserver is recommended for generation of homology models.

Metaservers generate models from several autonomous servers and critical evaluation of these models may provide a more robust model. CASP, CAFASP, EVA and LiveBench are useful resources for determining the best programs and servers currently available for homology modeling. With regards to p110, since the X-ray structure has been published it is recommended to use this structure for future studies.

During the study, three regions of the scaffold were modified as part of an SAR study which is discussed in chapter IV. These modifications were based on the docking of a key compound 28 in the p110 model (Chapter IV). In parallel to the SAR study, we also performed docking studies in order to evaluate the binding mode of these derivatives in the p110 model and understand their impact on the potency. The docking studies were performed with CAChe.

While working with CAChe, it was observed that docking the ligand without first placing it in the active site (“external” docking) would often lead to docking of the ligand outside

61

the binding site as it was more energetically favorable. However by superimposing with a reference (e.g. LY294002) already present in the binding site, the frequency of the ligands being docked outside the binding site was reduced considerably.

A common practice during docking is to keep the active site rigid and the ligand flexible.

Even though this approach is fast (~5m/run using an AMD Athlon 64 processor, 2.6 GHz with 1 GB memory), CAChe would often dislodge the ligand out of the active site during the run despite of its superimposition in the site. To overcome this, both the ligand and the active site were made flexible. This approach was advantageous with regards to the retention of the ligand in the binding site but it considerably increased the run time

(~3h/run) for the docking. As an alternative, a two step approach was used; first after superimposition of the ligand in the active site, the active site was relaxed keeping the ligand rigid to remove the initial steric clashes between the ligand and the active site; next the ligand was docked as a flexible unit keeping the relaxed active site rigid. This technique reduced the frequency of the ligand being dislodged from the active site and also provided moderate advantage with regards to the run time. The latter approach was used to evaluate the binding mode of the derivatives during the SAR study.

Reproducing the ligand binding pose for different runs using the same parameters was difficult. This is because CAChe uses a genetic algorithm (GA) for docking the ligands in the active site. GA is a random search algorithm which offers the advantage of relatively unbiased searching. GAs start by creating random populations of ligand related parameters such as random positions, orientations and conformations of the ligand and then optimizes these to find the best solution. Accordingly, obtaining the same pose for different runs would be difficult since the starting parameters for the docking in these

62

runs may be different. In such cases, the most frequently occurring and/or meaningful

pose from the different runs was selected.

Following the docking, a docking score is obtained which defines how well the ligand fits

in the binding site. We tried to evaluate whether the docking score can be used as a guide

to predict the relative potency of these compounds on the p110. The potency of these

compounds is defined in terms of their IC50 values on the p110 which was determined

using an isolated enzyme kinase assay (see chapters III and IV). As an example, the

docking score and the IC50 values of four compounds that were synthesized while

modifying the 2-position of the scaffold is shown in Table 1. According to the docking

score, one would expect that 47 containing the H-bond donor/acceptor piperazine should

be slightly more potent or at least comparable to the H-bond acceptor morpholine 28.

However, the IC50 values of these two compounds suggest that the H-bond

donor/acceptor leads to a significant loss of potency. Similarly, based on the docking, the

H-bond acceptor at the 3-position of aromatic substituent should have better potency than

at the 4-position (52 vs. 51). However, this is in contrast to the IC50 values and the H-

bond acceptor at the 3-position lead to a significant loss of potency. Moreover, the docking score also suggests that the potency of these compounds should improve as we

progress from the morpholine to the 3-pyridine which does not correlate to the observed

IC50. Hence the docking score obtained from CAChe was not predictive of the relative

potencies of these compounds on the p110.

Thus, while CAChe successfully predicted reasonable binding modes for LY294002 and

the proposed 2,6-disubstituted isonicotinic scaffold in the p110 as discussed in this

63

chapter, it was not successful in defining the binding modes of the SAR modifications

and correlating their impact on the potency.

One approach for corroborating the docking studies might be to use more than one

docking program. In addition to the low energy conformations, commonly occurring

ligand binding poses in the different programs should be considered as equally important.

Recommended docking software is AUTODOCK which uses a modified GA for docking

of ligands and has shown to be very successful.

Table 1 : Docking scores and IC50 of select compounds on the p110

COOH R1

CH3 H N N R2 R3 N R2 F

Docking score on IC (M) on No. R2 50 p110, (Kcal/mol)a p110b

N O 28 -97.952 5 ( 1) Morpholine

N NH 47 -97.969 >100 Piperazine

N 51 -101.973 29 ( 7) 4-pyridine N 52 -103.471 >100 3-pyridine a Compounds were docked using same procedure as for 63 ( see experimental section) . The binding site was set at 5Å around LY294002 b obtained from in vitro isolated enzyme kinase assay (chapter III and IV)

64

III: Evaluating the 2,6-disubstituted isonicotinic

scaffold

Introduction

The purpose of this specific aim was to evaluate the synthetic feasibility and biological

activity of the proposed 2,6-disubstituted isonicotinic scaffold for our study. For this, six

derivatives utilizing this scaffold were designed, synthesized and tested in an in vitro

kinase assay and two cell assays. As our scaffold was based on the arylmorpholine

LY294002, the selection of these derivatives was shaped by the reported structure-

activity relationship (SAR) of the Eli Lily (LY) and the Thrombogenix (TGX) series

which is briefly discussed below.47, 48

Most derivatives from the LY series and all from the TGX series contained the

morpholine at the 2-position while both contained the 4-ketone. These groups form

important H-bonds with the enzyme as exemplified by the LY294002-p110 X-ray

A B O 5 4 6 3

R 2 Ile963 Val882 7 O N 8 1 O

No. Name R IC50, M Ile881 20 LY292223 H 5 21 LY293684 Ph, 5-6 fused 14 Trp812 22 LY293646 Ph, 7-8 fused 6 Lys833 Ile831 6 LY294002 8 - Ph 1.4 Met804

Figure 23. (A) SAR of 2-Morpholinylchromone shows improvement in activity by substitution at the 8- position (B) X-ray of LY294002 in p110 shows important H-bonds and the 8-phenyl occupying a hydrophobic patch (dotted line) at the entrance of the binding pocket.

65

structure (Fig. 23). In the LY series, the presence of a fused 5-6 phenyl ring on the 2-

morpholinylchromone (21 vs. 20) caused a ~3-fold decrease in potency while the fused 7-

8 phenyl compound (22 vs. 20) was equipotent. The presence of phenyl at the 8-position i.e. 6 led to a slight increase in potency (Fig. 23A). This improvement can be attributed to interaction of the phenyl group with a hydrophobic patch at the entrance of the binding pocket consisting of Trp812, Ile881 and Met804 (Fig. 23B). Our docking studies show a similar orientation for LY294002 in p110 with the Ph group at the entrance (Fig. 18B).

In the TGX series, a wide range of aromatic groups was explored at the 8-position. The

compounds were tested at the PI3Kisoforms and a limited number were evaluated at the PI3K isoform as well. The 8-position aromatic substituents in the TGX series were not fused but were separated from the fused rings by at least one atom (O, NH or CH2).

O R Y

X N Z O IC50, M No. Name X Y RZPI3K PI3K

23 TGX-155 NH C H O F 0.02 5

H3C

HN F 0.2 ND 24 TGX-170 N NCH3

H3C

F H2CHN 0.01 ND 25 TGX-186 N NCH3

H3C 6 LY294002 1 - 1.5 2

Figure 24. Derivatives containing the 4-F-2-CH3PhXH group at the 8-position were the best compounds from the TGX series.

66

As compared to LY294002, these ligands were more flexible and covered more conformational space. Results indicate that a wide range of aromatic substituents were tolerated at the 8-position which can be rationalized by this region being the entrance of the pocket and hence solvent exposed and therefore not undergoing any major binding interactions. Several compounds from this series proved to be more potent than

LY294002. Among these, the quinazolone derivative TGX155 (23) containing the 4-F-2-

CH3PhOH group at the 8-position was the most potent with ~250-fold selectivity for the

PI3K isoforms (Fig. 24). A close analog TGX170 (24) was 10 times less potent at the

 isoform while TGX186 (25) with increased flexibility and conformational coverage showed slight improvement in activity when compared to TGX155 (Data for the PI3K

was not available). All these derivatives Interior contained the 4-F-2-MePhX substituent

at the 8-position. One feature of this

Arg852/Lys883 particular aromatic substituent that

might contribute towards the p110 Arg770/Lys802 Exterior selectivity is the 4-F atom. Though the

fluorine is weak H-bond acceptor it is Gln859/Lys890 His855/Thr886 shown to undergo multipolar C–F···H–

Figure 25. Overlap of PI3K (1e7v, magenta) with N, C–F···C=O and C–F···H–C the PI3K (2rd0, yellow) reveals important differences at the entrance of the binding pocket that interactions with peptide bonds in might influence the potency of 4-F-2-CH3-PhXH group in the TGX derivatives. proteins (particularly those in hydrophobic environments) as well as the side-chain amide residues of Asn and Gln and the positively charged guanidinium side chain of Arg.143 Overlap of the PI3K and  X-

67

ray structures suggest that the residues at the entrance of the binding pocket of PI3K provide a more fluorophilic environment (Arg 770, 852, Gln858 and His855) when compared to PI3K(Fig. 25) hence improving the PI3K potency and selectivity.

Additionally, the 2-CH3 on the phenyl can also contribute by enhancing the hydrophobic interactions.

Results and discussion

In accordance with the H-bonding 2-morpholine and 4-ketone in the above arylmorpholines, we chose the morpholine at the 2-position and the COOH at the 4- position of our scaffold. As several of the potent compounds contained the 4-F-2-

O OR CH3PhXH (where X = O or NH) at the 8-

position, we selected this substituent at

the analogues 6-position in our scaffold X N N Y O (Fig. 26). In addition to the COOH we

No. R X Y also evaluated their corresponding non-

26 H CH ionizable ethyl esters to study the 2 O F importance of the charged carboxylic acid 27 OEt " H3C (pH 7.4) and these were expected to 28 H " HN F provide better permeability properties for

29 OEt " H3C the cell assay. The esters did not require

30 H H Nil additional synthetic efforts as these were

31 OEt " " precursors for the corresponding

Figure 26. Derivatives based on the 2,6- carboxylic acids. The ethyl ester was disubstituted isonicotinic acid scaffold that were synthesized and tested.

68

chosen as it is more stable than a methyl ester and at the same time, less hydrophobic and less bulky than the propyl and other esters. At the 6-position the ether oxygen of 26 and

27 and the NH2 of 28 and 29 were selected to analyze the H-bond acceptor vs donor/acceptor preference in this region. Since absence of the 8-position aromatic substituent in the chromone scaffold (LY series) caused a decrease in PI3K potency, derivatives 30 and 31 were evaluated to study the contribution of the 6-substituent towards the potency of our isonicotinic scaffold.

Chemistry

For synthesis of 26-29, the scheme in Fig. 27 was proposed which was later modified as discussed below. The synthesis started from 2-chloro-6-methyl isonicotinic acid. The first step was esterification of the carboxylic acid followed by attachment of the morpholine via addition-elimination. Next the 6-CH3 was to be brominated using N- bromosuccinamide/ azobisisobutyronitrile (NBS/AIBN) which was to be then displaced by the 4-F-2-CH3PhXH (X = O/NH) thus providing the desired ethyl esters 27 and 29.

COOH COOEt COOEt HN O H2SO4/EtOH

Reflux EtOH/Reflux H C N N H3C N Cl H3C N Cl 3 COOEt 2-Chloro-6-methyl- O isonicotinic acid NBS Br

AIBN, CCl4

reflux H3C N N O

COOH COOEt COOEt

4-F-2-MePhXH HOH N N N N N N K2CO3/DMF, rt X O X O Br O X = O or NH

H3C F H3C F Figure 27. Proposed synthetic scheme. Bracket indicates reactions that could not be achieved due to the bromine substitution on the pyridine ring.

69

Finally hydrolysis would yield the carboxylic acids 26 and 28. This scheme would have been convenient as the 2-morpholine intermediate could be synthesized in bulk and then varying substituents could be attached at the 6-position. However, following attachment of morpholine it was observed that bromination took place on the ring rather than at the

6-CH3 as determined by NMR and mass spectrometry of the product. This effect is due to the o, p-directing electron donating morpholine nitrogen which caused substitution of the bromine either at the 3- or 5-position on the pyridine ring. Hence, attachment of the morpholine before brominating the 6-CH3 was not reasonable. To overcome this problem, we brominated the 6-CH3 before adding the morpholine and the scheme was modified as shown in Fig. 28. Derivatives 26-29 were synthesized using this scheme.

Derivatives 30, 31 were synthesized from 2-chloroisonicotinic acid as shown in Fig. 29.

COOH COOEt COOEt

H2SO4/EtOH NBS/AIBN, CCl4 reflux reflux

H3C N Cl H3C N Cl N Cl 2-chloro-6-methyl Br isonicotinic acid 32 33

4-F-2-MePhXH

K2CO3/DMF, rt

COOH COOEt COOEt

HN O H2O N N N N Microwave, N Cl 250°C, 90m X O X O X

H3C F H3C F H3C F 26, X = O 27, X = O 34, X = O 28, X = NH 29, X = NH 35, X = NH

Figure 28. Rearrangement of the synthetic steps led to the modified synthetic scheme for synthesis of 26-29.

70

O O N COOH COOEt COOH Microwave 200°C, 60m H2SO4/EtOH H2O

N Cl HN O N N N N N N O O O 2-Chloro- 36 31 30 isonicotinic acid

Figure 29. Synthesis of 30, 31.

Optimization of reactions:

The esterification, hydrolysis and nucleophilic displacement of the bromine by 4-F-2-

CH3PhXH were relatively straight forward; however, problems were encountered during the bromination and attachment of morpholine and had to be optimized.

1] Bromination

The bromide 33 is a key intermediate for the synthesis. Bromination of the 6-CH3 with

NBS/AIBN yielded mixtures of monobromide (40%), dibromide (~55%) and minor amounts (<2%) of tribromide as judged by HPLC and LC-MS analysis. Purification of this mixture by column chromatography provided ~37% 33 (Fig. 30). Fortunately, treatment of this crude mixture with 4eq. diethyl phosphite – diisopropylethylamine before purification selectively reduced the polybromides to the monobromide 33 and improved its yield to ~75%.144

COOEt COOEt COOEt column chromatography, 37% NBS/AIBN

CCl4 HOP(OC H ) /DIEA 2 5 2 N Cl H3C N Cl N Cl column chromatography, 75% Br Br 33 n = 1-3 Additional step improved yields Figure 30. (A) Treatment of the polybromide mixture with diethylphosphite/ diisopropylethylamine led to ~100% increase in yield of monobromide 33.

71

2] Addition of morpholine

Addition of the morpholine by refluxing in EtOH required long reaction times (~5d) for completion as judged by HPLC and LC-MS analysis. The major product of the reaction was the ester and small amounts of the carboxamide 37, 38 were also observed that had formed due to coupling between the morpholine and the ethyl ester. To circumvent the long reaction time microwave heating was used. However, under these conditions, the morpholine reacted with the ester much faster leading to formation of carboxamide first which upon further heating (total 1 – 1.5h) yielded 37, 38 as the major product (Fig. 31).

In order to avoid chromatographic separation, the carboxamides were transesterified back to the ethyl ester by refluxing in H2SO4/EtOH for 48h and the esters were obtained in high yields (85-90%).

COOEt CON O COOEt HN O

Microwave N Cl N Cl 250°C N N X X X O COOEt

H C F H3C F 3 H3C F EtOH Main product N N HN O + + H2SO4 X O Microwave O COOEt 250°C CON H3C F 27, X = O 29, X = NH N N N N X O X O

H3C F > 5eq HN O H3C F Main product Microwave 37, X = O 38, X = NH

Figure 31. Nucleophilic displacement of the chlorine by morpholine was accompanied with the formation of the amides 37 and 38 which were transesterified back to the ethyl esters.

72

Biological evaluation

Lipid Kinase Assay

The IC50 values of the compounds on PI3K  and  were determined with the Promega

Luminescent Kinase Glo® assay that measures kinase activity by quantifying the amount of ATP remaining in solution following a kinase reaction. 145 The unused ATP reacts with the Ultra-GloTM recombinant Luciferase enzyme in the presence of Mg2+, molecular oxygen and beetle luciferin and produces oxyluciferin and one photon of light per turnover that is measured via luminescence (Fig. 32). The signal produced is proportional to the amount of ATP and inversely proportional to the amount of kinase activity. Thus, an inhibitor of the PI3K will lead to a higher signal as higher amounts of ATP will remain in solution, whereas if the enzyme is not inhibited the ATP will be utilized for the kinase reaction and a lower signal is obtained.

Assay Principle

-2 OH OPO3 Kinase Substrate + ATP Product + ADP

COO- O- HO S N HO S N Luciferase ATP 1/2O AMP CO + + 2 2+ ++2 N S Mg N S + PPi Beetle Luciferin Oxyluciferin

+ Light

Figure 32. The Ultra-GloTM Recombinant Luciferase reaction.

The sigmoidal dose-response curves for the compounds are shown in Fig. 33. Solubility issues complicated the evaluation of esters 27 and 29. From Table 2, the 2,6-disubstituted isonicotinic derivatives 26 and 28 inhibited the PI3K and  enzymes in the low

73

100 100 6 (LY294002) 26 (PC477) 80 80 28 (PC481) 30 (PC386) 60 60 31 (PC483) 40 40 Inhibition (%) Inhibition Inhibition(%)   20 20 PI3K PI3K 0 0 4 3 2 1 0 -1 -2 -3 -4 -5 4 3 2 1 0 -1 -2 -3 -4 -5

Log10[Inhibitor]M Log10[Inhibitor]M Figure 33. Inhibition curves are the average of three experiments in triplicates  SEM determined with the Promega Luminescent Kinase Glo ™ assay and processed with GraphPad Prism 5.0 using the sigmoidal dose response curve (variable slope). micromolar range and showed ~3- and ~7-fold selectivity for PI3K respectively.

Compound 26 containing the H-bond acceptor ether oxygen was equipotent to 28

containing the H-bond donor-acceptor aniline NH. In the case of the esters 27 and 29,

solubility issues limited their concentration to 10M at which they showed <10% PI3K

inhibition. Analogous to the LY series removal of the 6-position aromatic group (30) led

to a decrease in potency (7-fold vs. 28 and 5-fold vs. 26). The corresponding ester 31 had

better solubility than 27 and 29 that allowed for its IC50 determination. The ester 31 was

slightly less active than the corresponding carboxylic acid 30. Against p110, 30 and 31

showed 26% and 19% inhibition at 100M respectively. In general all compounds were

more potent on p110 than p110 which is consistent with the trend observed for

reported arylmorpholines on PI3K enzymes. When compared to LY294002, the

isonicotinic derivatives were less potent.

74

a Table 2. Name, chemical structure and IC50 values of test compounds.

