A Dissertation Entitled Structural and Enzymatic Studies of Essential

Home , Maltokinase, Mycothiol, Mycothiol synthase, Mycothione reductase, Quinaldine Red, Thioacetazone

A Dissertation

entitled

Structural and Enzymatic Studies of Essential Enzymes in Mycobacterium tuberculosis

Jared J. Lindenberger

Submitted to the Graduate Faculty as partial fulfillment of the requirements for the

Doctor of Philosophy Degree in Chemistry

______Dr. Donald R. Ronning, Committee Chair

______Dr. James T. Slama, Committee Member

______Dr. Steven J. Sucheck, Committee Member

______Dr. John J. Bellizzi, Committee Member

______Dr. Patricia R. Komuniecki, Dean College of Graduate Studies

The University of Toledo

August 2015

This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author.

An Abstract of

Structural and Enzymatic Studies of Essential Enzymes in Mycobacterium tuberculosis

Jared J. Lindenberger

Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Chemistry

The University of Toledo

August 2015

Tuberculosis (TB) continues to be a global health threat. The World Health

Organization estimates that nearly 8.6 million new TB cases were reported, and 1.3 million people succumbed to the disease in 2013. The persistence of TB globally is due in part to the hardiness of the bacterium Mycobacterium tuberculosis (M. tb), the etiological agent of TB, and the ability of M. tb to enter a dormant phase that complicates treatment. Typical therapeutic regimens to treat M. tb infections require at least 6 months of anti-tubercular drugs and patient non-compliance during this treatment period is thought to contribute to the selection of antibiotic resistant strains. In 2012, 1 in 5 cases of TB were multiply drug resistant (MDR-TB), while nearly 1 in 10 of these MDR-TB cases was also extensively drug resistant (XDR-TB). Totally drug resistant TB (TDR-TB), which is untreatable by current TB drugs, has also emerged. With current therapies becoming ineffective and the high TB burden continuing worldwide, the need for new drug targets and new therapeutics is of paramount importance.

iii

The first part of this work describes the biochemical and structural characterization of

GlgE, a maltosyltransferase from M. tb. Knock-out of GlgE causes rapid death of the bacterium, which is the result of a toxic build-up of the substrate maltose-1-phosphate. This novel method of killing makes GlgE an intriguing drug target for the design of novel inhibitors to kill M. tb. Here we report the design of a fluorescent based assay to monitor

GlgE activity. This was utilized to characterize the kinetics of the enzyme, measure and characterize mutant forms of the enzyme, and screen a compound library for potential inhibitors. Additionally, we report the crystal structures of the M. tb GlgE in a binary complex with maltose, and a ternary complex with maltose and a maltooligosaccharide, maltohexaose.

The second part of this work utilized the structural data obtained from the M. tb GlgE to mutate a homologous GlgE from Streptomyces coelicolor to produce an enzyme that crystallizes to higher resolutions than the M. tb GlgE, while having an identical active site.

This S. co GlgEI-V279S was used to structurally characterize interactions between the enzyme and four different inhibitor compounds. We also report the design of a novel coupled assay to monitor GlgE activity. This assay was fast, reproducible, and highly amenable to high-throughput screening applications.

The next part of this work looks at an essential enzyme in the biosynthesis of trehalose in Mycobacterium tuberculosis. TPP2, a trehalose-6-phosphate phosphatase, is essential for the viability of the organism. It has been speculated that cellular death is the result of the build-up of trehalose-6-phosphate, the substrate. This method of killing makes it an interesting target for the design of inhibitors. Here we report the design of a novel, coupled assay to monitor the activity of TPP2. This assay was also highly amenable to high-

throughput screening applications. We also attempted to crystallize and solve the structure of the TPP2 enzyme. However, structural information has remained elusive to this date.

Crystallization experiments are ongoing, with several new conditions and crystals awaiting diffraction experiments.

The final chapter of this work looks at the structural and biochemical characterization of a mycothiol-S-transferase from M. tb (MtMST). MtMST is responsible for maintaining redox homeostasis in the bacterium, detoxifying xenobiotics and electrophilic toxins, and has been implicated in the neutralization of some of the common anti-tubercular drugs used in the treatment of TB. Here we report the first structure of the MtMST. Additionally; we developed an assay to monitor MtMST activity, and utilized this to determine the ability of

MtMST to modify antibiotics in vitro.