IC IC No Name Structure 50 50 PI3K, M PI3K, M

O OH

26 PC477 7 ( 1) O  23 ( 5) N N O F O O

b b 27 PC473 O N N O F O OH

28 PC481 H 5 ( 1) N  34 ( 3) N N O F O O

29 PC479 H b b N N N O F O OH

>100 30 PC386 36 ( 8) c N N (26%) O O O

>100 31 PC483 51( 32)  (19%)c N N O O HCL 6 0.2 ( 0.03) 1.2 ( 0.3) LY294002 O N [0.5-10uM]d [1.4-38uM]d O a Mean of three experiments  SEM b <10% inhibition c Percent inhibition at 100M d literature values

75

Cell assays

Our primary goal was to evaluate the ability of the synthesized compounds to inhibit

PI3K dependent cell survival and to check for off-target inhibition by examining their effect on non-PI3K dependent cell survival. Cell assays were done in collaboration with

Dr. David Plas, Department of Cancer and Cell Biology, University of Cincinnati

Medical Center. The cell lines were generated and/or maintained by Jennifer Barger (JB) from the Plas Lab and JB also designed the experiments. The cell lines for the assay were derived from the hematopoietic FL5.12 cells. Use of FL5.12 cells for studying the

PI3K/Akt pathway has been previously described.146-149 FL5.12 is an immortalized cell line that depends on a single growth factor, interlukin-3 (IL-3) for growth, proliferation and survival. Withdrawal of IL-3 leads to rapid induction of apoptosis and hence these cell lines are adaptable for high throughput screening (Fig. 34A). Expression of survival genes by the FL5.12 cells in response to stress conditions such as the withdrawal of IL-3 delays the induction of apoptosis and promotes cell survival (Fig. 34B).

A. Model system

Growth factor - IL-3 Survival gene Cell survival

Removal of IL-3

No survival gene

Cell death

76

B. FL5.12 cell lines protected by survival genes

MTJWT – wild type Tel-Jak2, D2.7 –PTEN shRNA, constitutive PI3K activity constitutive PI3K activity

Tel-Jak2 PI3K X PTEN

Akt 2M10 – myristolated Akt, constitutively active

Bcl-xL Cell Survival

Bcl-xL – protected by anti-apoptotic Bcl-xL, PI3K independent cell survival

Figure 34: A. FL5.12 cells with survival genes survive even after removal of IL-3 while those without a survival gene undergo cell death. Diagram provided by Dr. David Plas. B. Important cell lines derived from the FL 5.12 cells that were used during the assay.

Assay 1

To evaluate the PI3K dependent cell survival inhibition and off-target inhibition, compounds 28 – 31 were tested in the following four cell lines:

FL5.12 - i.e. the parent cell line which served as control for the assay.

MTJWT - cells expressing the wild type TEL-JAK2 oncogene that constitutively activates PI3K.150, 151 Under stress conditions, resistance to cell death is achieved by the constitutive activity of the PI3K/Akt pathway. Hence, addition of PI3K inhibitors to these should lead to cell death.

77

MTJKD - cells expressing a kinase-deficient (inactive) form of the TEL-JAK2 oncogene which served as control for the MTJWT.

Bcl-xL - cells over expressing the anti-apoptotic Bcl-xL gene and are not dependent on the PI3K/Akt pathway for survival. 146

Control data for the four cell lines is shown in Fig. 35A. After 48h, following removal of

IL-3, Bcl-xL and the MTJWT cells showed ~78% and ~24% viability respectively while the MTJKD and FL5.12 were not viable. For the assay, 28-31 were added at 50M while the control LY294002 at 10M based on their IC50 values from the kinase assay (Fig.

35B-E). The FL5.12 and MTJKD cells showed normal response with small percentage being viable at the end of 48h. In the PI3K dependent MTJWT cells (Fig. 35B), addition of 26-29 led to lower survival rates than the vehicle control (DMSO, neat). The esters 27 and 29 showed better inhibition than the corresponding carboxylic acids 26 and 28 (Fig.

35B). This is in contrast to the kinase assay and can be rationalized by the better cell permeability of the esters and their subsequent hydrolysis to the corresponding COOH by esterases within the cells. Thus the ethyl esters most likely serve as prodrugs which might be essential while designing inhibitors based on the isonicotinic scaffold. In FL5.12 cells expressing the Bcl-xL gene, none of the derivatives except 27 showed inhibition (Fig.

35C). In the presence of 27 these cells were 25-30% less viable as compared to vehicle control (DMSO, neat) thus suggesting possible off-target inhibition by this derivative.

This is interesting since the only difference between the esters 27 and 29 is at the 6- position where the H-bond acceptor ether oxygen is replaced by the H-bond donor- acceptor aniline NH respectively. Thus though carboxylic acids 26 and 28 were equipotent in the kinase assay (Table 1) and the corresponding esters 27 and 29 were

78

equipotent on the PI3K dependent MTJWT cells (Fig. 35B), these substitutions may have important implications towards achieving PI3K/Akt pathway selectivity.

Assay 2

Following the results from Assay 1, we tried to explore whether 28 and 29 inhibit additional enzymes within the PI3K/Akt pathway. This assay was instigated by the fact that the PI3K inhibitor LY294002, the compound on which our scaffold was based, inhibits two other enzymes, namely mTOR and DNA-PK of PI3K-related protein kinase family which contributes towards the poor specificity of this compound. 159 For the assay,

28 and 29 were tested in the following cell lines-

D2.7 - cells expressing PTEN shRNA (small hairpin RNA). Following IL-3 withdrawal, these survive via the PI3K/Akt pathway which is constitutively active as these express low amounts of the phosphatase PTEN that deactivates the PI3K.

DKD2 - cells containing the empty pKD vector which was used for transfecting the

PTEN shRNA. These were controls for D2.7.

2M10 - cells expressing myristoylated Akt (mAkt) which is constitutively active.

Following the withdrawal of IL-3, Akt allows the cells to survive for prolonged periods.146 As Akt operates downstream of the PI3K, compounds that inhibit PI3K alone would have no effect on survival of these cells.

Bcl-xL - as mentioned above.

Control data for these cells is shown in Fig. 36A. As expected cells protected by PI3K, mAkt and Bcl-xL show survival after 48h following removal of IL-3 while those with the empty pKD vector do not. For the assay, 28 and 29 were added at 50M while

LY294002 at 10M. Similarly to the previous experiment, the ester 29 showed better

79

inhibition than the corresponding COOH 28 in the PI3K dependent D2.7 cells (Fig. 36B).

In cells expressing mAkt, 28 and 29 showed some inhibition similar to LY294002 (Fig.

36C) Hence these compounds may inhibit other enzymes within the PI3K/Akt pathway, however additional experiments such as examining the expression levels of different downstream proteins is required. Though PI3K selectivity is desired, recent literature suggests that compounds possessing optimal non-selectivity may be advantageous which is exemplified by the dual PI3K/mTOR inhibitor BEZ235 (17) that is currently in Phase I clinical trials. Hence further exploration of this aspect may be beneficial. In the Bcl-xL cells, unlike previous experiment, 29 (Figs. 35C vs. 36D) showed some inhibition which may be artifactual.

80

A

B C

*

~ *

* * *

% Viability D E

Time (h) Figure 35. A. Interleukin-3 dependent FL5.12 cells protected by either constitutive PI3K activity downstream of the TEL- JAK2 oncogene (MTJWT) or Bcl-xl show viability after 48h of IL-3 removal while cells lacking survival genes - MTJKD and FL5.12 do not survive. Graphs B – E represent the activity of the test compounds on the four cell lines. Results are the average of two experiments  SD (LY294002 data is from 1 experiment). Synthesized compounds were tested at 50M and LY294002 at 10M. *, P < 0.001 and ~, P = 0.01 vs. Vehicle Ctrl (DMSO, neat), one-way ANOVA with Student-Newman-Keuls Method.

81

A

B C

# * *

*

* *

% Viability D E

*

Time (h) Figure 36. A. Interleukin-3 dependent FL5.12 cells protected by PI3K (D2.7), myristolated Akt (2M10) and Bcl-xl show viability after 48h of IL-3 removal while cells containing the pKD vector control (DKD2) do not survive. Graphs B – E represent the activity of 28 and 29 on the four cell lines. Graph A shows result of one experiment and B – E show the average of three experiments  SD. Synthesized compounds 28 and 29 were tested at 50M and LY294002 at 10M. *, P < 0.001 and #, P = 0.002 vs. Vehicle control (DMSO, neat), one-way ANOVA with Student-Newman-Keuls Method.

82

Conclusions:

In this study we evaluated the synthetic feasibility and biological activity of the proposed

2,6-disubstituted isonicotinic scaffold that we designed on the basis of modeling and literature analysis. The chemistry developed for the synthesis provides straightforward access to three chemical libraries. These compounds inhibit PI3K and  in the in vitro kinase assay and show inhibition in PI3K dependent FL5.12 cells. Overall we demonstrate the practicality of the 2,6-disubstituted isonicotinic scaffold for exploring the

PI3K and active sites.

Limitations of our experimental methods

The solubility of the esters 27 and 29 was of concern during the kinase assay as well as the cell assay. For the kinase assay, the test compounds were prepared using 5%

(v/v)DMSO solution in kinase assay buffer and the final DMSO concentration in the assay well was 1% (v/v) (5-fold dilution). For 27 and 29 the maximum feasible concentration was 50M (determined by visual inspection) which provided a final concentration of 10M for the kinase assay. At these concentrations the esters did not show inhibition of the recombinant purified PI3K and .

For the cell assay, the test compounds were prepared in 100% DMSO and added directly to the cells in the cell assay buffer. The final DMSO concentration in the cell assay was

0.1% (v/v). The esters were added at 50mM to presumably provide the final concentration of 50M for the assay (1000-fols dilution). Based on their behavior in the kinase assay, it is reasonable to assume that the esters may have encountered solubility problems in the cell assay as well. However, it is important to indicate that the buffers for

83

the two assays were different as were the dilution factors and the cell assay was performed at 37°C for 48h in contrast to the 3 hour kinase assay at room temperature. All these factors would have a strong impact on the solubility of the compounds in the two assays. Repeating the cell experiments provided consistent results as depicted in Figures

35 and 36 and the esters showed inhibition of PI3K dependent cell survival. Hence, though the esters could not be tested beyond 10M in the kinase assay they can attain a higher concentration in the cell assay where the presumed concentration was 50M.

During the discussion for Assay 1, it was suggested that the esters function as prodrugs in the cell assay and that the free carboxylic acid is indeed responsible for the PI3K inhibition inside cells. However, no experiments were performed to verify this claim.

This theory can be tested by using esterase inhibitors that would deactivate the esterases and prevent the hydrolysis of the ethyl esters.

84

IV: Structure-Activity Relationship (SAR) of 2,6-

disubstituted isonicotinic derivatives.

Introduction

The purpose of this specific aim was to explore differences between the PI3K and  active sites by a structure-activity relationship (SAR) study of isonicotinic derivatives. As our scaffold provided straightforward access to three libraries (Fig. 37), we designed a select number of derivatives based on the 3D Interior structures and variations in the sequence of the two isoforms. The SAR involved - O O N A. Modification of the carboxylic acid. A O B N B. Substitution of Morpholine.

C. Substitution at the 6-methyl position. X H3C As discussed previously docking of 28 into the C p110 model indicates that the 4-COOH and F 2-morpholine are located in the interior of the Entrance Figure 37: Three regions of the binding pocket while the 6-aromatic group is at isonicotinic scaffold that were modified for the structure-activity relationship the entrance of the pocket. A comparison of the (SAR) study. p110 and p110 sequences within 12Å of 28 in the p110 homology model is shown in

Fig. 38 along with residues within 6Å of each of the peripheral fragments. In both isoforms, the morpholine pocket consists mostly of hydrophobic residues that engage in limited polar interaction with the ligand through their main chain amides. This pocket contains three pairs of varying residues of which one was of particular interest to us - the

85

polar H-bonding Ser854 in p110 vs. the non-polar Ala885 in p110. Modifications that can engage this serine provide opportunity for improving potency as well as selectivity for p110. Unlike the morpholine pocket the region surrounding the 4-COOH is very polar and hence provides a better chance of improving the potency of these compounds through polar interactions. As the above two regions are in the interior of the binding pocket, modifications in these should have a larger impact on potency as compared to the

6-position which is located at the entrance and exposed to solvent. This was observed in the TGX series where a wide range of aromatic substituents were accommodated in this region with little effect on their potency. However, comparing the sequences indicates that a majority of the variations in the two isoforms are observed at the entrance of the binding pockets alongside the 6-aromatic group. Hence modifications in this region provide a better chance of discriminating between the two isoforms. In addition to the above characteristics, the size and physicochemical properties of the different amino acids affect the overall shape and flexibility of the binding pockets which in turn will influence the interaction of the ligands with the two isoforms. Overall these regions within the two enzymes provided opportunities for improving the potency and/or selectivity of ligands and we explored these through modifications of the 2-, 4- and 6- positions of our isonicotinic scaffold.

86

* + ^^^+ ^* p110 730 V--QMKFL VE QMRRPDFMDA LQGFLS P LNP AHQLGNLR L E 7 67 p110 763 V I S QLKQK LE NLQNSQLPES ---FRV P YDP GLKAGALA I E 799 +* + + * + *^ * + ****++ + ^^ ^ *** * + **** 768 ECRIMSSAKR PLWLNWENPD IMSELLFQNN EIIFKNGDDL 8 07 800 KCKVMASKKK PLWLEFKCAD –PTALSNETI GIIFKHGDDL 8 38 **** ^* * * ***** *^*^ ^ + * + 808 R QDML TLQII R IMENIWQNQ GLDLRMLPYG CLSIGDCVGL 847 839 R QDML ILQIL R IMESIWETE SLDLCLLPYG CISTGDKIGM 878 **+ * + + + ^ ** ^+ ** + ^ *** * 848 IEVVRNSHTI MQIQ-CKGGL KGALQFNSHT L HQWLKDK NK 886 879 IEIVKDATTI AKIQQSTVGN TGA--FKDEV L NHWLKEK SP 916 **+* + ***+ ++ 887 GE-IYDAAID LFTRSCAGY C VATFILGIGD RHNSNIMVKD 925 917 TEEKFQAAVE RFVYSCAGY C VATFVLGIGD RHNDNIMITE 956 +* + ** * * *** * 926 DGQLFHIDFG HFLDHKKKKF GYKRERVPFV LTQDFLIVIS 965 957 TGNLFH IDFG HILGNYKSFL GINKERVPFV LTPDFLFVMG 996

Figure 38: Amino acid sequences of p110 and p110 were aligned with BLAST. Residues within 12Å of 28 (determined by CAChe) are underlined and are labeled conserved (*), similar (+) and different (^). Colored bars are residues within 6Å of COOH (green), morpholine (blue) and 6-Ar (orange).

Results and discussion

A. Modifications of the carboxylic acid group.

Docking of 28 into the p110 model suggests that the COOH of the isonicotinic scaffold

occupies the same region as the triphosphate portion of ATP in the p110 X-ray structure

(Fig. 39A, B). This region is highly polar and residues contained in this region engage in

numerous polar interactions with each other as well as with the ligand. Hence altering the

H-bonding properties of our scaffold within this region was expected to affect the

potency of our compounds. In order to explore this we modified the COOH as seen in

Fig. 39C. As compared to the COOH which will be ionized at pH 7.4 none of these

derivatives are ionizable. Similar to the COOH the aldehyde 39 is an H-bond acceptor but

87

less polar and without the charge while the alcohol 40 is also less polar but is an H-bond donor/acceptor. The more polar H-bond donor/acceptor amides 41–43 can form additional H-bonds as compared the above derivatives and the hydroxamic acid 42 and the ethanolamide 43 can extend further into this pocket similar to the  and -phosphates of ATP.

A Asn951 ATP/p110 C Asp950 X-ray(1e8x) R Asp964 CH3 X Lys807 N N Asp836 Ser806 O F Lys833 Lys808 No. X R

B (28)/p110 Asn921 Model 39 NH CHO

Asp933 40 NH CH2OH

Asp805 41 O CONH2

Lys776 42 O CONH-OH Ser774 Lys802 OH 43 O COHN

Figure 39: (A) Triphosphate of ATP in p110 (1e8x). The figure shows H-bonding residues (yellow) and other polar residues (white) in the neighborhood. (B) The COOH of 28 interacts with Lys802 in the same region of p110. The figure shows polar residues within a 6 Å radius of COOH. Structures are viewed from same angle. (C) Modifications of the COOH group.

Chemistry

The derivatives were synthesized as shown in Fig. 40A. For 41-43 the oxygen atom was selected at the “X” position instead of NH to avoid side reactions during amide formation

(Fig. 39C). Reaction of 26 with ammonia in methanol provided carboxamide 41.

Compounds 42 and 43 were prepared by treating 26 with ethylchloroformate to form the mixed anhydride and then reacting with hydroxylamine or ethanolamine respectively.

88

Treatment of 29 with LiAlH4 reduced the ester to alcohol 40. Synthesis of 39 was somewhat complicated as compared to the rest (Fig. 40B). Attempts to reduce the ester with DIBAL-H or alternatively by converting the ester to the Weinreb amide (29w) and

152 treating with LiAlH4/Et2O led to complete reduction to the alcohol 40. Finally the aldehyde 39 was prepared by oxidizing 40 with Dess-Martin Periodinane.153-155

A O NH2 O OR R = Et; (40, 41) OH or H; (42, 43)

10%NH3/MeOH, 60°C 1M LiAlH4/Et2O

N N N N N N O O X O X = O NH O or NH

CH3 F CH3 F CH3 F 41 26-29 40

Dess-Martin Ethylchloroformate DCM Periodinane NMM/ Et2O/NH2-R

O NH R H O R = OH R = -(CH2)2-OH

N N N N O O NH O

CH3 F CH3 F 42, 43 39 B O OEt H O OH

DIBAL-H Reduction

N Y N Y N Y Oxidation X 29 X 39 X 40 1] PCC, No reaction CH3 O 2] Dess-Martin Ethylchloroformate O N Periodinane CH NHO(CH3)2. HCl 3 LiAlH4

Weinreb Amide N Y X 29w

Figure 40: A. Synthetic scheme for 39-43. B. Optimizing the synthesis of 39.

89

Biological evaluation

Lipid Kinase Assay

All compounds were to be tested at 100M initially; however, solubility problems limited the testing of the amide 41 and the hydroxamic acid 42 at 10M and the ethanolamide 43 at 50M. From Fig. 41, reducing the H-bond acceptor COO- to the less polar CHO (39) led to a decrease in potency on both isoforms with a larger difference observed for the  isoform. Further reduction to the H-bond donor/acceptor CH2OH (40) did not affect the potency on p110 but improved potency on the p110 when compared to 39 (Fig. 41).

The alcohol 40 is only 4-fold less potent than carboxylic acid 28 (Table 3) which suggests that the COOH interacts with the enzymes by H-bonding and not by salt bridge formation. The amides 41, 43 as well as the hydroxamic acid 42 showed <50% inhibition at their respective concentrations. Extending the H- bonding by addition of OH to the carboxamide (41 vs. 42) did not affect the potency on the  isoform but showed

Figure 41: Percent inhibition of compounds determined using the Promega Luminescent Kinase Glo™ assay. Compounds were tested at 26, 28, 39, 40 and LY294002 at 100M, 41 and 42 at 10M and 43 at 50M. Results are average of three experiments  SD.

90

considerable improvement on the p110 (Fig. 41). These differences in potency between the aldehyde vs. alcohol as well as the amide vs. hydroxamic acid on the two isoforms suggests that subtle differences exist in this region of the enzymes in addition to the high polarity. Thus though the COOH was the best with regards to potency in the above series it is not vital for activity and can be replaced by other polar functional groups and varying the H-bonding properties of the scaffold within this region also influences the selectivity of the compounds.

Table 3: Biological activity of 39-43.a IC , M No. 50 PI3K PI3K 26 7 ( 1) 23 ( 5) 28 5 ( 1) 34 ( 3) 39 ~100 ~100 40 20 ( 5) >100 41 >10 - 42 >10 10 43 >50 >50 a IC50 values represent means of three experiments  SEM.

B. Substitution of Morpholine.

The morpholine of 28 in the p110 model occupies the same pocket as the morpholine of

LY294002 and the adenine of ATP in the p110 X-ray (Fig. 42). Within this region residues involved in polar interactions with the ligand include Val851 which H-bonds with the morpholine oxygen of 28 through the backbone amide NH (similar to morpholine of LY294002) and the adjacent Glu880 which interacts with the 6-NH2 of the

ATP adenine through the backbone amide C=O. A 6Å radius around this morpholine consists of 11 residues of which 8 are identical and the remaining 3 are conservative (Fig.

91

38). The 3 conservative residues include

Ser854/Ala885

Val851/Val882 Val850, Arg852 and Ser854 in p110 that

are substituted by Ile881, Lys883 and

Ala885 respectively in p110 The

Glu849/Glu880 substitution of the polar H-bonding

Figure 42: Important residues of p110 (green) and Ser854 in p110 by the nonpolar Val885 p110 (yellow) in the morpholine pocket in p110 presents a favorable opportunity for discriminating between the two isoforms.

Accordingly we hypothesized that addition or removal of H-bonding fragments to the morpholine template will impact the interaction with either the PI3K or  isoforms.

Hence we explored this pocket using a set of flexible 44-46, semi-rigid 47-50 and rigid

51-55 analogs of morpholine in 28 (Fig. 43). The flexible derivatives 44-46 can be considered as structural fragments of the morpholine ring which not only provided flexibility but also H-bond donor/acceptor oxygen in the corresponding position as the H- bond acceptor oxygen of morpholine. Removal of the CH3 from the nitrogen (45 vs. 44) provided an additional site on the scaffold for exploring the H-bonding while the more polar diethanolamine 46 allowed for scanning additional conformational space. Steric analog piperazine 47 allowed for exploring the H-bond acceptor vs. donor effect as the 4-

NH will be ionized at pH 7.4 while the non-ionizable 48 and 49 were selected to further evaluate the importance of H-bonding in this position. The 2-pyrrolidinol 50 introduced the H-bond donor/acceptor OH to the semi-flexible series. Rigidity in the set was introduced by the use of aromatic groups which included the isosteric 4-pyridine 51, the

3-pyridine 52 with the H-bond acceptor shifted to adjacent position and the H-bond donor/acceptor hydroxyphenyls 53-55 extend the H-bonding by one atom. In addition to

92

the above, two bis-derivatives 45a (ethanolamide) and 50a (pyrrolidinol amide) which were isolated during synthesis of 45 and 50 respectively and the 2- chloride derivative

35a were also tested (Fig. 44).

Flexible Semi-flexible

N N N N N 44 CH3 OH N N-Methylethanolamine C OH H H2 Piperazine Piperidine Pyrrolidine 2-Pyrrolidinol COOH 47 48 49 50 N 45 H CH3 OH H Ethanolamine N Rigid N N O F 28 N 51 52 46 N OH N 4-Pyridine 3-Pyridine OH Diethanolamine OH

OH OH

4-hydroxyPhenyl 3-hydroxyPhenyl 2-hydroxyPhenyl 53 54 55

Figure 43: Substitution of morpholine

Chemistry

All analogs were synthesized from 35 by displacement of the chlorine (Fig. 44A). In case of cyclic amines regular heating with 35 provided the desired analogs 47-50. For the aliphatic amines regular heating was insufficient and the only product obtained was the amide that formed due to coupling between the ethyl ester and the amine (Fig. 44B).

Hence 44-46 were prepared by microwave heating of 35 with excess amine that formed the bis-derivatives (e.g. 45a) which were then hydrolyzed to the corresponding carboxylic acids. The aromatic substituents in 51-55 were added by Suzuki reaction.

93

A HNR1R2, MW, 200°C O OEt 44-46 H2O

H O 35a 2 HNR3R4,1,4-dioxane, 80°C N Cl 47-50 H O NH 2

CH3 F 35 Ar-B(OH)2, Pd(dppf), 2M Na2CO3, 1,4-Dioxane, N2 51-55 H2O B H H O N O OEt O N O OH OH OH NH (CH ) OH NH2(CH2)2OH 2 2 2 H2O 1,4-dioxane, 80°C MW, 200°C OH OH N Cl N Cl N N N N H H NH NH NH NH

CH F CH F 3 3 CH3 F CH3 F 35 45a 45 Amide, Only Product Bis-derivative

Figure 44: A. Synthetic scheme for 44-55. B. Optimizing the synthesis of 45. Compounds 44-46 were prepared in similar manner.

Biological evaluation

Lipid Kinase Assay

Replacement of the morpholine with flexible counterparts 44, 45 led to considerable drop in potency while the more polar and flexible diethanolamine 46 was practically inactive

(Fig. 45). Analog 45 which lacked the N-CH3 showed comparable potency to 44 suggesting no involvement of the H-bond donor/acceptor nitrogen at this position (Fig.

9). Replacement of the morpholine oxygen with NH (47) or CH2 (48), its complete removal (pyrrolidine derivative 49) or removal with addition of OH group (3-OH pyrrolidine 50) led to a significant drop in potency (Fig. 9) indicating lack of direct

94

interaction with Ser854. The observed decrease in potency (similar to that for LY294002) for protonated 47 and neutral non-basic 48 was most likely caused by loss of H-bonding with the backbone NH. While conversion of the hydrophilic H-bond acceptor oxygen in

+ morpholine 28 into hydrophilic H-bond donor NH2 in 47 did not affect selectivity, the

+ replacement of NH2 with the isosteric non-polar CH2 (48) considerably reduced potency at PI3K but not at PI3K(Fig. 45). Compounds 48 and 49 showed similar potency and selectivity (Fig. 45), i.e. the 4-CH2 in piperidine is not essential. At the same time, addition of 3-OH (50) to the pyrrolidine 49 which enables polar interactions does not affect the activity at PI3K but increases activity at PI3K (Fig. 45). This biological profile agrees with the above difference in properties of the Ser/Ala pairs and potential involvement of the hydrophobic Ile881 in p110 vs. corresponding Val850 in p110 located in the morpholine binding pocket.

Replacement of the morpholine in 28 by flat and rigid 4-pyridine (51) retained PI3K

Figure 45: Percent inhibition was determined as before. All compounds were tested at 100M. Average of three experiments  SD.

95

inhibitory activity; however, shifting the H-bond acceptor to the adjacent position (52) led to significant drop in potency (Fig. 45). IC50 determination indicates that replacement of the morpholine with 4-pyridine led to a 6x drop in potency on PI3K but no change on

PI3K (Table 4). Extending the H-bond by means of hydroxyphenyls 53-55 led to further loss of potency than the 4-pyridine (51) (Fig. 45). Unlike the 3- and 4-pyridines

(52, 51) which show significant difference in potency the 3- and 4-OHPh (53, 54) showed comparable potency with the 3-OHPh (53) being slightly better on PI3K (Table 4 and

Fig. 45). The IC50 value of 3-OHPh (54) on PI3K was similar to 28. The 2-OHPh (55) derivative led to further loss than 53 and 54.

Overall, the flexible 44-46, semi-flexible 47-50, the bis-derivatives 45a, 50a and the aromatic 3-pyridine (52) were only equipotent to the chloride 35a which suggests lack of the necessary H-bonding by these derivatives within this region (Fig. 45). None of the substituents (44-55) was able to engage Ser854 as no improvement in PI3Kselectivity was achieved. The aromatic substituents 51-55 were generally more potent on PI3K vs.

PI3K. The retention of activity by 4-pyridine (51) as compared to 3-pyridine (52) corroborates the importance of H-bond acceptor in the 4-position of the ring. It is significant that the morpholine (28), the 4-pyridine (51) and the 3-OHPh (54) show similar potency on PI3K but show considerably different potency on the PI3K (Table

4) which suggests that modifications in this pocket might be more sensitive for p110 vs. p110. The results also suggest that the morpholine is not vital for activity and can be replaced by aromatic substituents with polar substituents in the 3 or 4-position; however as compared to the aromatic rings, the morpholine does show preference for PI3Kover

PI3K.

96

Table 4: Biological activity of 44-55. a

IC50, M IC50, M No. No. PI3K PI3K PI3K PI3K 28 5 ( 1) 34 (  3) 51 29 ( 7) 42 (  3) 44 >100 >100 52 >100 >100 45 ~ 100 >100 53 ~ 100 ~ 100 46 >100 >100 54 ~ 100 42 (  6) 47 >100 >100 55 >100 >100 48 b >100 35a >100 >100 49 b >100 45a >100 >100 50 >100 >100 50a >100 >100 a IC50 are average of three experiments  SEM. b<10% inhibition at 100M.

C. Substitution at the 6-methyl position

In the p110 model the 6-position aromatic group of 28 is located at the entrance of the binding pocket. Previously it was mentioned that this portion of the enzymes contains a hydrophobic patch that may be responsible for enhancing the potency of these compounds by hydrophobic interactions with the aromatic ring (as seen in the LY series)

(28)/p110 model and secondly this region shows the highest

variability in the amino acid sequence of the Gln859 His855 Ser774 two isoforms. Further examination suggests

that this region also contains several polar Ser773 residues adjacent to the hydrophobic patch Arg770 Arg852 (Fig. 46). Therefore aromatic substituents

Figure 46: Docked structure of 28 in p110 model that can engage these residues in H-bonding shows several hydrophilic residues at the entrance of the binding pocket.

97

may improve the potency of these compounds. Furthermore the Arg770 and Trp780 within this region are possibly involved in cation- interactions which are frequently observed in proteins and contribute towards their stability (Fig. 47).156-158 Hence disruption of this complex by engaging the arginine via H-bonding with the ligand presents additional opportunity for improving the potency. In light of these observations we synthesized derivatives containing varying H-bonding properties (Fig. 48). These included the H-bond acceptor nitro derivatives 56, 57 and 65, the H-bond donor/acceptor sulfonamides 58, 59, phenolic 62 and the anilines 60, 61. Additionally in contrast to the above polar series we also tested the hydrophobic and extended 4-OBzl derivative 63 which was an intermediate for synthesis of 62. The acetylated 64 allowed for further evaluating the H-bonding at the aniline NH of these derivatives.

A (28)/p110 model B

Trp780

T-shaped Parallel Arg770 Cation- complex

Figure 47: A. Cation- interaction between the Arg 770 and Trp780 in the docked model of 28 in p110. B. Two geometries of cation- interaction involving arginine and tryptophan.7 Other amino acids involved in such interactions include the cationic lysine and the aromatic phenylalanine and tyrosine. Histidine can function either as cation or  group depending on its protonation state.

98

O Hydrophobic Polar H HN H3C N N

62 64 63

O O O H

HN HN

O H N HOOC N N H H H N 60 61 H N HN NH HN H2C

CH3 65 O O N N N O O F O O 56 57 28 HN HN

58 59 O S NH2 S O O O NH2

Figure 48: Substitution at the 6-methyl position.

Chemistry

The aromatic substitution was achieve by nucleophilic displacement of 6-Br in 33 (Fig.

49). Addition of the p-OBzl aniline and the m-nitrobenzylamine proceeded at room temperature (63b, 65b). The addition of the p-NO2 derivative under reflux in CH3CN was extremely slow and required extended reaction times for completion. Hence the less

99

nucleophilic p/m-NO2 and p/m-SO2NH2 derivatives were added by microwave heating at

200°C (56b-59b). Next, displacement of the 2-chlorine by morpholine under microwave heating followed by alkaline hydrolysis provided 56-59, 63 and 65.

COOEt COOEt

DMF, K2CO3, rt, 24h 63b, 65b m-NO PhCH NH 2 2 2 N Cl N Cl or Br 33 p-OBzlPhNH2 HN R

1] MW, 150°C, 30m H2N Morpholine X MW, 200°C, 75m 63a, 65a KF/Celite, CH CN 3 2] HOH

63, 65 COOEt 1] MW, 150°C, 30m Morpholine 56a - 59a N Cl 56 - 59 2] H O HN 2 X X = m/p- NO2, SO2NH2 56b - 59b

Figure 49: Synthetic scheme for 56-59, 63 and 65. The corresponding ethyl esters of the desired products are labeled with the suffix “a”; e.g. 63a is the ethyl ester of 63. Similarly the 2-chloro intermediate are labeled with suffix “b”.

The amine derivatives 60 and 61 were synthesized by reduction of the corresponding nitro derivatives 56 and 57. The reaction was complicated by benzylic reduction and other factors (Fig. 50) and was ultimately achieved by use of Fe/CH3COOH.

100

Figure 50: Synthesis of amino derivatives.

O OEt O OEt O OEt

Conditions below + N N N N H3C N N HN O HN O 66 O

NO2 NH2 60a Side product

Yield(%)a Conditions for reduction of NO 2 60ab 66

Pd/C, H2, MeOH,rt, 2h, solubility issues on scale-up 85 15

Zn, NH2-NH2. HCOOH, MeOH, rt No reaction

Al/NH4Cl, sonication No reaction

Fe/CH3COOH, 2h, 50°C 100 --- a Approximate values based on HPLC. b Ethyl ester of 60.

For synthesis of 62 conditions had to be optimized to achieve selective O-debenzylation in the presence of N-benzyl group (Fig. 51). Reduction under standard conditions using

Pd/C led to the N-debenzylation product 66. This issue was solved by addition of 5mol% of n-butylamine to the reaction mixture that led predominantly to O-debenzylation and the 62a was obtained in excellent yields. Nevertheless, hydrolysis of ethyl ester 62a was complicated by formation of the aldehyde 67 as determined by NMR and MS analysis.

This reaction is probably due to the combination of presence of atmospheric oxygen, base and refluxing temperature that led to the cleavage of the 4-OHPhNH in 62 via a quinone type intermediate. Under these conditions the yellow reaction mixture turned dark brown upon heating. Hence 62 was synthesized by selective O-debenzylation of 63 which is very stable. An alternative route would be to hydrolyze 62a using mild conditions while maintaining N2 atmosphere. Compound 64 was synthesized by acetylation of 63a using

101

acetic anhydride followed by mild alkaline hydrolysis to avoid cleavage of the acetyl group (Fig. 52).

Figure 51: Optimization of O-debenzylation in presence of N-benzyl group.

O OH O OEt O OEt

Conditions KOH/H20/EtOH A or B + 66 reflux, 1.5h (Side product) N N N N N N HN O HN O HN O

OBzl OBzl OH 63 63a 62a

Condition B KOH/H20/EtOH reflux, 1.5h O OH O OH

N N H HN O N N O O OH 67 62 (Undesired product)

Yield(%)a Conditions for O-Bzl reduction 62ab 66

A = Pd/C, H2, DCM,rt, 3h <1 30 B = Pd/C, DCM,rt, 1h, 5mol% n-butylamine 85 15 a Approximate values based on HPLC. b Ethyl ester of 62.

Figure 52: Synthesis of 64.

O OH O O O O

Acetic Anhydride, CHCl3 LiOH/THF/H2O N N N N rt, 1h N N rt, 3h O N O HN O O N O

CH3 CH3 OBzl OBzl OBzl 63a 64a 64

102

Biological evaluation

Lipid Kinase Assay

Modification in this region was better tolerated than the previous two and the substituents generally showed better potency on the p110 than p110 (Fig. 53). From Table 5, replacement of the 4-F-2-CH3Ph by the 4-NO2Ph (56) showed a 2-fold decrease in potency on p110 but 2-fold improvement on p110. As compared to the 56, shifting the

NO2 to the 3-position 57 retained potency on p110 but led to 2-fold decrease on p110.

Replacement of the H-bond acceptor NO2 by the H-bond donor/acceptor SO2NH2 (58,

59) or the aniline NH2 (60, 61) led to further decrease in potency with the 3-NH2 (61) being the most detrimental. Of the above three pairs of 3- and 4-substituents, the 4- substituents were generally more potent than the corresponding 3-substituents on both the isoforms. Replacement of the weak H-bond acceptor 4-F in 28 by the isosteric H-bond donor/acceptor 4-OH (62) led to >4-fold decrease in activity. Interestingly, replacement

Figure 53: Percent inhibition was determined as before. All compounds were tested at 100M. Results represent average of two experiments  SD.

103

of the 4-polar substituents by the extended and hydrophobic 4-OBzl group (63) retained potency on the p110 but showed additional loss on the  isoform. Accordingly 63 is

>10-fold selective for p110 and in this respect is better than 28 which shows 7-fold selectivity. Furthermore, the extended 3- NO2 benzylamine (65) also showed better selectivity for the p110 over the p110with comparable potency on the p110 to the other polar derivatives. These results suggest that extended substituents within this region are better tolerated by the p110 than p110and the activity of the polar substituents may not be due to the H-bonding alone but also due to steric factors. Previously it was observed that replacing the H-bond donor/acceptor aniline NH in the 6-position by H- bond acceptor ether O (28 vs. 26) did not affect the potency. In this series removing the

H-bonding of the 6-position aniline NH by acetylation also did not cause any change in potency (64 vs. 63); however, this does not exclude the possibility of polar interaction via the amide C=O at this position. Nevertheless the aniline NH provides another site on our scaffold that can be further explored e.g. for designing prodrugs of these compounds.

Table 5: Biological activity of 56-65.a

IC50, M IC50, M No. No. PI3K PI3K PI3K PI3K 28 5 ( 1) 34 (  3) 61 42 ( 10) >100 56 10 ( 2) 18 ( 6) 62 22 ( 3) ~100 57 11 ( 2) 38 ( 11) 63 10 ( 1) >100 58 13 ( 2) 41 ( 6) 64 10 ( 2) >100 59 15 ( 4) 63 ( 16) 65 14 ( 2) ~100 60 19 ( 4) ~100 a IC50 represents the means of three experiments  SEM.

104

Docking of 63 with CAChe

In order to understand the selectivity shown by 63, it was docked into the p110 model and the p110 X-ray structure (Fig. 54). Docking suggests that the 4-OBzl group in 63 can access a cavity near the entrance of the binding pocket that may not be available to the shorter 4-F-2-CH3Ph (28) as well as the 3/4-substituted compounds 56-62. This cavity is formed by 5 residues of which 4 are variable between the two isoforms and includes the cation- pair of Arg770/Trp780 in p110and corresponding Lys802/Trp812 in p110. The other 3 residues include the Glu768, Asn782 and Glu789 in p110 that are replaced by Lys800, Glu814 and Gly829 respectively in p110. These variable residues will affect the overall size and shape of this cavity which in turn can influence the interaction of the ligand within this cavity and thereby the selectivity.

Docking shows a more compact fit of 63 in the p110 pocket as compared to p110with the 4-OBzl extending deeper within the cavity in p110 vs. p110(Fig. 54A, B).