Acknowledgements

First and foremost, I would like to thank my boss, Dr. Donald Ronning. He has been instrumental towards my success in the lab and I will forever be grateful for the opportunities, and knowledge I have obtained in his lab. I would also like to thank my committee members,

Drs. Sucheck, Bellizzi, and Slama, for helping me, and giving me guidance through the various projects.

Next, I need to thank all my lab members who have come and gone throughout my time in the lab. This includes, but is not limited to: Dan, Vidhi, Lorenza, Lucile, Erica, Chris,

Mike, Cecile, Paniz, all the undergraduates who have been in the lab, and if I missed someone, thank you too. Everyone has been helpful and made every day in the lab enjoyable.

I especially need to thank both Sri and Sandeep for helping make the GlgE project successful.

I also need to thank all the friends I have made these five years. I love you all, and appreciate everything.

Lastly, and most importantly, I want to thank my family for all their love and support.

To my mom and pops for their constant love and support over the years, and for all the free meals you’ve supplied, and to my brothers, Japey and Biggins, for always keeping it real with me and never missing an opportunity to make fun of each other.

Table of Contents

Abstract ...... iii

Acknowledgements ...... vi

Table of Contents ...... vii

List of Tables ...... xiv

Table of Figures ...... xv

List of Equations ...... xix

List of Abbreviations ...... xx

List of Symbols ...... xxiii

1 Chapter 1 ...... 24

A primer on Mycobacterium tuberculosis ...... 24

1.1 Background on Tuberculosis and Mycobacterium tuberculosis ...... 24

1.1.1 The Tuberculosis Burden ...... 24

1.2 Characterization of Mycobacterium tuberculosis ...... 25

1.2.1 The pathogenic Mycobacterium ...... 25

1.2.2 Description of Mycobacterium tuberculosis ...... 25

1.2.2 Infection and pathogenesis of the disease ...... 26

1.2.3 Treatments ...... 27 vii

1.2.4 Challenge of drug resistance and co-infection with HIV ...... 29

1.3 Overview and significance of the different projects ...... 31

1.3.1 GlgE ...... 31

1.3.2 TPP2 ...... 32

1.3.3 Mycothiol-S-transferase ...... 32

2 Chapter 2 ...... 34

Kinetic and structural characterization of the Mycobacterium tuberculosis maltosyltransferase, GlgE ...... 34

2.1 Background ...... 34

2.1.1 The many roles of glycogen ...... 34

2.1.2 The classical pathway of glucan synthesis: the GlgA pathway ...... 35

2.1.2 The non-classical pathway of glucan synthesis: the Rv3032 pathway ...... 38

2.1.3 The new kid on the block: the GlgE pathway and glucan biosynthesis ...... 39

2.1.3 Interplay of the glucan producing pathways ...... 40

2.1.4 The GlgE pathway and its significance for drug discovery ...... 41

2.2 Material and Methods ...... 42

2.2.1 Molecular cloning ...... 42

2.2.2 Expression and purification ...... 42

2.2.3 Quinaldine Red assay for GlgE activity ...... 43

2.2.4 Crystallization and structural determination of GlgE complexes ...... 44

2.3 Results and Discussion ...... 46

2.3.1 Optimization of the fluorescence based Quinaldine Red assay for GlgE activity

assessment ...... 46

viii

2.3.1.1 Kinetic characterization of GlgE using Quinaldine Red assay for phosphate

detection ...... 46

2.3.1.2 Assessing the quality of the Quinaldine Red assay for high-throughput

screening applications ...... 51

2.3.1.3 NIH Clinical Collection Screening of GlgE ...... 53

2.3.3 Determination of the wild-type M. tb GlgE structure in a binary complex with

maltose ...... 54

2.3.3.1 Crystallization strategies for the Mtb GlgE-maltose ...... 54

2.3.3.2 Data collection and structural refinement of the GlgE-maltose structure .. 56

2.3.3.3 Analysis of the Mtb GlgE-maltose structure ...... 58

2.3.4 Identification of a high-affinity maltohexaose-binding site in the M. tb GlgE ...... 65

2.3.4.1 Crystallization strategies for the Mtb GlgE-maltohexaose ...... 65

2.3.4.2 Data collection and structural refinement of the Mtb GlgE-maltohexaose

structure ...... 65

2.3.4.3 Structural analysis of the Mtb GlgE-maltohexaose structure ...... 67

2.3.2 Identification of important catalytic residues in GlgE ...... 73

2.3.2.1 Rationale for choosing Asp383 and Asp418 ...... 73

2.3.2.2 Docking and enzymatic studies on identified residues ...... 74

2.5 Conclusions and future works ...... 76

3 Chapter 3 ...... 79

Structural Characterization of the Streptomyces coelicolor GlgEI-V279S in complexes with non-covalent inhibitors ...... 79