Accordingly, one reason for the selectivity can be the hydrophobicity of the cavity. As mentioned above the Arg770 in p110 is replaced by Lys802 in p110 and arginine is relatively more hydrophobic than lysine due to its delocalized  system. Hence the

Trp780 combined with the flat structure of Arg770 in p110 provides a better environment for hydrophobic interaction with the 4-OBzl as compared to corresponding

Trp812 and Lys802 in p110. Another reason for the selectivity can be cation- interactions. It has been observed that these interactions prefer arginine over lysine arguably due to its relative hydrophobicity and these cationic residues are capable of forming multiple cation- interactions with adjacent aromatic groups in the protein.151-153

Hence it is possible that the Arg770 in p110 is involved in cation- interaction with the

105

A B

C D

Gln859 Lys890

Trp780 Trp812 Arg770 Lys802

Figure 54: Docking of 63 in p110 model [A, C] and in LY294002/p110X-ray (1e7v) [B, D] viewed from the same angle. Ph of the 4-OBzl in addition to the Trp780 while this interaction may not be occurring in the p110.

Docking also shows that the Gln859 in p110 is facing towards the aniline NH of 63 while the corresponding Lys890 in p110 is directed away form the binding pocket (Fig.

54C, D). Hence polar interactions between the ligand and the enzyme within this region may be an additional factor contributing towards the selectivity. All these observations

106

suggest that 63 would be better accommodated in p110 than p110 which parallels the observed selectivity profile.

Cell Assay

The 4-OBzl 63, the 4-NO2 56 and their corresponding ethyl esters 63a and 53a were evaluated for PI3K dependent cell survival inhibition and off-target inhibition (Fig. 55).

These compounds were selected since 56 was the least selective (1.8-fold) between

PI3K and  while 63 was the most selective (>10-fold) from the series (Table 5). These compounds were also compared to the previously tested 28 and corresponding ester 29.

The cell lines are described in chapter III. For this study all compounds were tested at

10M.

In cells protected by PI3K (MTJWT, Fig. 55A), the esters were more potent than the corresponding carboxylic acids similarly to the previous assays. Compounds 63a and 56a were more potent than 29 which is in contrast to the IC50 of their carboxylates where 28 is 2x more potent than 56 and 63 (Table 5). This effect is most likely due to the different physicochemical properties of these compounds rather than direct PI3K inhibition. One obvious physicochemical advantage for 63a over 29 is the hydrophobic 4-OBzl group which can improve its permeability in the cells. For 56a, the metabolism of the 4-NO2 group is a concern and these results are being treated with caution. Further evaluation is necessary to determine the cellular effects of these compounds and also the effects of their selectivity seen in the kinase assay. The inhibition shown by 63a and 56a was comparable to each other and to LY294002.

In cells protected by the downstream mAkt (2M10, Fig. 55B), none of the isonicotinic derivatives showed inhibition suggesting that they may not inhibit the other enzymes

107

within the PI3K/Akt pathway. As mentioned earlier, one of the disadvantages of

LY294002 is its non-specificity and it inhibits the mTOR and DNA-PK downstream of

the PI3K. This is evident from the inhibition shown by LY294002 in the cells protected

by mAkt. Thus, 56a and 63a provide the advantage of being more specific for the PI3K

as compared to LY294002. However further examination is warranted to corroborate

these results. The lack of inhibition shown by 28 and 29 in contrast to the previous cell

A B

* * ^ *

% Viability C D

Time (h) Figure 55. Compounds 28, 56 and 63 and their corresponding esters 29, 56a and 63a were tested at 10M in four FL5.12 derived cell lines. Results are the average of at least two experiments done in triplicates  SD. *, P<0.001 and ^, P = 0.001 vs. Vehicle control (DMSO, neat), one-way ANOVA with Student-Newman-Keuls method.

108

assay (Fig. 36A) can be explained on the basis of the lower concentration (10M vs.

50M) used in this assay. For evaluating the non-PI3K related inhibition, these compounds were tested in cells protected by the anti-apoptotic BclxL (Bcl-xL, Fig. 55C) in which none of the isonicotinic derivatives showed inhibition.

In summary, our modifications at the 6-position did not improve the potency over 28 suggesting that none were able to engage the polar residues at the entrance of the binding pockets. As expected, these modifications had a strong impact on the PI3K vs.  selectivity with the 4-Obzl derivative 63 proving to be the most PI3K selective compound from the isonicotinic series. Two of the derivatives 63 and 56 were tested in the cell assay and found to better than 28.

Conclusions

Based on our modeling and the protein sequences, we modified the 2-, 4- and 6-positions of the isonicotinic scaffold with the intent of exploring the p110 and  binding sites. In agreement with the modeling, modifications of the 2- and 4-positions had a greater impact on the potency as compared to the 6-position. Though the corresponding regions surrounding the 2- and 4- positions are similar in the two isoforms subtle variations exist that affect the selectivity of these compounds. Modification of the 6-position provided the highest selectivity between PI3K and  arguably since the region surrounding the 6- position is the most variable in the two isoforms and these modifications also affected the potency in the cell assay. The highlight of the SAR study was compound 63 which not only shows >10-fold selectivity for the PI3Kover the PI3K but based on our cell assay is also more specific for the PI3K than the known inhibitor LY294002 (Fig. 56). Overall,

109

this study provides valuable insight regarding the PI3K and  binding sites that may be applicable towards designing better inhibitors.

Figure 56. Improvement in specificity provided by 63 over LY294002

PI3K-related protein kinases

O OH

63 PI3K N N HN O O mTOR O O N ATR O

Smg-1 LY294002

DNA-PK

110

V. General discussion and Future directions

Generation of the PIP3 by the PI3Ks influences multiple cellular processes including growth, survival, motility and immunity. Deregulation of the class I PI3K activity is common in various pathological conditions and these enzymes are essential targets for the treatment of inflammation, autoimmune diseases, cardiovascular disorders and cancer. Of the four class I PI3Ks (p110, ,  and ), the PI3K is an important chemotherapeutic target as the PIK3CA gene that encodes for the p110 catalytic subunit is frequently mutated in various cancers. Even though several classes of compounds that inhibit PI3Ks have been reported developing PI3K selective inhibitors is still a major challenge. In this study we explored the PI3K and  binding sites using homology modeling and inhibitors utilizing a 2,6-disubstituted isonicotinic scaffold as summarized in Figure 57.

The highlights of the study include the p110 model which proved to be in good agreement with the later published X-ray structure (2rd0), the isonicotinic scaffold that provides straightforward access to three chemical libraries and compound 63 which showed >10-fold selectivity for PI3K over PI3K and improved specificity over the widely used PI3K inhibitor LY294002. The findings that replacement of the morpholine at the 2-position by 4-pyridine causes loss of potency on the PI3K vs. PI3K while extended substituents at the 6-position causes loss of potency on the PI3K vs. PI3K are important while designing isoform selective inhibitors. The improvement in PI3K- dependent FL5.12 cell survival inhibition upon conversion of the carboxylic acid to the ethyl ester demonstrates the prodrug potential of these compounds.

111

Figure 57. Summary and Future directions

OH PI3K Y O sequence Modeller 1] Literature analysis N YASARA 2] Docking- CAChe PI3K X X-ray Proposed 2,6-disubstituted isonicotinic scaffold

1]Design PI3K model 2] Synthesis

3] Testing - in vitro kinase assay 63 docked in PI3K model - FL5.12 cell assay

Asp805 Asp933 COO OOC HN Glu849 H A 2 O O B Lys802 N O O O H O N N H NH Val851 O NH O 3 N N HO Lys776 Ser854 1] SAR -Exploring A, B and C NH 63 NH O 2] Docking -analyzing selectivity of 63 H3C Gln859 H N 2 C F O PI3Kselectivity 28 NH2 rationalized by interaction of 7x PI3K vs. PI3K selectivity HN NH the 4-OBzl Arg770 with Arg770 and Trp780 NH Trp780

>10x PI3K vs. PI3K selectivity

Future directions

1] Substitution at 6-Me 2] Substitution of Morpholine 3] Modification of COOH

- explore interaction of 4-OBzl - explore H-bonding and engage - explore H-bonding with Arg770 & Trp780 in p110 Ser854 in p110 N N O N NH NH N y = OH, OCH3, F, Y CH F2, CF3 Tetrazole 2 and 3-substituted

O O COOH, CONH , F, CF N 2 3 O X = 1,4-dioxa-8-aza-spirodecane 4 2 2 3 F F X 4 3 3 N HN F F H N N 4,4-Difluoro- 3,3-Difluoro- Indoles piperidine pyrrolidine

112

The future direction of this project should entail the further exploration of the scaffold based on the generated data. With regards to the 6-position, the interaction of the 4-OBzl of 63 with the Arg770 and Trp780 of p110 should be further examined. Addition of a

COOH or CONH2 to the 4-OBzl might be able to engage the Arg770 in polar interactions by salt bridge formation (COO-) or H-bonding. A fluorine or CF3 on the benzyl ring can interact with the positively charged central C atom of the guanidinium side-chain. The use of indoles may provide better hydrophobic interactions with the Trp780 as well as compete for cation- interaction with the Arg770.

At the 2-position, the H-bond acceptor morpholine oxygen is important for PI3K inhibition and secondly the morpholine provides selectivity for p110 over . Hence further exploration of the H-bonding within this region should be performed using substituted morpholines containing polar groups at the 2- or 3-positions which also provide better means of engaging the Ser854 in p110. The difluoro and dioxospiro derivatives would be good additions to the previously explored semi-flexible series

(chapter IVB) which included the unsubstituted/non-bonding piperidine and pyrrolidine as well as the H-bond donor acceptor pyrrolidinol. Finally the replacement of the 4-

COOH by the bioisosteric tetrazole should be explored. The tetrazole provides the acidity and the planarity of the COOH but as it is larger in size it should be able to access this pocket further.

In conclusion, we successfully demonstrated the utility of homology modeling and 2,6- disubstituted isonicotinic scaffold for exploring the p110 and  isoforms and anticipate that the data generated during this study may be useful toward designing more potent and selective PI3K and  inhibitors.

113

VI. Experimental

The 2-chloro-6-methylisonicotoinic acid and 2-chloroisonicotinic acid were purchased from Sigma-Aldrich. The 2-chloro-6-methylisonicotoinic acid ethyl ester was purchased from Oakwood Products Inc, South Carolina. All other chemicals and anhydrous solvents were purchased from either Sigma-Aldrich or Fisher Scientific. Solvents were generally purchased from Pharmaco-Aaper, Kentucky and were HPLC grade. Microwave heating was performed in a Biotage initiator ™ Sixty using microwave vessels (Biotage) sealed with aluminium caps and Teflon septa. Analytical HPLC was performed on a Thermo

Scientific system using an Altima C18 5M column (A - 0.1%TFA/H2O, B - AcCN, 5–

95% B in 20mins), UV3000 dual wave length detector and processed with Chromquest.

NMR was recorded on a Varian Oxford 300MHz instrument and processed with Glide, multiplicity (s = singlet, d = doublet, dd = doublet of doublets, dt = doublet of triplets, t = triplet, q = quadruplet, m = multiplet). MS data were obtained using a Waters Micromass

ZQ 4000 instrument and processed with Masslynx 4.0. High-Resolution MS of all tested compounds were obtained with Micromass Q-TOF-2 at the R. Marshall Wilson Mass

Spectrometry Facility, Department of Chemistry - University of Cincinnati and processed with MassLynx 4.0. Preparative RP HPLC was performed on a Varian Prostar system with Polaris C18-A 10μ column and UV detector (A = 0.1% TFA/H2O, B = CH3CN; 5-

100% B over 44 m; Flow Rate = 60 mL/min). Fractions from RP HPLC were combined and lyophilized using a Labconco Freezone® 6 freeze dry system; products were lyophilized at least twice to remove excess TFA before using for biological evaluation.

Flash column chromatography was performed with either Biotage Flash+™ semi- automated or Biotage SP1 automated purification systems. Melting points were

114

determined using a Fisher-Johns Melting Point Apparatus and are uncorrected. All tested compounds were >98% pure except for 39 which was 90% pure.

2-(4-Fluoro-2-methyl-phenoxymethyl)-6-morpholin-4-yl-isonicotinic acid (26): A mixture of 134mg (.36mmol, 1eq) 27, 9ml EtOH, 6ml H2O and 60mg (1.074mmol, 3eq)

KOH were refluxed for 2h. Next the EtOH was evaporated under vacuum and 0.1N HCl was added to the solution until precipitation stopped. The solid was filtered, washed with

H2O (2 x 15ml) and recrystallized from MeOH/H2O to yield 93mg (69%) of 26 as an off-

1 white solid. HNMR (CD3OD)  7.37 (s, 1H),  7.25 (s, 1H),  6.95 – 6.85 (m, 3H), 

5.08 (s, 2H),  3.82 (t, 4H),  3.59 (t, 4H),  2.33 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 347.1402, observed 347.1422. Mp = 168- 171°C.

2-(4-Fluoro-2-methyl-phenoxymethyl)-6-morpholin-4-yl-isonicotinic acid ethyl ester

(27): 174mg (.63mmol, 1eq) of 34 and excess morpholine (2.1ml, 24mmol, 38.5eq) were added to 5ml microwave reaction vessel. The reaction mixture was heated at 250°C for

75min to provide the carboxamide 37. The mixture was transferred to a round bottom flask and morpholine was evaporated under vacuum. Next 0.5ml H2SO4 and 10ml EtOH were added and the mixture refluxed for 48h.The EtOH was evaporated and 25ml

5%NaHCO3 were added to yield an off-white precipitate. The precipitate was filtered, washed with H2O (2 x 25ml) and recrystallized from MeOH/H2O to provide 190mg

1 (82%) of 27. H NMR (CDCl3)  7.46 (s, 1H),  7.23 (s, 1H),  6.94 – 6.83 (m, 3H), 

5.18 (s, 2H),  4.43 (q, 2H),  3.88 (t, 4H),  3.69 (t, 4H),  2.35 (s, 3H),  1.42 (t, 3H).

ESI-HRMS (M+H)+ m/z calculated 375.1714 observed 375.1719. Mp = 117-119°C.

2- [(4-Fluoro-2-methyl-phenylamino)-methyl] -6-morpholin-4-yl-isonicotinic acid

(28): Obtained by alkaline hydrolysis of 29 using the same procedure as 26. Yield was

115

1 42mg (56%). HNMR (CD3OD)  7.22 (s, 1H),  7.19 (s, 1H),  6.84 – 6.69 (m, 2H), 

6.49 (q, 1H),  4.37 (s, 2H),  3.83 (t, 4H),  3.59 (t, 4H),  2.25 (s, 3H). ESI-HRMS

(M+H)+ m/z calculated 346.1562 observed 346.1563. Mp = 86-88°C.

2- [(4-Fluoro-2-methyl-phenylamino)-methyl] -6-morpholin-4-yl-isonicotinic acid ethyl ester (29): Synthesized from 35 using the same procedure as 27. Yield was 81mg

1 (78%). H NMR (CDCl3)  7.22 (s, 1H),  7.23 (s, 1H),  6.87 – 6.80 (m, 2H),  6.58 (q,

1H),  4.45 – 4.37 (m, 4H),  3.88 (t, 4H),  3.65 (t, 4H),  2.26 (s, 3H),  1.42 (t, 3H).

ESI-HRMS (M+H)+ m/z calculated 374.1874 observed 374.1880. Mp = 78-80°C.

2-Morpholin-4-yl-isonicotinic acid (30): Obtained by alkaline hydrolysis of 31 using the

1 same procedure as 26. Yield was 60mg (26%). HNMR (CD3OD)  8.26 (d, 1H),  7.36

(s, 1H),  7.22 (d, 1H),  3.83 (t, 4H),  3.56 (t, 4H). ESI-HRMS (M+H)+ m/z calculated

209.0921 observed 209.0926. Mp = 53-55°C.

2-Morpholin-4-yl-isonicotinic acid ethyl ester (31): 100mg (.64mmol, 1eq) of 2- chloroisonicotinic acid and excess morpholine (1.5ml, 17mmol, 27eq) were placed into a

2ml microwave reaction vessel. The reaction mixture was heated at 215°C for 15min to provide the carboxamide 36. The mixture was transferred to a round bottom flask and the morpholine was evaporated under vacuum. Next 0.3ml H2SO4 and 5ml EtOH were added and the mixture refluxed for 48h.The EtOH was evaporated and 25ml

10%NaHCO3 were added to yield a white precipitate. The precipitate was filtered, washed with H2O (2 x 25ml) and recrystallized from MeOH/H2O to yield 21mg (13%) of

1 31 as needle shaped crystals. HNMR (CD3Cl3)  8.34 (s, 1H),  7.22 (s, 2H),  4.44 (q,

2H),  3.87 (t, 4H),  3.65 (t, 4H),  1.43 (t, 3H). ESI-HRMS (M+H)+ m/z calculated

237.1234 observed 237.1234. Mp = 187-189°C.

116

2-Chloro-6-methyl-Isonicotinic acid ethyl ester (32): 171.58mg (1mmol) of 2-chloro-

6-methyl isonicotinic acid, 10ml of ethanol and 500l of conc.H2SO4 were refluxed for

12h. The solution was cooled to room temperature and 25ml 5%NaHCO3 was added to provide a white precipitate which was extracted into EtOAc (3 x 10ml). The combined

EtOAc layers were washed with brine and dried using anhydrous MgSO4. The EtOAc layer was evaporated and 192.7mg (90%) of 32 were obtained as a white powder. 1H

NMR (CDCl3)  7.94 (s, 1H),  7.84 (s, 1H),  4.57 (s, 2H),  4.42 (q, 2H),  1.43 (t,

3H). ESI-MS (M+H)+ m/z calculated. 200.64, observed 200.40.

2-Bromomethyl-6-chloro-isonicotoinic acid ethyl ester (33): In a 250ml round bottom flask, 1.9g (9.32mmol, 1eq) of 32 and 30ml of CCl4 were placed. To this solution 2.5g

(13.92mmol, 1.5eq) N-bromosuccinamide (NBS) and 153mg (0.932mmol, 0.1eq) of azobisisobutyronitrile (AIBN) were added and the mixture refluxed for 4h after which another portion of NBS/AIBN was added. After a total of 8h, the reaction mixture was cooled to room temperature and filtered to remove the succinamide. The filtrate was then washed with 5%NaHCO3 (2 x 15ml), then brine (15ml) and dried with anhydrous

MgSO4 to obtain yellowish oil (~ 3g) which was dried under vacuum overnight. Under a nitrogen atmosphere, the oil was dissolved in 15ml anhydrous THF with constant stirring.

The solution cooled to 0-4°C by placing in an ice water bath for 30min. To the cold solution, 4eq (with respect to the dibromide in mixture) of diethylphosphite (2.9ml,

22.34mmol) and diisopropylethylamine (3.9ml, 22.34mmol) were added via syringe.

After addition the ice bath was removed and the mixture stirred for 30min. The mixture was poured into crushed ice/ water and extracted with EtOAc (3 x 30ml). The EtOAc layers were combined, washed with brine (30ml) and dried with MgSO4 to yield a dark

117

brown oil. The crude mixture was purified by column chromatography with

Hexane/CHCl3 (2:1  1:1, step gradient) to provide 1.9g (73%) of 33 as a light yellow

1 oil. H NMR (CDCl3)  7.94 (s, 1H),  7.84 (s, 1H),  4.57 (s, 2H),  4.47 (q, 2H),  1.45

(t, 3H). ESI-MS (M+H)+ m/z calculated 279.53, observed 279.96.