3.1 Rationale for using the S. co GlgEI as a surrogate for the M. tb GlgE ...... 79

3.2 Material and Methods ...... 81

3.2.1 Molecular cloning of the Sco glgEI ...... 81

3.2.2 Expression and purification ...... 82

3.2.3 Crystallization of Sco GlgEI-V279S-MCP, GlgEI-V279S- αMTF, and GlgEI-

V279S-DDGIM, and GlgEI-V279S-DDGMP complexes ...... 82

3.2.4 Diffraction experiments...... 82

3.2.5 Structure Determination of the Sco GlgE-V279S complexes...... 83

3.2.4 Mass Spec labeling studies for covalent modification ...... 83

3.2.5 Sco-GlgE-V279S maltooligosaccharide extension using M1F and α-MTF as

substrates ...... 84

3.2.6 Determination of maltooligosaccharide elongation by Sco GlgEI-V279S size using

MALDI-MS ...... 84

3.2.7 Development and optimization of a real-time coupled assay for GlgE activity and

high-throughput screening...... 84

3.3 Results and Discussion ...... 85

3.3.1 Crystallization strategies for the Sco GlgEI-V279S ...... 85

3.3.2 Sco GlgE-V279S in a binary complex with Maltose-C-phosphonate...... 87

3.3.3 Sco GlgE-V279S in a binary complex with an oxocarbenium mimic ...... 95

3.3.4 Sco GlgE-V279S in a binary complex with an imino-mannitol-N-methyl-C-

phosphonate, DDGMP ...... 99

3.3.5 Confirming D418/394 as the genuine nucleophile of GlgE enzymatic activity .. 107

3.3.5.1 Probing covalent modification of GlgE by α-MTF ...... 107

3.3.5.2 MALDI-MS experiments ...... 108

3.3.5.3 Inhibition studies of the α-MTF ...... 109

3.3.5.4 Assessment of α-MTF incorporation into linear glucans ...... 110

3.3.5.5 Sco GlgEI-V279S in a binary complex with α-MTF ...... 113

3.3.6 Development of a new, quantitative, real-time assay for GlgE activity ...... 116

3.3.6.1 Attempts to utilize PNP-maltose and PNP-glucose ...... 116

3.3.6.2 A real-time assay utilizing the reverse GlgE reaction and PNP-

maltopentaose ...... 118

3.4.5 Screening of potential GlgE inhibitors utilizing the PNP-M5 assay ...... 122

3.5 Conclusions and future works ...... 123

4 Chapter 4 ...... 125

Biochemical characterization and structural studies of the Mycobacterium tuberculosis treahalose-6-phosphate phosphatase, TPP2 ...... 125

4.1 Introduction ...... 125

4.1.1 Trehalose and its role in Mycobacteria ...... 125

4.1.2 Trehalose biosynthesis and storage in Mycobacterium tuberculosis ...... 127

4.1.3 TPP2 and its potential as a drug target in the treatment of TB ...... 128

4.2 Materials and Methods ...... 129

4.2.1 Molecular cloning of the otsB2 gene, D147N mutant, and E. coli treA...... 129

4.2.2 Expression and purification of TPP2, TreA, and mutants...... 129

4.2.3 Preparation of the Amplex Red working solution ...... 131

4.2.4 Determination of the enzyme concentrations to be used in the assay and control

reactions ...... 131

4.2.5 Michaelis-Menten kinetics determination of TPP2 using the Amplex Red assay 132

4.2.6 Assay validation for high-throughput screening applications ...... 132

4.2.7 NIH Clinical Collection screening ...... 133

4.2.8 Crystallization attempts for TPP2 ...... 133

4.3 Results and Discussion ...... 134

4.3.1 Developing a new assay to measure TPP2 activity ...... 134

4.3.1.2 Optimization of enzyme concentrations for the assay ...... 136

4.3.1.3 Michaelis-Menten kinetics for TPP2 ...... 139

4.3.1.4 Z’ determination and NIH screening ...... 141

4.3.1.5 Piceid inhibition against TPP2 ...... 142

4.3.2 Crystallization of TPP2 ...... 144

4.3.2.3 TPP2 crystallization utilizing a TPP2-n-dodecyl-β-D-maltoside -AlFx

complex ...... 147

4.3.2.4 TPP2 crystallization utilizing a TPP2 4-dodecyl-trehalose vanadate

complex ...... 151

4.3.2.5 Crystallization of an inactive mutant, TPP2-D147, and T6P ...... 153

4.4 Conclusions and Future Works ...... 155

5 Chapter 5 ...... 157

Structural characterization of the Mycobacterium tuberculosis Mycothiol-S-transferase .... 157

5.1 Background ...... 157

5.1.1 Biosynthesis of mycothiol (MSH)...... 157

5.1.2 Utilization of Mycothiol ...... 159

5.1.3 The Mycobacterium tuberculosis mycothiol-S-transferase, MtMST ...... 161

5.1.3 Mycothiol, drug resistance, and a new therapeutic target ...... 161

xii

5.2 Materials and Methods ...... 162

5.2.1 Molecular cloning of Rv0443c ...... 162

5.2.2. Expression and purification of MtMST ...... 163

5.2.3 Crystallization and structural determination ...... 164

5.2.4 Virtual Screening using Glide ...... 164

5.2.5 Assay for assessing MtMST activity and antibiotic modification ...... 165

5.3 Results and Discussion ...... 165

5.3.1 Crystallization strategies for MtMST ...... 165

5.3.2 Data collection and structural refinement of the MtMST structure ...... 168

5.3.3 MtMST monomeric structure contains the highly conserved, 4-helical bundle

domain of the DinB superfamily ...... 170

5.3.4 Docking Studies reveal a potential Mycothiol binding site in MtMST ...... 175

5.3.5 Assessing the activity of MtMST using monoChlorobimane ...... 177

5.3.6 Modification of antibiotics by MSH ...... 179

5.4 Conclusions and future works ...... 180

References ...... 181

xiii

List of Tables

Table 1: Common first and second line drugs that are currently in clinical use as part of the

TB treatment regimens...... 29

Table 2: Primers used for PCR amplification of the various glgE mutants ...... 42

Table 3: Comparison of the kinetic parameters of GlgE enzymes ...... 51

Table 4: Z’ Determination of the GlgE Quinaldine Red assay ...... 53

Table 5: Data Collection and Refinement statistics for GlgE-MAL and GlgE-M6 ...... 58

Table 6: Data collection and refinement statistics for Sco-GlgE-V279S structures...... 89

Table 7: Bond distances observed between residues in the Sco GlgEI-V279S active site and the atom in the inhibitors...... 105

Table 8: Z’ determination of the PNP-M5 assay...... 122

Table 9: Screens and conditions used during TPP2 crystallization experiments...... 134

Table 10: Comparison of kinetic parameters of different TPP enzymes and the assay used to determine them...... 141

Table 11: TPP2 Amplex Red Z’ determination ...... 142

Table 12: Data collection and refinement statistics for the MtMST...... 169

xiv

Table of Figures

Figure 1: Bacterial and eukaryotic glycogen...... 35

Figure 2: Mycobacterium tuberculosis classical and non-classical glucan biosynthetic pathways...... 36

Figure 3: M. tb glucans ...... 38

Figure 4: Quinaldine Red based assay for phosphate detection...... 48

Figure 5: GlgE Quinaldine Red progress curve...... 49

Figure 6: Michaelis-Menten curve for M1P...... 50

Figure 7: Initial GlgE crystals ...... 55

Figure 8: Improved GlgE crystals ...... 56

Figure 9: Mtb GlgE overall structure...... 60

Figure 10: Dynamic loop and active site of the Mtb GlgE...... 62

Figure 11: Active sites of the Mtb GlgE and Sco GlgEI...... 64

Figure 12: M6 docking site of GlgE...... 67

Figure 13: A high affinity docking site for M6...... 68

Figure 14: Maltohexaose binding site sequence alignment between the Sco GlgEI and Mtb

GlgE...... 69

Figure 15: Surface conservation comparison of the Mtb and Sco GlgE enzymes...... 71