2-Chloro-6-(4-fluoro-2-methyl-phenoxymethyl)-isonicotinic acid ethyl ester (34):

Under an N2 atmosphere 45mg (.352mmol, 1.1eq) of 4-F-2-MePhOH and 89mg K2CO3

(.64mmol, 2eq) were added to 90mg (.32mmol, 1eq) of 33 dissolved in 5ml anhydrous

DMF. After 24h, 50ml of H2O were added to the reaction mixture and a yellow precipitate was obtained. The precipitate was filtered, washed with H2O (2 x 25ml) and

1 dried under high vacuum to yield 89mg (85%) of 34. H NMR (CDCl3)  8.05 (s, 1H), 

7.85 (s, 1H),  6.96 – 6.75 (m, 3H),  5.20 (s, 2H),  4.46 (q, 2H),  2.36 (s, 3H)  1.44 (t,

3H). ESI-MS (M+H) + m/z calculated 324.75, observed. 324.06.

2-Chloro-6-[(4-fluoro-2-methyl-phenylamino)-methyl]-isonicotinic acid ethyl ester

1 (35): Same procedure as 34. Yield was 245mg (90%). H NMR (CDCl3)  7.92 (s, 1H), 

7.81 (s, 1H),  6.92 – 6.81 (m, 3H),  4.59 (s, 2H),  4.44 (q, 2H),  2.38 (s, 3H)  1.42 (t,

3H). ESI-MS (M+H) + m/z calculated 323.76, observed. 323.00.

2-Chloro-6-[(4-fluoro-2-methyl-phenylamino)-methyl]-isonicotinic acid (35a):

Obtained by alkaline hydrolysis of 35 using the same procedure as 26. Yield was 245mg

1 (90%). H NMR (CD3OD)  7.85 (s, 1H),  7.77 (s, 1H),  6.84 (d, 1H),  6.70 (t, 1H), 

6.36 (q, 1H),  4.52 (s, 2H),  2.27 (s, 3H). ESI-HRMS (M+H) + m/z calculated

295.0645, observed. 295.0644.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-morpholin-4-yl-pyridine-4- carbaldehyde (39). To a solution of 150mg (0.45mmol, 1eq) of the alcohol 40 in 5ml

118

DCM were added 1ml (212mg, 0.5mmol, 1.1eq) of 15% solution of Dess-Martin

Periodinane in DCM with constant stirring. After 20m the reaction mixture was diluted with 25ml ether and poured into 10ml 1.3M NaOH. After stirring for 20m the ether layer was extracted with 10ml 1.3M NaOH and 10ml H2O. The ether layer was finally washed with brine, dried with MgSO4 and evaporated under vacuum to provide an orange solid.

This solid was triturated with MeOH and dried under vacuum to yield 44mg (30%) of 39.

1 H NMR (CDCl3).  9.98 (s, 1H),  7.08 (s, 1H),  6.95 (s, 1H),  6.85 (m, 2H),  6.49

(q, 1H),  4.41 (s, 2H),  3.87 (t, 4H),  3.65 (t, 4H),  2.26 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 330.1612 observed 330.1613. Mp = decomposed at room temperature.

{2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-morpholin-4-yl-pyridin-4-yl}- methanol (40). 2.4ml (95mg, 1.9mmol, 6eq) of 1M LiAlH4/Et2O was added in three portions over 4h to a solution of 140mg (0.371mmol, 1eq) of the ester 29 in 7ml anhydrous THF maintained under argon atm. The reaction mixture was diluted with 15ml ether and cooled to 0-4°C in an ice bath. To this cold solution were added 95l H2O,

95l 15% NaOH followed by additional 285l H2O. The reaction mixture was warmed to room temperature and stirred for 15min. Next small amount of anhydrous MgSO4 was added to the reaction mixture and stirred for another 15min. The mixture was filtered and the filtrate evaporated under vacuum to provide a yellow solid. The solid was recrystallized from Hex/EtOAc (2:1) to yield 47mg (38%) of 40 as yellow needle shaped

1 crystals. H NMR (CDCl3).  6.87-6.83(m, 3H),  6.67 (s, 1H),  6.60 (s, 1H),  6.53 (q,

1H),  4.68 (s, 2H),  4.32 (s, 2H),  3.87 (t, 4H),  3.59 (t, 4H),  2.52 (s, 3H). ESI-

HRMS (M+H)+ m/z calculated 332.1769 observed 332.1778. Mp = 139-141°C.

119

2-(4-Fluoro-2-methyl-phenoxymethyl)-6-morpholin-4-yl-isonicotinamide (41). To

400mg (1.068mmol, 1eq) of 27 placed in a pressure tube were added 30ml of

10%NH4OH/MeOH. The tube was sealed with Teflon stopper and heated at 60°C in a glycerol bath. After 48h the tube was cooled to room temperature during which the product precipitated out. The precipitate was filtered, washed with 10% H2O/MeOH and air dried. The solid was recrystallized from acetone to yield 156mg (30%) of 41 as white solid. 1H NMR (DMSO).  8.14 (s, 1H),  7.6 (s, 1H),  7.18 (dd, 1H),  7.08 (dd, 1H), 

6.96 (dd, 1H),  5.04 (s, 2H),  3.73 (t, 4H),  3.53 (t, 4H),  2.62 (s, 3H). ESI-HRMS

(M+H)+ m/z calculated 346.1562 observed 346.1552. Mp = 235-237°C.

2-(4-Fluoro-2-methyl-phenoxymethyl)-N-hydroxy-6-morpholin-4-yl-isonicotinamide

(42). To a mixture of 120mg (0.346mmol, 1eq) of 26 in 12ml Et2O placed in an ice bath were added 40l (0.415mmol, 1.2eq) ethyl chloroformate and 49.5l (0.45mmol, 1.3eq)

N-methyl morpholine and stirred for 30min. The suspension was filtered to provide a solution of the mixed anhydride. Simultaneously in another flask a solution of 72mg

(1.04mmol, 1.5eq) NH2OH.HCl in 1ml MeOH was neutralized by addition to a cold solution of 58.4 mg (1.04mmol, 1.5eq) KOH in 2ml MeOH and stirring for 30min. The precipitate was filtered to remove the KCl salt and the filtrate was added to the above mixed anhydride solution and stirred for 30min at room temperature. The reaction mixture was evaporated under vacuum, the crude product was triturated with MeOH and

THF and recrystallized from acetone to yield 38mg (24%) of 42 as a white solid. 1H

NMR (CD3OD)  7.09 – 6.94 (m, 5H),  5.04 (s, 2H),  3.71 (t, 4H),  3.52 (t, 4H), 

2.26 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 362.1511 observed 362.1510. Mp =

209-211°C.

120

2-(4-Fluoro-2-methyl-phenoxymethyl)-N-(2-hydroxy-ethyl)-6-morpholin-4-yl- isonicotinamide (43). To the mixed anhydride of 26 (prepared using the same procedure as 42), a solution of 16l (0.261mmol, 1.5eq) ethanolamine in 1.5ml MeOH was added and the mixture stirred for 15min. The solvent was evaporated under vacuum and the product recrystallized from Hex/EtOAc (1:1) to provide 33mg (50%) of 43. 1H NMR

(CD3OD).  7.22 (s, 1H),  7.09 (s, 1H),  6.95-6.80 (m, 3H),  5.06 (s, 2H),  3.83 (t,

4H),  3.72 (t, 2H),  3.60 (t, 4H),  3.53 (t, 2H),  2.32 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 390.1824 observed 390.1810. Mp = 179-181°C.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-[(2-hydroxy-ethyl)-methyl-amino]- isonicotinic acid (44). 200mg (0.62mmol, 1eq) 35 and 1ml (12.5mmol, 20eq) 2-

(methylamino) ethanol were added to a 2ml microwave reaction vessel and heated at

200°C for 30min to provide the bis-derivative (e.g. refer to 45a). The reaction mixture was then transferred to a round bottom flask and the amide hydrolyzed by refluxing with

56.11mg (1.24mmol, 2eq) KOH, 6ml EtOH and 9ml H2O for 2h. The crude mixture was purified by preparative RP HPLC and 118mg (57%) of 44 was obtained as TFA salt. 1H

NMR (CD3OD).  7.47 (s, 1H),  7.29 (s, 1H),  6.90 (dd, 1H),  6.79 (td, 1H),  6.57

(q, 1H),  4.56 (s, 2H),  3.86 (t, 4H),  3.75 (t, 2H),  3.33 (s, 3H),  2.32 (s, 3H). ESI-

HRMS (M+H)+ m/z calculated 334.1562 observed 334.1577. Mp = TFA salt was very hygroscopic.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-(2-hydroxy-ethylamino)-isonicotinic acid (45). Same procedure as 44. Yield was 112mg (33%). 1H NMR (DMSO).  6.97 (s,

1H),  6.89 (m, 2H),  6.80 (td, 1H),  6.30 (q, 1H),  4.27 (s, 2H),  3.57 (t, 2H),  3.41

121

(t, 2H),  2.19 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 320.1405 observed 320.1409.

Mp = TFA salt was very hygroscopic.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-N-(2-hydroxy-ethyl)-6-(2-hydroxy- ethylamino)-isonicotinamide (45a). Bis-derivative isolated during preparation of 45. 1H

NMR (CD3OD).  7.31 (s, 1H),  7.20 (s, 1H),  6.90 (d, 1H),  6.90 (m, 1H),  6.41 (q,

1H),  4.54 (s, 2H),  3.74 (dt, 4H),  3.54 (m, 4H),  2.30 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 363.1827 observed 363.1826. Mp = TFA salt was very hygroscopic.

2-[Bis-(2-hydroxy-ethyl)-amino]-6-[(4-fluoro-2-methyl-phenylamino)-methyl]- isonicotinic acid (46). Same procedure as 44. Yield was 44mg (20%). 1H NMR

(CD3OD).  7.62 (s, 1H),  7.34 (s, 1H),  6.90 (dd, 1H),  6.78 (td, 1H),  6.52 (q, 1H),

 4.54 (s, 2H),  3.85 (dd, 8H),  2.32 (s, 3H). ESI-HRMS (M+H)+ m/z calculated

364.1667 observed 364.1668. Mp = TFA salt was very hygroscopic.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-piperazin-1-yl-isonicotinic acid (47).

200mg (0.62mmol, 1eq) of 35, 267mg (3.1 mmol, 5eq) piperazine and 5ml 1,4-dioxane were added to 5ml microwave reaction vessel and heated normally at 85°C for 72h. The reaction mixture was then transferred to a round bottom flask and the ester was hydrolyzed by refluxing with 56.11mg (1.24mmol, 2eq) KOH, 6ml EtOH and 9ml H2O for 2h. The crude mixture was purified by preparative RP HPLC and 185mg (86.5%) of

1 47 was obtained as TFA salt. H NMR (CD3OD).  7.33 (s, 2H),  6.90 (dd, 1H),  6.79

(td, 1H),  6.61 (q, 1H),  4.47 (s, 2H),  3.91 (t, 4H),  3.34 (t, 4H),  2.31 (s, 3H). ESI-

HRMS (M+H)+ m/z calculated 345.1721 observed 345.1729. Mp = TFA salt was very hygroscopic.

122

6'-[(4-Fluoro-2-methyl-phenylamino)-methyl]-3,4,5,6-tetrahydro-2H-[1,2'] bipyridinyl-4'-carboxylic acid (48). 200mg (0.62mmol, 1eq) 35, 306l (3.1 mmol, 5eq) piperidine and 5ml 1,4-dioxane were added to 5ml microwave reaction vessel and heated normally at 85°C for 72h. The crude mixture was purified by column chromatography

(Hex/EtOAc, 11:1) and 213mg (92%) ethyl ester of 48 was obtained. Next 186mg

(0.5mmol, 1eq) of the ester were subjected to alkaline hydrolysis and the product was precipitated by neutralization with 0.1NHCl. The product was recrystallized from

1 MeOH/H2O to provide 114mg [53% over 2 steps from 35] of 48 as yellow solid. H

NMR (CD3OD).  7.17 (s, 1H),  7.11 (s, 1H),  6.83 (dd, 1H),  6.74 (td, 1H),  6.51

(q, 1H),  4.34 (s, 2H),  3.66 (t, 4H),  3.34 (t, 4H),  2.25 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 344.1769 observed 344.1764. Mp = 153-155°C.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-pyrrolidin-1-yl-isonicotinicacid (49).

Same procedure as 48. Ethyl ester of 49 was purified by column chromatography with

Hex/EtOAc (7:1). Yield of the ethyl ester was 95mg (43%). Yield of 49 was 88mg [43%

1 over 2 steps from 35]. H NMR (CD3OD) 7.09 (s, 1H),  6.93 (s, 1H),  6.83 (dd, 1H),

 6.74 (td, 1H),  6.65 (q, 1H),  4.36 (s, 2H),  3.54 (t, 4H),  2.25 (s, 3H), 2.08 (t,

4H). ESI-HRMS (M+H)+ m/z calculated 330.1612 observed 330.1612. Mp = 93-95°C.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-(3-hydroxy-pyrrolidin-1-yl)-isonico- tinic acid (50) and ethyl ester (50a). To 100mg (0.31mmol, 1eq) 35 and 126ml

(1.55mmol, 5eq) 3-pyrrolidinol and 5ml of 1, 4 dioxane were heated at 85°C. After 24h, an additional portion of 3-pyrroldinol was added and the mixture heated for another 24h.

The dioxane was evaporated under vacuum and the reaction mixture subjected to alkaline hydrolysis. The crude mixture was purified by preparative RP HPLC to provide 68mg

123

1 (48%) of 50 and 20mg (10%) of 50a as TFA salts. (50) H NMR (CD3OD).  7.18 (s,

2H),  6.89 (dd, 1H),  6.77 (td, 1H),  6.54 (q, 1H),  4.63 (s, 1H),  4.49 (s, 2H),  3.75

(m, 4H),  2.29 (s, 3H), 2.18 (m, 2H). ESI-HRMS (M+H)+ m/z calculated 353.1562 observed 353.1568. Mp = TFA salt was very hygroscopic.

1 (50a) H NMR (CD3OD).  6.96 (dd, 2H),  6.78 (t, 2H),  6.65 (q, 1H),  4.63 (s, 1H), 

4.57 (s, 2H),  4.45 (s, 1H),  3.39 (s, 1H),  3.80-3.59 (m, 8H),  2.28 (s, 3H), 2.18-

1.95 (m, 4H). ESI-HRMS (M+H)+ m/z calculated 415.2140 observed 415.2138. Mp =

TFA salt was very hygroscopic.

6-[(4-Fluoro-2-methyl-phenylamino)-methyl]-[2,4']bipyridinyl-4-carboxylic acid

(51). Under an argon atmosphere 200mg (0.62mmol, 1eq) 35, 92mg (0.744mmol, 1.2eq)

4-pyridine boronic acid (C6H5BNO2) and 5mol% of Pd(dppf)Cl2:DCM (26mg,

0.031mmol, 0.05eq) were added to 5ml microwave reaction vessel. The vessel was sealed and 5ml of 1, 4-dioxane and 623ml of 2M Na2CO3 (132mg, 1.24mmol, 2eq) were added via syringe. The mixture was heated normally at 80°C for 24h under a constant N2 stream. Next the dioxane was evaporated under vacuum, the crude mixture purified by column chromatography (Hex/EtOAc, 2:11:1, step gradient) and 180mg (81%) ethyl ester of 51 was obtained. Next 155mg (0.43mmol, 1eq) of the ester were subjected to alkaline hydrolysis and the product was precipitated by neutralization with 0.1NHCl. The product was recrystallized from MeOH/H2O to provide 142mg [68% over 2 steps from

1 35] of 51. H NMR (CD3OD).  8.71 (s, 2H),  8.38 (s, 1H),  8.21 (d, 2H),  7.99 (s,

1H),  6.83 (dd, 1H),  6.71 (td, 1H),  6.47 (q, 1H),  4.65 (s, 2H),  2.30 (s, 3H). ESI-

HRMS (M+H)+ m/z calculated 338.1299 observed 338.1318. Mp = 186-188°C.

124

6- [(4-Fluoro-2-methyl-phenylamino)-methyl]- [2,3']bipyridinyl-4-carboxylic acid

(52). Same procedure as 51. Yield of the ethyl ester was 268mg (73%). Yield of 52 was

175mg (56% over 2 steps from 35). 1H NMR (DMSO).  9.31 (s, 1H),  8.66 (s, 1H), 

8.51 (d, 1H),  8.22 (s, 1H),  7.81 (s, 1H),  7.54 (s, 1H),  6.90 (dd, 1H),  6.73 (t, 1H),

 6.34 (q, 1H),  4.55 (s, 2H),  2.21 (s, 3H). ESI-HRMS (M+H)+ m/z calculated

338.1299 observed 338.1310. Mp = 252-254°C.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-(4-hydroxy-phenyl)-isonicotinic acid

(53). Same procedure as (51). Ethyl ester was purified by column chromatography with

Hex/EtOAc (4:1). Yield of the ethyl ester was 140mg (59%). Yield of 53 was 121mg

1 (55% over 2 steps from 35). H NMR (CD3OD).  8.12 (s, 1H),  7.96 (d, 2H),  7.74 (s,

1H),  6.95 (dd, 2H),  6.90 (dd, 1H),  6.85 (td, 1H),  6.45 (q, 1H),  4.58 (s, 2H), 

2.30 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 353.1296 observed 353.1299. Mp =

122-124°C.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-(3-hydroxy-phenyl)-isonicotinic acid

(54). Same procedure as 51. Ethyl ester was purified by column chromatography with

Hex/EtOAc (4:1). Yield of ester was 213mg (90%). Yield of 54 was 103mg (47% over 2

1 steps from 35). H NMR (CD3OD).  8.15 (s, 1H),  7.84 (s, 1H),  7.54 (d, 2H), 

7.34(t, 1H),  6.95 (dd, 1H),  6.90 (dd, 1H),  6.85 (td, 1H),  6.45 (q, 1H),  4.59 (s,

2H),  2.30 (s, 3H). ESI-HRMS (M+H)+ m/z calculated 353.1296 observed 353.1303.

Mp = 156-158°C.

2-[(4-Fluoro-2-methyl-phenylamino)-methyl]-6-(2-hydroxy-phenyl)-isonicotinic acid

(55). Same procedure as 51. Ethyl ester was purified by column chromatography with

Hex/EtOAc (4:1). Yield of the ethyl ester was 60mg (25%). The carboxylic acid 55 was

125

purified by trituration with ether. Yield of 55 was 13mg (24% over 2 steps from 35). 1H

NMR (CD3OD).  8.36 (s, 1H),  7.98 (d, 1H),  7.82 (m, 1H),  7.35(t, 1H),  6.92 (m,

2H),  6.90 (dd, 1H),  6.85 (m, 1H),  6.45 (q, 1H),  4.61 (s, 2H),  2.26 (s, 3H). ESI-

HRMS (M+H)+ m/z calculated 353.1296 observed 353.1291. Mp = 227-229°C.

2-Morpholin-4-yl-6- [(4-nitro-phenylamino)-methyl] -isonicotinic acid (56). Obtained by alkaline hydrolysis of 56a. Same procedure as 26. Yield was 100mg (99%). 1H NMR

(CD3OD).  8.05 (d, 2H),  7.20 (d, 2H),  6.70 (d, 2H),  4.46 (s, 2H),  3.81 (t, 4H), 

3.58 (t, 4H). ESI-HRMS (M+H)+ m/z calculated 359.1350 observed 359.1369. Mp =

216-218°C.