Figure 16: GlgE bound to model linear glucan...... 73

Figure 17: Structural docking studies of M1P and GlgE...... 75

Figure 18: Relative activities of GlgE and mutants...... 76

Figure 19: GlgE dimers with active sites...... 81

Figure 20: Initial Sco GlgEI-V279S crystals...... 86

Figure 21: Optimized Sco GlgEI-V279S crystals...... 87

Figure 22: Malose-1-phosphate and the first GlgE inhibitor, α-maltose-C-phosphonate...... 88

Figure 23: Sco GlgEI-V279S in complex with MCP...... 90

Figure 24: Interactions between the MCP and Sco GlgEI-V279S...... 92

Figure 25: The second inhibitor for GlgE DDGIM and the oxocarbenium transition state. .. 95

Figure 26: Sco GlgEI-V279S in complex with DDGIM...... 96

Figure 27: Sco GlgE-V279S in complex with DDGIM...... 98

Figure 28: The third GlgE inhibitor DDGMP...... 100

Figure 29: Sco GlgEI-V279S in complex with DDGMP...... 101

Figure 30: Interactions between the Sco GlgE-V279S and DDGMP...... 103

Figure 31: Potential inhibitory compounds based on the DDGMP structure...... 107

Figure 32: α-maltose-1-fluoride and the α-MTF...... 108

Figure 33: Evaluation of α-MTF as a substrate for Sco GlgEI-V279S...... 112

Figure 34: Sco GlgEI-V279S in complex with α-MTF...... 114

Figure 35: Interactions between Sco GlgE-V279S and α-MTF...... 115

Figure 36: Attempts to utilize new substrates for GlgE activity assessment...... 117

Figure 37: The PNP-M5 GlgE assay...... 120

Figure 38: Progress curve of the PNP-M5 assay...... 120

Figure 39: α-glucosidase concentration dependence determination...... 121

xvi

Figure 40: Screening of GlgE inhibitory compounds...... 123

Figure 41: Cartoon image of trehalose...... 126

Figure 42: Trehalose biosynthetic and utilization pathways...... 127

Figure 43: Schematic of the TPP2 Amplex Red assay ...... 136

Figure 44: Initial progress curves from the Amplex Red assay ...... 137

Figure 45: Determination of the TreA concentration dependence...... 138

Figure 46: TPP2 concentration dependence on the Amplex Red assay...... 139

Figure 47: Michaelis-Menten curves for T6P...... 140

Figure 48: Piceid causes lag time during the Amplex Red assay...... 143

Figure 49: Diffraction of the TPP2-trehalose-AlFx crystals...... 146

Figure 50: Reproduced TPP2-trehalose-AlFx crystals after optimization...... 147

Figure 51: Sequence alignment between the M. tb and T. acidophilum TPP enzymes...... 148

Figure 52: TPP2 with modeled in N-terminal domain...... 149

Figure 53: Crystals from the TPP2-n-dodecyl-β-D-maltoside -AlFx complex...... 151

Figure 54: Crystals obtained from the TPP2 4-dodecyl-treahalose vanadate complex...... 152

Figure 55: Activity comparison of the wild-type TPP2 to the D147N variant...... 154

Figure 56: Initial crystals obtained from the TPP2-D147N and T6P complex...... 155

Figure 57: The three most common reducing thiols...... 158

Figure 58: Mycothiol biosynthetic pathway...... 159

Figure 59: Detoxification by the mycothiol...... 160

Figure 60: Structures of Mycothiol and the advanced synthetic intermediate used for crystallization attempts...... 166

Figure 61: Initial crystals of MtMST...... 167

xvii

Figure 62: Diffraction quality crystal of MtMST...... 168

Figure 63: Overall MtMST structure...... 170

Figure 64: Structural elements of MtMST...... 172

Figure 65: Determination of the MtMST dimer...... 173

Figure 66: MtMST dimer and binding pocket...... 174

Figure 67: The MtMST active site...... 176

Figure 68: Glide docking of MSH...... 177

Figure 69: monoChlorobimane assay for MtMST activity...... 178

Figure 70: Modification of anti-tubercular drugs by MSH...... 179

xviii

List of Equations

Equation 1: Z' calculation ...... 52

xix

List of Abbreviations

Abs ...... Absorbance AcCoA ...... Acetyl coenzyme A Add ...... Additive ADP...... Adenine diphosphate ADP-Glu ...... Adenine diphosphate-glucose AIDS ...... Acquired Immune Deficiency Syndrome AMK ...... Amikacin APS-ANL...... Advance Photon Source Argonne National Laboratory ATP ...... Adenine triphosphate