2-Morpholin-4-yl-6-[(4-nitro-phenylamino)-methyl]-isonicotinic acid ethyl ester

(56a). 185mg (0.55mmol, 1eq) of 56b and 7ml (80mmol, 145eq) morpholine were added to a microwave vessel and heated at 150°C for 1h. The morpholine was evaporated under vacuum and the crude mixture dissolved in 10ml EtOAc. The EtOAc layer was washed with H2O (10ml X 2) then 10ml brine and dried with anhydrous MgSO4 to provide a yellow solid containing a reddish impurity. The solid was triturated with MeOH to remove the impurity and recrystallized from acetone to yield 113mg (53%) of 56a as

1 yellow solid. H NMR (CD3Cl3).  8.15 (d, 2H),  7.18 (d, 2H),  6.64 (d, 2H),  4.47 (s,

2H),  4.42 (q, 2H),  3.88 (t, 4H),  3.62 (t, 4H),  1.42 (t, 3H). ESI-MS (M+H)+ m/z calculated 387.40 observed 387.60. Mp = 179 -181°C.

2-Chloro-6-[(4-nitro-phenylamino)-methyl]-isonicotinic acid ethyl ester (56b). To a

30ml microwave vessel were added 570mg (2.05mmol, 1eq) 33, 340mg (2.46mmol,

1.2eq) p-nitroaniline, 373mg (2.57mmol, 1.25eq) KF/Celite (40% loading) and 10ml

CH3CN and heated at 200°C for 75min. The mixture was then filtered and the solvent

126

evaporated under vacuum. The crude mixture was purified by column chromatography using Hex/EtOAc (3:1) and the product recrystallized from MeOH to give 185mg (27%) of 56b. This product was only 86% pure as determined by HPLC analysis and was used as is in the preparation of 56a. The impurity was the bis-substituted (2 and 6-position) p-

1 nitroaniline product as determined by mass spectrometric analysis. H NMR (CD3Cl3). 

8.16 (d, 2H),  7.85 (d, 2H),  6.68 (d, 2H),  4.62 (d, 2H),  4.45 (q, 2H),  1.42 (t, 3H).

ESI-MS (M+H)+ m/z calculated 336.74 observed 336.49.

2-Morpholin-4-yl-6-[(3-nitro-phenylamino)-methyl]-isonicotinic acid (57). Obtained by alkaline hydrolysis of 57a using the same procedure as 26. Yield was 48mg (99%). 1H

NMR (CD3OD).  7.51 (t, 1H),  7.44 (dd, 1H),  7.31-7.19 (m, 3H),  7.01 (m, 1H), 

4.41 (s, 2H),  3.82 (t, 4H),  3.59 (t, 4H). ESI-HRMS (M+H)+ m/z calculated 359.1350 observed 359.1367. Mp = 204-206°C.

2-Morpholin-4-yl-6-[(3-nitro-phenylamino)-methyl]-isonicotinic acid ethyl ester

(57a). To a 30ml microwave vessel were added 304mg (1.09mmol, 1eq) of 33, 196mg

(1.42mmol, 1.3eq) m-nitroaniline, 317mg (2.18mmol, 2eq) KF/Celite (40% loading) and

15ml CH3CN and heated at 200°C for 75min. The mixture was then filtered and the solvent evaporated under vacuum. The crude mixture was added to 8ml morpholine placed in a microwave reaction vessel and heated at 150°C for 1h. The morpholine was rotavaped and the reaction mixture added to 10ml EtOAc. The EtOAc layer was washed with H2O (10mlx 2) then by 10ml brine and dried with anhydrous MgSO4. The EtOAc layer was evaporated under vacuum and the crude mixture purified by column chromatography [DCM DCM:EtOAc (1:0.04), step gradient] to yield 100mg (24%) of

1 57a. H NMR (CD3Cl3).  7.59-7.53 (m, 2H),  7.34 (dd, 1H),  7.21 (s, 1H),  7.15 (s,

127

1H),  6.98(dd, 1H),  4.42 (s, 2H),  4.39 (q, 2H),  3.88 (t, 4H),  3.63 (t, 4H),  1.42

(t, 3H). ESI-MS (M+H)+ m/z calculated 387.40 observed 387.73. Mp = 109-111°C.

2-Morpholin-4-yl -6- [(4-sulfamoyl-phenylamino)-methyl] -isonicotinic acid (58).

Obtained by alkaline hydrolysis of 58a using the same procedure as 26. Yield was 84mg

1 (89%). H NMR (CD3OD).  7.64 (d, 2H),  7.19 (d, 2H),  6.71 (d, 2H),  4.41 (s, 2H),

 3.82 (t, 4H),  3.58 (t, 4H). ESI-HRMS (M+H)+ m/z calculated 393.1227 observed

393.1227. Mp = 219-221°C.

2-Morpholin-4-yl-6-[(4-sulfamoyl-phenylamino)-methyl]-isonicotinic acid ethyl ester

(58a). Same procedure as 57a except KF/Celite was not used. Crude mixture was purified by column chromatography using Hex/EtOAc (2:11:1, step gradient). Yield was

1 119mg (16%). H NMR (CD3Cl3).  7.76 (d, 2H),  7.15 (d, 2H),  6.72 (d, 2H),  4.43

(s, 2H),  4.42 (q, 2H),  3.88 (t, 4H),  3.62 (t, 4H),  1.42 (t, 3H). ESI-MS (M+H)+ m/z calculated 421.48 observed 421.72. Mp = 175-178°C.

2-Morpholin-4-yl -6- [(3-sulfamoyl-phenylamino)-methyl] -isonicotinic acid (59).

Obtained by alkaline hydrolysis of 59a using the same procedure as 26. Yield was 75mg

1 (95%). H NMR (CD3OD).  7.28-7.12 (m, 5H),  6.84 (dd, 1H),  4.39 (s, 2H),  3.81

(t, 4H),  3.58 (t, 4H). ESI-HRMS (M+H)+ m/z calculated 393.1227 observed 393.1221.

Mp = 211-213°C.

2-Morpholin-4-yl-6-[(3-sulfamoyl-phenylamino)-methyl]-isonicotinic acid ethyl ester

(59a). Same procedure as 58a. Crude mixture was purified by column chromatography using Hex/EtOAc (2:11:1, step gradient) and recrystallized from MeOH to yield

1 115mg (23%) of 59a. H NMR (CD3OD).  7.28-7.12 (m, 5H),  6.84 (dd, 1H),  4.43 (s,

128

2H),  4.38 (q, 2H),  3.83 (t, 4H),  3.60 (t, 4H),  1.37 (t, 3H). ESI-MS (M+H)+ m/z calculated 421.48 observed 421.69. Mp = 195-198°C.

2-[(4-Amino-phenylamino)-methyl]-6-morpholin-4-yl-isonicotinic acid (60). 132mg

(0.342mmol, 1eq) 56a were added to 10ml of CH3COOH and heated at 50°C for 30min.

To aid the solubilization of 56a, an additional 10ml of CH3COOH was added and the mixture boiled for 20min. Next more 10ml of CH3COOH was added followed by 2ml

H2O and the mixture refluxed for 30min which provided complete dissolution of 56a.

Next the reaction mixture was cooled to 60°C and the round bottom flask was sealed with rubber cap. The air was carefully evacuated under vacuum and replaced with argon (2x).

Under the argon blanket were added 96mg (1.7mmol, 5eq) Fe powder and the reaction mixture stirred for 2h. The CH3COOH was evaporated under vacuum and 5% NaHCO3 added to the crude mixture. The product was extracted into EtOAc and washed with 0.1%

NaCN (to ensure complete removal of Fe) then by brine. Following drying with MgSO4, the EtOAc was evaporated under vacuum to provide 60a as a dark red oil which was hydrolyzed as mentioned for 26 and purified by preparative HPLC to yield 70mg (34%)

1 of 60. H NMR (CD3OD).  7.24 (d, 2H),  7.12 (d, 2H),  6.79 (d, 2H),  4.41 (s, 2H), 

3.83 (t, 4H),  3.59 (t, 4H). ESI-HRMS (M+H)+ m/z calculated 329.1608 observed

329.1598. Mp = TFA salt was very hygroscopic.

2-[(3-Amino-phenylamino)-methyl]-6-morpholin-4-yl-isonicotinic acid (61). 125mg

(0.323mmol, 1eq) 57a was added to round bottom flask followed by small amount of

DCM. The solution was then evaporated under vacuum which caused 57a to form a thin film on the wall of the flask. Next 10ml of CH3COOH was added and a clear yellow solution of 57a was obtained (the dissolution was aided probably by the DCM trapped

129

within the film of 57a). Next 500l of H2O was added and the round bottom flask evacuated of air and filled with argon (2x). The reaction mixture was stirred at 60°C for

2h. The reaction mixture was cooled to room temperature and filtered through a short

Celite path to remove solid particles. The consecutive steps of the synthesis are the same as 60a. Following alkaline hydrolysis of 61a, preparative HPLC provided 61mg (34%) of

1 61. H NMR (CD3OD). 7.23 (m, 3H),  6.78 (dd, 1H),  6.61 (dd, 2H),  4.41 (s, 2H), 

3.82 (t, 4H),  3.60 (t, 4H). ESI-HRMS (M+H)+ m/z calculated 329.1608 observed

329.1605. Mp = TFA salt was very hygroscopic.

2-[(4-Hydroxy-phenylamino)-methyl]-6-morpholin-4-yl-isonicotinic acid (62). 53mg

(0.13mmol, 1eq) 63, 14l (0.14mmol, 1.1eq), 16mg (30 wt %) Pd/C and 10ml MeOH were added to a round bottom flask. The flask was evacuated and filled with nitrogen

(3x). Next the flask was evacuated and filled with H2 (3x). After 2h stirring under H2, the reaction mixture was quickly filtered with 0.45M disc filter into a flask containing a few drops of acetic acid [as exposure of the product to atmospheric oxygen under basic conditions led to the formation of aldehyde 67]. The filtrate was purified by preparative

1 HPLC to yield 15mg (30%) of 62. H NMR (CD3OD).  7.31 (s, 1H),  7.24 (s, 1H), 

7.21 (d, 2H),  6.89 (dt, 2H),  4.56 (s, 2H),  3.83 (t, 4H),  3.66 (t, 4H). ESI-HRMS

(M+H)+ m/z calculated 330.1448 observed 330.1455. Mp = TFA salt was very hygroscopic.

2- [(4-Benzyloxy-phenylamino)-methyl] -6-morpholin-4-yl-isonicotinic acid (63).

Obtained by alkaline hydrolysis of 63a using the same procedure as 26 except the product

1 was recrystallized from MeOH. Yield was 27mg (40%). H NMR (CD3OD).  7.43-7.27

(m, 5H),  7.24 (s 1H),  7.17 (s 1H),  6.86 (dt, 2H),  6.72 (dt, 2H),  4.98 (s, 2H), 

130

4.31 (s, 2H),  3.81 (t, 4H),  3.57 (t, 4H). ESI-HRMS (M+H)+ m/z calculated 420.1918 observed 420.1934. Mp = 127-129°C.

2- [(4-Benzyloxy-phenylamino)-methyl] -6-morpholin-4-yl-isonicotinic acid ethyl ester (63a). Under N2 atmosphere 630mg (2.26mmol, 1eq) (33), 586mg (2.49mmol,

1.1eq) 4-Benzyloxyamine. HCl, 656mg (4.75mmol, 2.1eq) 33 and 15ml DMF were stirred for 24h. Next the reaction mixture was diluted with H2O and the precipitate extracted into EtOAc. The EtOAc layer was washed with brine, dried with MgSO4 and evaporated under vacuum. The obtained solid was added to a microwave reaction vessel followed by excess morpholine (8ml) and heated at 150°C for 30min. The morpholine was evaporated and the residue dissolved in 15ml EtOAc. The EtOAc was washed with

H2O (15ml x 2), followed by 15ml brine and dried with MgSO4. The crude mixture was purified by column chromatography using Hex/EtOAc (3:1) and recrystallized from

1 MeOH to yield 472mg (52%) of 63a as yellow solid. H NMR (CD3Cl3).  7.44-7.32 (m,

5H),  7.14 (d, 2H),  6.96 (m, 2H),  6.85 (dt, 2H),  5.01 (s, 2H),  4.39 (s, 2H),  4.31

(q, 2H),  3.86 (t, 4H),  3.66 (t, 4H),  1.40 (t, 3H). ESI-HRMS (M+H)+ m/z calculated

448.2231 observed 448.2251. Mp = 100-102°C.

2-{[Acetyl-(4-benzyloxy-phenyl)-amino]-methyl}-6-morpholin-4-yl-isonicotinic acid

(64). 100mg (0.24mmol, 1eq) 63a, 25l (0.26mmol, 1.2eq) acetic anhydride were dissolved in 2ml CHCl3 and stirred for 1h. The reaction mixture was diluted with 10ml

EtOAc and washed with 10ml 5% NaHCO3, 10ml brine and dried with MgSO4. The

EtOAc layer was evaporated to provide 111mg of 64a. This intermediate was dissolved in 4ml THF and 460l (0.46mmol, 2eq) of 1M LiOH added. This mixture was stirred for

3h which provided 64 in quantitative yields. Next the pH of the reaction mixture was

131

lowered to 3 using 0.1N HCl. The THF was evaporated under vacuum and the aqueous layer extracted with EtOAc (10ml x 2). The EtOAc layers were combined, washed with

10ml brine and dried with MgSO4. The EtOAc layer was evaporated under vacuum to provide 64 as oil. The oil was dissolved in 500l acetone in a round bottom flask and

10ml ether added followed by 20ml hexanes. The flask was sealed with parafilm to avoid moisture and allowed to stand for 5 days following which an off-white semisolid was obtained. Next the mother liquor was decanted and more ether was added to force the semi-solid to solidify. As no change was observed after 24h, the ether was rotavaped and the semi-solid was dried under high vacuum for 24h which provided ~15mg (15%) 64 as

1 a solid. H NMR (CD3Cl3).  7.434 (m, 5H),  7.27 (s 1H),  7.17 (t, 3H),  6.99 (d, 2H),

 5.07 (s, 4H),  3.81 (t, 4H),  3.57 (t, 4H),  2.02 (t, 3H). ESI-HRMS (M+H)+ m/z calculated 462.2024 observed 462.2041. Mp = 69-71°C.

2-Morpholin-4-yl-6-[(3-nitro-benzylamino)-methyl]-isonicotinic acid (65). The ethyl ester 65a was prepared using the same procedure as 63a and purified by column chromatography using Hex/EtOAc (1:11:2, step gradient). Alkaline hydrolysis of 65a followed by preparative HPLC provided 33mg (34%) of 65 as TFA salt. 1H NMR

(CD3OD).  8.46(t, 1H),  8.37(dd, 1H),  7.94 (d, 1H),  7.75 (t, 1H),  7.37 (s, 1H), 

7.22 (s, 1H),  4.49 (s, 2H),  4.44 (s, 2H),  3.83 (t, 4H),  3.66 (t, 4H). ESI-HRMS

(M+H)+ m/z calculated 373.1506 observed 373.1508. Mp = TFA salt was very hygroscopic.

Biological methods:

In vitro Lipid Kinase Assay. The PI3K (cat# 14-602) and  (cat# 14-558) were purchased from Millipore. The Kinase-Glo® Luminescent Kinase Assay kit (Cat#

132

V6713) was purchased from the Promega Corporation. The buffer for the assay contained

50mM HEPES HCl (pH 7.4), 5mM MgCl2.6H2O, 50mM NaCl, 0.05% CHAPS and

5mM DTT. The assay was performed in black 96 well plates. Following the literature provided with the kit, the assay was optimized for the amounts of ATP and PI3K substrate L--Phosphatidylinositol (PI) which was determined to be 1M and 10g/well respectively. Using these values the amount of PI3Kandneeded per well was determined (ranged between 175-400ng/well for different batches of enzymes). Each well consisted of 50l of kinase reaction mixture. For a typical assay, to each well were added

10l each of PI3K, 10g L--PI and inhibitor. The mixture was incubated for 30min at room temperature after which 20l of 2.5M ATP (final conc. was 1M) were added.

After incubating for 90min at room temperature the reaction was stopped by addition of

50l Kinase-Glo® reagent. Following 30min incubation, the luminescence was measured using a Polarstar Optima microplate reader. The IC50 values of the compounds were determined using the sigmoidal dose-response curve (variable slope, due to the large number of data points) with GraphPad Prism 5.01 for Windows, GraphPad Software, San

Diego California USA.

FL 5.12 Cell Assay: Cell culture and flow cytometry. All cell lines for the assay were maintained by the Plas lab. FL5.12 cells were cultured as described previously.146 For the assay the cells were deprived of the growth factor by washing the cells with RPMI buffer

(3X) followed by addition of PBS buffer devoid of IL-3. The cells were incubated at

37°C with test compounds (final concentrations of 10M or 50M) or vehicle (DMSO, neat). Cell viability was determined at 0, 24 and 48h by aliquoting 500l sample into test tube containing 1l propidium iodide (PI) and analyzing the cells for uptake of PI by

133

fluorescence-activated cell sorting (FACS) using BD FACSAria ™ Cell-Sorting System.

The graphs were generated with Microsoft Excel.

Plasmid Constructs and Retroviral Transductions. Constructs used to generate FL5.12 cells expressing wild type TelJak2 (MTJWT) and the kinase deficient form (MTJKD),

BSKS TelJak2WT and BSKS TelJak2KD (K882E) were generously provided by V

Penard-Lacronique from the Necker Children’s Hospital, Paris. The TelJak2 gene was excised and inserted into the MigR1 (MSCV-IRES-EGFP) vector containing a CMV promoter. Stable populations were generated by sorting the GFP expressing cells using the BD FACSAria.

Generation of FL5.12 cells expressing mAkt (2M10) and Bcl-xL (Bcl-xL) has been described previously.146 FL5.12 cells expressing the PTEN shRNA (D2.7) were generated using a pKD vector.149

Statistical analysis. The effect of the test compounds on cell viability was analyzed by one-way ANOVA. Multiple pairwise comparisons were performed using the Student-

Newman-Keuls Method. Effect of the test compounds on the percent viability of the cells was compared to the vehicle control and was considered to be statistically significant at p< 0.05. The analyses were performed using Sigmastat 3.5. (Systat Software, Chicago,

IL)

Computational methods:

Methods have been described in detail in Chapter II.

Docking of 63 with CAChe. In order to accommodate the longer 4-OBzl group, the radii of the p110 and  binding sites were increased from 5Å to 12Å. Compound 63 was initially placed in the binding site by superimposition with LY294002 as described in

134

Chapter II. Docking was performed using a two step process - initially the steric clashes were removed by using the flexible side chains - rigid ligand docking option. This allowed the side chains to relax and better accommodate the ligand. Next the ligand was docked using the flexible ligand - rigid side chains option. The lowest energy conformation was used for explaining the observed difference in potency of 63 on the

PI3K and  isoforms in the Lipid Kinase Assay.

135

VII. References

1] Fruman, D. A.; Meyers, R.A.; Cantley, L.C. Phosphoinositide Kinases. Annu. Rev.

Biochem.1998, 67, 481-507.

2] Vanhaesebroeck, B.; Leevers, S. J.; Khatereh, A.; Timms, J.; Katso, R.; Driscoll, P. C.;

Woscholski, R.; Parker, P. J.; Waterfield, M. D. Synthesis and function of 3-

phosphorylated inositol lipids. Annu. Rev. Biochem. 2001, 70, 535-602.

3] Hong, W.; Xu, Y.; Seet, L.; Hanson, B. The Phox (PX) domain, a new player in

phosphoinositides signaling. Biochem. J. 2001, 360, 513-530.

4] Kraub, M.; Haucke, V. Phosphoinositide-metabolizing enzymes at the interface

between membrane traffic and cell signaling. EMBO reports 2007, 8(3), 241-246.

5] Takenawa, T.; Itoh, T. Membrane targeting and remodeling through phosphoinositide-

binding domains. IUBMB Life 2006, 58(5), 296 – 303.

6] Kouchi, Z.; Shikano, T.; Nakamura, Y.; Shirakawa, H.; Fukami, K.; Miyazaki, S. The

role of EF-hand domains and C2 domain in regulation of enzymatic activity of

Phospholipase CJ.Biol.Chem. 2005, 280(22), 21015–21021.

7] Stenmark, H.; Aasland, R. FYVE-finger proteins – effectors of an inositol lipid. J.Cell.

Sci. 1999, 112, 4175-4183.

8] Suire, S.; Coadwell, J.; Ferguson, G. J.; Davidson, K.; Stephens, L. A new G-

activated regulatory subunit of the type IB Phosphoinositide 3-Kinase p110. Curr.

Biol. 2005, 15, 566–570.

9] Voigt, P.; Dorner, M. B.; Schaefer, M. Characterization of p87PIKAP, a novel

regulatory subunit of Phosphoinositide 3-Kinase  that is highly expressed in heart

and interacts with PDE3B. J. Bio. Chem. 2006, 281(15), 9977-9986.

136

10] Huang, C.; Mandelker, D.; Schmidt-Kittler, O.; Samuels, Y.; Velculescu, V. E.;

Kinzler, K. W.; Vogelstein, B.; Gabelli, S. B.; Amzel, L. M. The structure of a human

p110α/p85α complex elucidates the effects of oncogenic PI3Kα mutations. Science

2007, 318, 1744.

11] Stokoe, D.; The phosphoinositides-3-kinase pathway and cancer. Expert Rev. Mol.

Med. 2005, 7(10), 1-22.

12] Rohrschneider, L. R.; Fuller, J. F.; Wolf, I.; Liu, Y.; Lucas, D. M. Structure, function,

and biology of SHIP proteins. Genes & Dev. 2000, 14, 505-520.

13] Sawyers, C. L.; Vivanco, I. The Phosphatidylinositol 3-kinase –Akt pathway in

human cancer. Nat. Rev. Cancer 2002, 2, 489-501.

14] Bi, L.; Okabe, I.; Bernard, D. J.; Wynshaw-Boris, A.; Nussbaum, R. L. Proliferative

defect and embryonic lethality in mice homozygous for a deletion in the p110

subunit of Phosphoinositide 3-Kinase. J. Bio. Chem. 1999, 274(16), 10963-10968.

15] Bi, L.; Okabe, I.; Bernard, D. J.; Wynshaw-Boris, A.; Nussbaum, R. L. Early

embryonic lethality in mice deficient in the p110 catalytic subunit of PI 3-kinase.

Mammalian Genome 2002, 13, 169–172.

16] Wymann M. P.; Zvelebil, M.; Laffargue, M.; Phosphoinositide 3-kinase signaling –

which way to target? Trends Pharm. Sci. 2003, 24(7), 366-376.

17] Welch H. C. E.; Coadwell, W. J.; Stephens, L. R.; Hawkins, P. T. Phosphoinositide

3-kinase-dependent activation of Rac. FEBS Lett. 2003, 546, 93-97.

18] Maffucci, T.; Falasca, M. Role of class II phosphoinositides 3-kinase in cell

signaling. Biochem. Soc. Trans. 2007, 35(2), 211-214.

137

19] Falasca, M.; Hughes, W. E.; Dominguez, V.; Sala, G.; Fostira, F.; Fang, M. Q.;

Cazzolli, R.; Shepherd, P. R.; James, D. E.; Maffucci, T. The Role of

Phosphoinositide 3-Kinase C2 in Insulin Signaling. J. Bio. Chem. 2007, 282(38),

28226 – 28236.

20] Backer, J. M. The regulation and function of Class III PI3Ks: novel role for Vps34.

Biochem. J. 2008, 410, 1-17.

21] Di Cristofano, A.; Pesce, B.; Cordon-Cardo, C.; Pandolfi, P. P. Pten is essential for

Embryonic development and tumour suppression. Nat. Genet. 1998, 19, 348-355.

22] (a) Velculescu, V. E; Vogelstein, B; Silliman, N; Ptak, J; Szabo, S; Yan, H; Gazdar,

A; Powell, S.M Riggins,G.J; Willson,J; Markowitz,S; Kinzler,K.W.; Victor E. High

frequency of mutations of the PIK3CA gene in human cancers; Science 2004, 304,

554. (b) Karakas, B.; Bachman, K. E.; Park, B. H. Mutations of the PIK3CA

oncogene in human cancers. Br. J. Cancer 2006, 94, 455-459.

23] Philip, A. J.; Campbell, I. G.; Leet, C.; Vincan, E.; Rockman, S. P.; Whitehead, R.

H.; Thomas, R. J. S.; Phillips, W. A. The Phosphatidylinositol 3’-kinase p85 gene is

an oncogene in human ovarian and colon tumors. Cancer Res. 2001, 61, 7426–7429.

24] Hosoi, Y.; Miyachi, H.; Matsumoto, Y.; Ikehata, H.; Komura, J.; Shii, K.; Zhao, H.

J.; Yoshida, M.; Takai, Y.; Yamada, S.; Suzuki, N.; Ono, T. A phosphatidylinositol

3-kinase inhibitor Wortmannin induces radioresistance DNA synthesis and sensitizes

cells to bleomycin and ionizing radiation. Int. J. Cancer 1998, 78, 642- 647.

25] Edwards, E.; Geng , L.; Tan, J.; Onishko, H.; Donnelly, E.; Hallahan, D. E.

Phosphatidylinositol3-kinase/Alt signaling in the response of vascular endothelium to

ionizing radiation. Cancer Res. 2002, 62, 4671-4677.

138

26] Naga Prasad, S. V.; Laporte, S. A.; Chamberlain, D.; Caron, M. G.; Barak, L.;

Rockman, H. A.; Phosphoinositide 3-kinase regulates 2-adrenergic receptor

endocytosis by AP-2 recruitment to the receptor/ -arrestin complex. J. Cell. Biol.

2002, 158(3), 563-575.

27] Michael A. Crackower et al. Regulation of myocardial contractility and cell size by

distinct PI3K-PTEN signaling pathways. Cell 2002, 110, 737-749.

28] Vecchione, C.; Patrucco, E.; Marino, G.; Barberis, L.; Poulet, R.; Aretini, A.; Maffei,

A.; Gentile, M. T.; Storto, M.; Azzolino, O.; Brancaccio, M.; Colussi, G. L.;

Bettarini, U.; Altruda, F.; Silengo, L.; Tarone, G.; Wymann, M. P.; Hirsh, E.;

Lembo. G. Protection from angiotensin II-mediated vasculotoxic and hypertensive

response in mice lacking PI3K. J. Exp. Med. 2005, 201(8), 1217-1228.

29] Chang, J. D.; Sukhova, G. K.; Libby, P.; Schvartz, E.; Lichtenstein, A. H.; Field, S.

J.; Kennedy, C.; Madhavarapu, S.; Luo, J.; Wu, D.; Cantley, L.C. Deletion of the

phosphoinositide 3-kinase p110 gene attenuates murine atherosclerosis. Proc. Natl.

Acad. Sci. USA. 2007, 104(19), 807-8082.

30] Selheim, F.; Holmsen, H.; Vassbotn, F. S.; PI 3-kinase signalling in platelets: the

significance of synergistic, autocrine stimulation. Platelets 2000, 11, 69-82.

31] Hirsh, E.; Bosco, O.; Tropel, P.; Laffargue, M.; Calvez, R.; Altruda, F.; Wymann, M.;

Montrucchio, G.; Resistance to thromboembolism in PI3K- deficient mice. FASEB

J. 2001, 15, 2019-2021.

32] Lian, L.; Wang, Y.; Draznin, J.; Eslin, D.; Bennett, J.S.; Poncz, M.; Wu, D.; Abrams,

C. S.; The relative role of PLC and PI3K in platelet activation. Blood 2005, 106(1),

110- 117.

139

33] Jackson S. P., Schoenwaelder S. M., Goncalves I., Nesbitt W. S.,Yap C. L., Wright

C. E. Kenche, V.; Anderson K. E.; Dopheide S. M.; Yuan Y.; Sturgeon S.e A.;

Prabaharan H.; Thompson P. E.; Smith G. D.; Shepherd P. R.; Daniele N.; Kulkarni

S.; Abbott B.; Saylik D.; Jones C.; Lu L.; Giuliano S.; Hughan S. C.; Angus J. A.;

Robertson A. D; Salem H. H. PI 3-kinase p110beta: a new target for antithrombotic

therapy. Nat. Med. 2005, 11, 507–514.

34] Cosemans, J. M. E. M.; Munnix, I. C. A.; Wetzker, R.; Heller, R.; Jackson, S. P.;

Heemskerk, J. W. M. Continuous signaling via PI3K isoforms  and  is required for

platelet ADP receptor function in dynamic thrombus stabilization. Blood 2006,

108(9), 3045-3052.

35] Hirsch, E.; Katanaev, V. L.; Garlanda, C.; Azzolino, O.; Pirola, L.; Silengo,L.;

Sozzani, S.; Mantovani, A.; Altruda, F.; Wymann, M. P. Central role for G protein-

coupled phosphoinositide 3-kinase  in inflammation. Science 2000, 287, 1049-1053.

36] Oh, S.; Zheng, T.; Bailey, M. L.; Barber, D. W.; Schroeder, J. T.; Kim, Y.; Zhu, Z.

Src homology 2 domain–containing inositol 5-phosphatase 1 deficiency leads to a

spontaneous allergic inflammation in the murine lung. J. Allergy Clin. Immunol.

2007, 119(1), 123-131.

37] Camps, M.; Reuckle, T.; Rommel, C.; et al. Blockade of PI3Ksuppresses joint

inflammation and damage in mouse models of rheumatoid arthritis. Nat. Med. 2005,

11(9), 936-943.

38] Okkenhaug, K.; Bilancio, A.; Farjot, G.; Priddle, H.; Sancho, S.; Peskett E., Pearce,

W.; Meek, S. E.; Salpekar, A.; Waterfield, M. D.; Smith, A. J.; Vanhaesebroeck, B.

140

Impaired B and T cell antigen receptor signaling in p110delta PI 3-kinase mutant

mice. Science 2002, 297(5583), 1031-1034.

39] Ward, S. G.; Finan, P. Isoform-specific phosphoinositide 3-kinase inhibitors as

therapeutic agents. Curr. Opp. Pharmacol. 2003, 3, 426-434.

40] Sasaki, T.; Irie-Sasaki, J.; Jones, R. G.; Oliveira-dos-Santos, A. J.; Stanford, W. L.;

Bolon, B.; Wakeham, A.; Itie, A.; Bouchard, D.; Kozieradzki, I.; Joza, N.; Mak, T.

W.; Ohashi, P. S.; Suzuki, A.; Penninger, J. M. Function of PI3Kgamma in

thymocyte development, T cell activation and neutrophil migration. Science 2000,

287, 1040-1044.

41] Powis, G.; Bonjouklian, R.; Berggren, M. M.; Gallegos, A.; Abraham, R.; Ashendel,

C.; Zalkow, L.; Matter, W. F.; Dodge, J.; Grindey, G.; Vlahos, C. J. Wortmannin, a

potent and selective inhibitor of phosphatidylinositol-3-kinase. Cancer Res. 1994,

54, 2419-2423.

42] Norman, B. H.; Shih, C.; Toth, J. E.; Ray, J. E.; Dodge, J. A.; Johnson, D. W.;

Rutherford, P. G.; Schultz, R. M.; Worzalla, J. F.; Vlahos, C. J. Studies on the

mechanism of Phosphatidylinositol 3-Kinase inhibition by wortmannin and related

analogs. J. Med. Chem. 1996, 39, 1106-1111.

43] Wymann, M. P.; Bulgarelli-Leva, G.; Zvelebil, M. J.; Pirola, L.; Vanhaesebroek, B.;

Waterfield, M. D.; Panayotou, G. Wortmannin inactivates Phosphoinositide 3-Kinase

by covalent modification of Lys-802, a residue involved in the phosphate transfer

reaction. Mol. Cell. Biol. 1996, 1722–1733.

44] Ihle, N. T.; Williams, R.; Chow, S.; Chew, W.; Berggren, M. I.; Murrieta, G. P.;

Minion, D. J.; Halter, R. J.; Wipf, P.; Abraham, R.; Kirkpatrick, L.; Powis, G.

141

Molecular pharmacology and antitumor activity of PX-866, a novel inhibitor of

phosphoinositide-3-kinase signaling. Mol. Cancer Ther. 2004, 3(7), 763-772.

45] Marion, F.; Williams, D. E.; Patrick, B. O.; Hollander, I.; Mallon, R.; Kim, S. C.;

Roll, D. M.; Feldberg, L.; Van Soest, R.; Andersen, R. J. Liphagal, a selective

inhibitor of PI3Kinase isolated from the sponge Aka coralliphaga: structure

elucidation and biomimetic synthesis. Org. Lett. 2006, 8(2), 321-324.

46] Frojdo, S.; Cozzone, D.; Vidal, H.; Pirola, L.; Resveratrol is a class IA

phosphoinositide 3-kinase inhibitor. Biochem. J. 2007, 406, 511–518.

47] Vlahos, C. J.; Matter, W. F.; Hui, K. Y.; Brown, R. F. A specific inhibitor of

Phosphatidylinositol-3-kinase, 2-(4-Morpholinyl)-8-phenyl-4H-1-benzopyran-4-one

LY294002). J. Bio. Chem. 1994, 269(7), 5241-5248.

48] Robertson, A.; Jackson, S.; Kenche, V.; Yaip, C.; Parbaharan, H.; Thompson, P.

Therapeutic Morpholino-Substituted Compounds.International Patent Application.

WO 01/53266 A1, 2001.

49] Sadhu, C; et al.. Quinazoline derivatives as inhibitors of human phospatidylinositol 3-

kinase delta. International Patent Application WO 0181346 A2, 2001.

50] Hayakawa, M.; Kaizawa, H.; Kawaguchi, K.; Ishikawa, N.; Koizumi, T.; Ohishi, T.;

Yamano, M.; Okada, M.; Ohta, M.; Tsukamoto, S.; Raynaud, F. I.; Waterfield, M. D.;

Parker, P.; Workman, P. Synthesis and biological evaluation of imidazo[1,2- a]

pyridine derivatives as novel PI3 kinase p110 inhibitors. Bioorg. Med. Chem. 2007,

15, 403–412.

51] Knight, Z. A.; Gonzalez, B.; Feldman, M. E.; Zunder, E. R.; Goldenberg, D. D.;

Williams, O.; Loewith, R.; Stokoe, D.; Balla, A.; Toth, B.; Balla, T.; Weiss, W. A.;

142

Williams, R. L.; Shokat, K. M. A pharmacological map of the PI3-K family defines a

role for p110 in insulin signaling. Cell 2006, 125, 733–747.

52] Wymann, M. P.; Schneiter, R.; Lipid signalling in disease. Nature Rev. Mol. Cell

Biol., 2008, 9, 162-176.

53] http://www.nature.com/nrm/posters/lipidsignalling-disease/lipidsignalling.pdf (Last

accessed 01/21/09)

54] http://www.novartis.com/downloads/research/key-project-therapeutic.pdf (Last

accessed 04/16/09)

55] Bonnans, C.; Fukunaga, K.; Keledjian, R.; Petasis, N. A.; Levy, B. D.; Regulation of

Phosphatidylinositol 3–kinase by polyisoprenyl phosphates in neutrophil-mediated

tissue injury. J. Exp. Med. 2006, 203(4), 857- 863.

56] Maronea, R.; Cmiljanovic, V.; Giese, B.; Wymann, M. P. Targeting phosphoinositide

3-kinase—Moving towards therapy. Biochim. Biophys. Acta 2008, 1784,159–185.

57] http://clinicaltrials.gov/ (Last accessed 04/16/09)

58] Fan, Q-W; Knight, Z. A.; Goldenberg, D. D.; Yu, W.; Mostov, K. E.; Stokoe, D.;

Shokat, K. M.; Weiss, W. A. A dual PI3 kinase/mTOR inhibitor reveals emergent

efficacy in glioma. Cancer Cell 2006, 9(5), 341-349.

59] Fan, Q.W.; Cheng, C. K.; Nicolaides, T. P.; Hackett, C. S.; Knight, Z. A.; Shokat, K.

M.; Weiss W. A. A dual phosphoinositide-3-kinase alpha/mTOR inhibitor cooperates

with blockade of epidermal growth factor receptor in PTEN-mutant glioma. Cancer

Res. 2007, 67(17), 7960-7965.

60] Serra, V.; Markman, B.; Scaltriti, M.; Eichhorn, P. J.; Valero, V.; Guzman, M.;

Botero, M. L.; Llonch, E.; Atzori, F.; Di Cosimo, S.; Maira. M.; Garcia-Echeverria,

143

C.; Parra J. L.; Arribas, J.; Baselga, J. NVP-BEZ235, a dual PI3K/mTOR inhibitor,

prevents PI3K signaling and inhibits the growth of cancer cells with activating PI3K

mutations. Cancer Res. 2008, 68(19), 8022-30.

61] Folkes, A. J.; Ahmadi,K.; Alderton,W.K.; Alix, S.; Baker,S. J.; Box, G.;

Chuckowree, I. S.; Clarke, P. A.; Depledge,P.; Eccles, S. A.; Friedman, L. S.; Hayes,

A.; Hancox, T. C.; Kugendradas, A.; Lensun, L.; Moore, P.; Olivero, A. G.; Pang, J.;

Patel, S.; Pergl-Wilson, G. H.; Raynaud, F. I.; Robson, A.; Saghir, N.; Salphati, L.;

Sohal, S.; Ultsch, M. H.; Valenti, M.;. Wallweber, H. J. A.; Wan, N. C.; Wiesmann,

C.; Workman, P. Zhyvoloup, A.; Zvelebil, M. J.; Shuttleworth, S. J. The

Identification of 2-(1H-Indazol-4-yl)-6-(4-methanesulfonyl-piperazin-1-ylmethyl)-4-

morpholin-4-yl-thieno[3,2-d]pyrimidine (GDC-0941) as a potent, selective, orally

bioavailable inhibitor of class I PI3 kinase for the treatment of cancer. J. Med.

Chem.2008, 51(18), 5522-5532.

62] Walker, E. H.; Perisic, O.; Ried, R.; Stephens, L.; Williams, R. L. Structural insights

into phosphoinositide 3-kinase catalysis and signaling. Nature 1999, 402, 313-320.

63] Walker, E. H.; Pacold, M. E.; Perisic, O.; Ried, R.; Stephens, L.; Williams, R. L.

Structural Determinants of Phosphoinositide 3-Kinase Inhibition by Wortmannin,

LY294002, Quercetin, Myricetin, and Staurosporine. Mol. Cell 2000, 6, 909–919.

64] http://www.rcsb.org/pdb/statistics/holdings.do (Last accessed 01/16/2009)

65] http://www.expasy.ch/sprot/ (Last accessed 01/16/2009)

66] http://predictioncenter.org/index.cgi (Last accessed 01/16/2009)

67] Schwede, T.; Sali, A.; Eswar, N.and Peitsch, M. C. in Computational Structural

Biology: Methods and Applications; Schwede, T. and Peitsch, M. C., Eds.; World

144

Scientific Publishing Company; 1 edition (September 2, 2008); Chapter I: Protein

Structure Prediction, pp 3 – 35.

68] Haber, E.; Anfinsen, C. B.; Side-chain interactions governing the pairing of half-

cystine residues in ribonuclease. J. Biol. Chem. 1962, 237(6), 1839-1844

69] Wooley, J. C. and Ye, Y. in Computational Methods for Protein Structure Prediction

and Modeling: Volume 1: Basic Characterization; Xu, Y.; Xu, D. and Liang, J. Eds.;

Springer New York; 2007; Chapter 1: A Historical Perspective and Overview of

Protein Structure Prediction, pp 1- 43.

70] Zhang, Y., and Skolnick, J.; Automated structure prediction of weakly homologous

proteins on a genomic scale. Proc. Natl. Acad. Sci. USA. 2004, 101, 7594–7599.

71] Klepeis, J. L.; Wei, Y.; Hecht, M. H.; and Floudas, C. A.; Ab Initio prediction of the

three-dimensional structure of a de Novo designed protein: A double-blind case study.

Proteins 2005, 58, 560–570.

72] Kihara,D.; Lu, H.; Kolinski, A.; and Skolnick, J. TOUCHSTONE: An ab initio

protein structure prediction method that uses threading-based tertiary restraints. Proc.

Natl. Acad. Sci. USA. 2001, 98, 10125–10130.

73] (a) Bonneau,R.; Tsai,J.; Ruczinski, I.; Chivian, D.; Rohl,C.; Strauss, C. E. M.; and

Baker, D.; Rosetta in CASP4: Progress in ab initio protein structure prediction.

Proteins 2001, Suppl 5, 119-126. (b) Rohl, C.A.; Strauss, C.E.; Misura, K.M.S.; &

Baker, D. Protein structure prediction using Rosetta. Methods Enzymol. 2004, 383,

66-93. (c) Bradley, P.; Misura, K.M.; & Baker, D. Toward high-resolution de novo

structure prediction for small proteins. Science 2005, 309, 1868-1871.

145

74] Govindarajan, S.; Recabarren,R.; and Goldstein, R. A.; Estimating the total number

of protein folds. Proteins 1999, 38, 408–414.

75] http://scop.mrc-lmb.cam.ac.uk/scop/count.html#scop-1.73 (Last accessed

01/16/2009).

76] Mizuguchi, K. Fold recognition for drug discovery. Drug Dis. Today: Targets 2004,

3(1), 18-23.

77] Altschul, S.F.; Madden, T. L.; Schäffer, A. A.; Zhang, J.; Zhang, Z.; Miller, W.; and

Lipman, D. J. Gapped BLAST and PSI-BLAST: a new generation of protein database

search programs. Nucleic Acids Res. 1997, 25, 3389–3402.

78] Rychlewski, L.; Jaroszewski, L.; Li, W. and Godzik, A. Comparison of sequence

profiles. Strategies for structural predictions using sequence information. Protein Sci.

2000, 9, 232-241.

79] Sadreyev, R. I.; Baker, D.; Grishin, N. V. Profile-profile comparisons by COMPASS

predict intricate homologies between protein families. Protein Sci. 2003, 12, 2262-

2272.

80] Edgar, R. C.; Sjolander, K. COACH: profile-profile alignment of protein families

using hidden Markov models. Bioinformatics 2004, 20(8), 1309-1318.

81] Jones, D.T.; Taylor, W. R.; Thornton, J. M. A new approach to protein fold

recognition. Nature 1992, 358, 86–89.

82] Jones,D.T. GenTHREADER: An efficient and reliable protein fold recognition

method for genomic sequences. J. Mol. Biol. 1999, 287, 797–815.

83] Xu,Y.; and Xu, D. Protein threading using PROSPECT: Design and evaluation.

Proteins 2000, 40, 343–354.

146

84] Xu, J.; Li, M.; Kim, D.; Xu, Y. RAPTOR: Optimal protein threading by linear

programming. J. Bioinform. Comput. Biol. 2003, 1, 95–117.

85] Kelley, L.A.; MacCalluma, R. M.; Sternberg, M. J. E. Enhanced genome annotation

using structural profiles in the program 3D- PSSM. J. Mol. Biol. 2000, 299, 499–520.

86] Shi, J.; Blundell, T. L. Mizuguchi, K. FUGUE: sequence-structure homology

recognition using environment-specific substitution tables and structure-dependent

gap penalties. J. Mol. Biol. 2001, 310, 243–257.

87] Höltje, H. D.; Sippl, W.; Rognan, D.; Folkers, G.; Molecular Modeling: Basic

Principles and Applications. Wiley-VCH; 2008, 3rd edition, pg 111-178.

88] Tramontano, A. Protein Structure Prediction: Concepts and Applications. Wiley-

VCH; 2006, pp. 73-109.

89] Pearson, W. R. Rapid and sensitive sequence comparison with FASTP and FASTA.

Methods Enzymol. 1990, 183, pp. 63-98.

90] Altschul S. F.; Gish W.; Miller W.; Myers E. W.; Lipman D. J. Basic local alignment

search tool. J. Mol. Biol. 1990, 215(3), 403-410.

91] Altschul, S. F.; Madden, T. L.; Schäffer, A. A.; Zhang, J.; Zhang. Z.; Miller, W.;

Lipman, D.J.; Gapped BLAST and PSI-BLAST: a new generation of protein database

search programs. Nucleic Acids Res. 1997, 25(17), 3389-3402.

92] Needleman, S. B.; Wunsch. C. D. A general method applicable to the search of

similarities in the amino acid sequence of two proteins. J. Mol. Biol. 1970, 48, 442-

453.

93] http://www.hku.hk/bruhk/gcg.html (Last Accesses 01/16/2009).

147

94] Thompson, J. D.; Higgins, D. G.; Gibson, T. J. CLUSTAL W: improving the

sensitivity of progressive multiple sequence alignment through sequence weighting,

position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 1994,

22(22), 4673–4680.

95] Thompson, J. D.; Gibson, T. J.; Plewniak, F.; Jeanmougin F.; and Higgins, D. G. The

CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment

aided by quality analysis tools. Nucleic Acids Res. 1997, 25(24), 4876-4882.

96] Sander, C.; Schneider, R. Database of homology-derived structures and the structural

meaning of sequence alignment. Proteins 1991, 9, 56-68.

97] M.O. Dayhoff; R.V. Eck, Atlas of Protein Sequence and Structure; M.O. Dayhoff and

R.V. Eck, Eds.; National Biomedical Research Foundation. 1968, A model of

evolutionary change in proteins; pp. 33–41.

98] Henikoff, S.; Henikoff, J. G. Amino acid substitution matrices from protein blocks.

Proc. Natl. Acad. Sci. U S A. 1992, 89(22), 10915–10919.

99] Canutescu, A. A.; Shelenko, A. A.; Dunbrack, R. L. Jr. A graph-theory algorithm for

rapid protein side-chain prediction. Protein Sci. 2003, 12, 2001-2014.

100] Desmet, J.; Demaeyer, M.; Hazes, B.; Lasters, I. The dead-end elimination theorem

and its use in protein side-chain positioning. Nature 1992, 356, 539–542.

101] Eswar, N.; John, B.; Mirkovic, N.; Fiser, A.; Ilyin, V. A.; Pieper, U.; Stuart, A. C.;

Marti-Renom, M. A.; Madhusudan, M. S.; Yerkovich, B.; Sali, A. Tools for

comparative protein structure modeling and analysis. Nucleic Acids Res. 2003,

31(13), 3375-3380.

102] http://swissmodel.expasy.org//course/text/chapter8.htm (Last Accessed 01/16/2009)

148

103] Laskowski R. A.; MacArthur M. W.; Moss D. S.; Thornton J. M.; PROCHECK: a

program to check the stereochemical quality of protein structures. J. Appl. Cryst.

1993, 26, 283-291.

104] Hooft, R. W. W.; Vriend, G.; Sander, C.; Abola, E. E.; Errors in protein structures.

Nature 1996, 381, 272.

105] Willard, L.; Ranjan, A.; Zhang, H.; Monzavi, H.; Boyko, R. F.; Sykes, B. D.

Wishart, D. S. VADAR: a web server for quantitative evaluation of protein structure

quality. Nucleic Acids Res. 2003, 31(13), 3316-3319.

106] Kabsch, W.; Sander, C. Dictionary of protein secondary structure: pattern

recognition of hydrogen-bonded and geometrical features. Biopolymers, 1983, 22,

2577-2637.

107] Hooft, R. W. W.; Sander, C. and Vriend, G. Objectively judging the quality of a

protein structure from the Ramachandran plot. Bioinformatics 1997, 13, 425-430.

108] (a) http://nihserver.mbi.ucla.edu/Verify_3D/ (Last Accesses 01/16/2009) (b) Bowie,

J.U.; Lüthy, R.; Eisenberg, D. A method to identify protein sequences that fold into a

known three-dimensional structure. Science 1991, 253(5016),164-70.

109] Wiederstein, M. and Sippl, M. J. ProSA-web: interactive web service for the

recognition of errors in three-dimensional structures of proteins. Nucleic Acids Res.

2007, 35, Web Server issue W407–W410.

110] http://www.molsoft.com/homology.html (Last Accessed 01/16/2009).

111] http://accelrys.com/products/discovery-studio/protein-modeling/index.html (Last

Accessed 01/16/2009).

112] http://www.tripos.com/ (Last Accessed 01/16/2009).

149

113] http://swift.cmbi.kun.nl/whatif/ (Last Accessed 01/16/2009).

114] http://modbase.compbio.ucsf.edu/ModWeb20-html/modweb.html (Last Accessed

01/16/2009).

115] http://zhang.bioinformatics.ku.edu/I-TASSER/ (Last Accessed 01/16/2009).

116] http://swissmodel.expasy.org/SWISS-MODEL.html (Last Accessed 01/16/2009).

117] http://www.fundp.ac.be/sciences/biologie/urbm/bioinfo/esypred/ (Last Accessed

01/16/2009).

118] http://www.sbg.bio.ic.ac.uk/~phyre/index.cgi (Last Accessed 01/16/2009).

119] http://meta.bioinfo.pl/submit_wizard.pl (Last Accessed 01/16/2009).

120] Fischer D: 3D-SHOTGUN: a novel, cooperative, fold recognition meta-predictor.

Proteins 2003, 51, 434-441.

121] http://robetta.bakerlab.org/ (Last Accessed 01/16/2009).

122] Sali, A.; Sanchez, R. Large-scale protein structure modeling of the Saccharomyces

cerevisiae genome. Proc. Natl. Acad. Sci. USA. 1998, 95, 13597–13602.

123] Schäffer A. A.; Wolf, Y. I.; Ponting, C.P.; Koonin, E. V.; Aravind, L.; Altshul, S. F.

IMPALA: matching a protein sequence against a collection of PSI-BLAST-

constructed position-specific score matrices. Bioinformatics 1999, 15(11), 1000-

1011.

124] Kleywegt, G. J.; Validation of protein crystal structures. Acta Crystallogr., Sect D

2000, 56(3), 249-265.

125] Acharya, K. R. and Lloyd, M. D. The advantages and limitations of protein crystal

structures. Trends Pharm. Sci. 2005, 26(1), 10-14.

150

126] (a) Burley, S. K.; Almo, S. C.; Bonanno, J. B.; Cape, M.; Chance, M. R.;

Gaasterland, T.; Lin, D.; Sali, A.; Studier, F.W.; Swaminathan, S. Structural

genomics: beyond the Human Genome Project. Nat. Genet. 1999, 23, 151-157. (b)

http://www.structuralgenomics.org/ (Last Accessed 01/16/2009).

127] Hillisch, A.; Pineda, L. P.; Hilgenfeld, R. Utility of homology models in the drug

discovery process. Drug Dis. Today 2004, 9(15), 659-669.

128] Hopkins, A. L.; Groom, C. R. The druggable genome. Nat. Rev. Drug Dis. 2002, 1,

727-730.

129] Mol, C. D.; Fabbro, D.and Hosfield, D, J. Structural insights into the conformational

selectivity of STI-571 and related kinase inhibitors. Curr. Opin. Drug Discov. Devel.

2004, 7(5), 639-48.

130] Honma, T.; Hayashi, K.; Aoyama, T.; Hashimoto, N.; Machida, T.; Fukasawa, K.;

Iwama, T.; Ikeura, C.; Ikuta, M.; Suzuki-Takahashi, I.; Iwasawa, Y.; Hayama, T.;

Nishimura, S.; and Morishima, H. Structure-Based Generation of a New Class of

Potent Cdk4 Inhibitors: New de Novo Design Strategy and Library Design. J. Med.

Chem. 2001, 44, 4615-4627.

131] Siedlecki, P.; Boy, R. G.; Comagic, S.; Schirrmacher, R.; Wiessler,

M.; Zielenkiewicz, P.; Suhai, S.; Lyko, F. Establishment and functional validation of

a structural homology model for human DNA methyltransferase 1. Biochem. Biophys.

Res. Commun. 2003, 306, 558–563.

132] Hardy, L. W. and Malikayil, A. The impact of structure-guided drug design on

clinical agents. Curr. Drug Discov. 2003, (Dec), 15-20.

151

133] Hillisch, A.; Peters, O.; Kosemund, D.; Müller, G.; Walter, A.; Schneider,

B.; Reddersen, G.; Elger, W.; Fritzemeier, K. H. Dissecting physiological roles of

estrogen receptor alpha and beta with potent selective ligands from structure based

design. Mol. Endocrinol. 2004, 18, 1599–1609.

134] http://molprobity.biochem.duke.edu/ (Last Accesses 01/16/2009).

135] http://www.yasara.org/products.htm#dynamics (Last Accesses 01/16/2009).

136] Morris, A. L.; MacArthur, M. W.; Hutchinson, E. G. and Thorton, J. M.

Stereochemical Quality of Protein Structure Coordinates. Proteins: Struct. Funct.

Genet. 1992, 12, 345–364.

137] http://pymol.sourceforge.net/ (Last Accesses 01/16/2009).

138] Knight, Z. A.; Chiang, G. G.; Alaimo, P. J.; Kenski, D. M.; Ho,C. B.; Coan, K.;

Abraham, R. T.; Shokat, K. M. Isoform-specific phosphoinositide 3-kinase inhibitors

from an arylmorpholine scaffold. Bioorg. Med. Chem. 2004, 12, 4749 – 4759.

139] Knight, Z. A.; Shokat, K. M. Chemically targeting the PI3K family. Biochem Soc

Trans. 2007, 35, 245- 249.

140] http://www.cache.fujitsu.com/biomedcache/ (Last Accesses 01/16/2009).

141] Xie, P.; Williams, D. S.; Atilla-Gokcumen, G. E.; Milk, L.; Xiao M.; Smalley, K. S.

M.; Herlyn, M.; Marmorstein, R.; Structure-based design of an organoruthenium

phosphatidyl-inositol-3-kinase inhibitor reveals a switch governing lipid kinase

potency and selectivity. ACS Chem. Biol. 2008, 3(5) 305–316.

142] www.schrodinger.com

143] Müller, K.; Faeh, C.; Diederich, F. Fluorine in Pharmaceuticals: looking beyond

intuition. Science 2007, 317, 1881-1886.

152

144] Liu, P.; Chen, Y.; Deng, J.; Tu Y.An efficient method for the preparation of

benzylic bromides. Synthesis 2001, 14, 2078 – 2080.

145] Goueli, S.; Hsaio K. Assay virtually any kinase with Kinase-Glo® Plus

Luminescent kinase assay, a homogenous high-throughput assay. Promega

Corporation Cell Notes, 2005, 12, 8-12.

146] Plas, D. R.; Talapatra, S.; Edinger, A. L.; Rathmell, J. C.; Thompson, C. B.; Akt and

Bcl-xL promote growth factor-independent survival through distinct effects on

mitochondrial physiology. J. Biol. Chem. 2001, 276(15), 12041–12048.

147] Vander Heiden, M. G.; Chandel, N. S.; Williamson, E. K.; Schumacker, P. T.;

Thompson, C. B. Bcl-xL regulates the membrane potential and volume homeostasis

of mitochondria. Cell 1997, 91, 627–637.

148] Vander Heiden, M. G.; Plas, D. R.; Rathmell, J. C.; Fox, C. J.; Harris, M. H.;

Thompson, C. B. Growth factors can influence cell growth and survival through

effects on glucose metabolism. Mol. Cell. Biol. 2001, 21(17), 5899–5912.

149] Edinger, A. L.; Cinali, R. M.; Thompson, C. B. Rab7 prevents growth factor

independent survival by inhibiting cell-autonomous nutrient transporter expression.

Dev. Cell 2003, 5, 571–582.

150] Nguyen, M. H.-H.; Ho, J. M.-Y.; Beattie, B. K.; Barber, D. L. TEL-JAK2 mediates

constitutive activation of the phosphatidylinositol 3’- kinase/protein kinase B

signaling pathway. J. Biol. Chem. 2001, 276(35), 32704-32713.

151] Pecquet, C.; Nyga, R.; Penard-Lacronique, V.; Smithgall, T. E.; Murakami, H.;

Régnier, A.; Lassoued, K.; Gouilleux, F. The Src tyrosine kinase Hck is required for

153

Tel-Abl- but not for Tel-Jak2-induced cell transformation. Oncogene 2007, 26, 1577–

1585.

152] Howard, N.; Abell,C.; Blakemore,W.; Chessari,G.; Congreve,M.; Howard, S.; Jhoti,

H.; Murray, C. W.; Seavers, L. C. A.; Van Montfort, R. L. M. Application of

fragment screening and fragment linking to the discovery of novel thrombin

inhibitors. J.Med. Chem. 2006, 49, 1346-1355.

153] Cha, F. S.; Chun, J. H.; Kim, J. M.; Lee, D. Y.; Cho, S. D. Preparation of aldehydes

from carboxylic esters by reductive oxidation with lithium aluminum hydride and

pyridinium chlorochromate or pyridinium dichromate. Bull. Korean Chem. Soc. 1999,

20 (11), 1373-1374.

154] Corey, E. J.; Suggs, J.W. Pyridinium chlorochromate. An efficient reagent for

oxidation of primary and secondary alcohols to carbonyl compounds. Tet. Lett. 1975,

31, 2647-2650.

155] Dess, D. B.; Martin, J. C. Readily accessible 12-I-51 oxidant for the conversion of

primary and secondary alcohols to aldehydes and ketones. J. Org. Chem. 1983, 48,

4155-4156.

156] Dougherty, D. A. Cation- interactions involving aromatic amino acids. J. Nutr.

2007, 137, 1504S–1508S.

157] Shi, Z.; Olson, C. A.; Kallenbach, N. R. Cation− interaction in model -helical

peptides. J. Am. Chem. Soc., 2002, 124 (13), 3284-3291.

158] Gallivan, J. P.; Dougherty, D. A. Cation- interactions in structural biology. Proc.

Natl. Acad. Sci. USA 1999, 96, 9459–9464.

154

159] Gharbi, S. I.; Zvelebil, M. J.; Shuttleworth, S. J.; Hancox, T.; Saghir, N.; Timms, J.

F.; Waterfiled, M. D. Exploring the specificity of the PI3K family inhibitor

LY294002. Biochem. J. 2007, 404, 15–21.

155