BCG ...... Bacille-Calmette-Guérin CHCA ...... α-Cyano-4-hydroxycinnamic acid CoA ...... Coenzyme A

DDAO ...... N,N-dimethyldodecylamine N-oxide DDGEP ...... 2,5-dideoxy-3-O-α-D-glucopyranosyl-2,5-imino-D-mannitol- N-ethyl-C-phosphonate DDGIM ...... 2,5-dideoxy-3-O--D-glucopyranosyl-2,5-imino-D-mannitol DDGMP ...... 2,5-dideoxy-3-O-α-D-glucopyranosyl-2,5-imino-D-mannitol- N-methyl-C-phosphonate DMSO ...... Dimethyl sulfoxide DTT ...... Dithiothreitol

E. coli ...... Escherichia coli EDTA ...... Ethylenediaminetetraacetic acid eLBOW ...... electronic Ligand Builder and Optimization Workbench EMB ...... Ethambutol ETH ...... Ethionamide

G1P ...... Glucose-1-phosphate G6P ...... Glucose-6-phosphate Glu...... Glucose

HEPES ...... 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid HIV ...... Human Immunodeficiency Virus HTS ...... High-throughput screening xx

INH ...... Isoniazid IPTG ...... Isopropyl β-D-1-thiogalactopyranoside KAN ...... Kanamycin KCl ...... Potassium chloride keV ...... kiloelectron volt

LC ...... Liquid Chromatography LS-CAT...... Life Sciences Collaborative Access Team

M. bovis ...... Mycobacterium bovis M1P ...... Maltose-1-phosphate M6 ...... Maltohexaose MAL ...... Maltose MALDI ...... Matrix-assisted laser desorption/ionization MAME ...... Cell wall-bound mycolic acids MCP ...... Maltose-1-C-phosphonate MDR ...... Multidrug-resistant MDR-TB ...... Multiple Drug Resistant Tuberculosis MG ...... Malachite Green MgCl2 ...... Magnesium chloride MGLP ...... Methyl-glucose lipopolysaccharide MIC ...... Minimum Inhibitory Concentration MOM...... Mycobacterial outer membrane MSH ...... Mycothiol MST ...... Mycothiol-S-transferase Mtb ...... Mycobacterium tuberculosis

NCC ...... NIH Clinical Collection

PMPS ...... Polymethylated polysaccharides PNP-Glu ...... para-nitrophenol-glucose PNP-M5 ...... para-nitrophenol-maltopentaose PNP-Mal ...... para-nitrophenol-maltose PZA ...... Pyrazinamide

Qred...... Quinaldine Red

RMP ...... Rifampicin

Sco...... Streptomyces coelicolor

TB ...... Tuberculosis TDR-TB ...... Totally Drug Resistant Tuberculosis TPP2 ...... Trehalose-6-phosphate phosphatase TPS ...... Trehalose phosphate synthase

xxi

UDP...... Uridine diphosphate

XDR-TB ...... Extensively Drug Resistant Tuberculosis

α-MTF ...... 2-deoxy-2,2-difluoro-α-maltose-1-fluroide βME ...... β-mercaptoethanol

xxii

List of Symbols

°C...... degree Celsius Å...... Angstrom M...... molar mg...... milligrams mL...... milliliter mM...... millimolar nm...... nanometer nM...... nanomolar pH...... potential hydrogen rpm...... revolution per minute v/v...... volume/volume w/v...... weight/volume μL...... microliter μM...... micromolar

xxiii

Chapter 1

A primer on Mycobacterium tuberculosis

1.1 Background on Tuberculosis and Mycobacterium tuberculosis

1.1.1 The Tuberculosis Burden

Despite near eradication, the Tuberculosis (TB) pandemic continues to remain one of the most important health threats throughout the world. Worldwide, TB related deaths are second only to HIV for all infectious diseases 1. TB is the result of an infection by the bacterium Mycobacterium tuberculosis (M. tb). It has been estimated that nearly one-third of the world’s population is infected by M. tb 1. In healthy individuals, infection by M. tb is controlled by the immune system and results in a latent phase of the infection that can last for decades. During this time, those infected are not contagious and appear normally healthy.

However, of the nearly 2 billion people infected, nearly 9 million of those will develop an active infection in any particular year 1. During this time, people are highly contagious, and can easily spread the bacterium and infect others. Of the 9 million active infections, nearly

1.3 million of those will succumb to the disease 1. Exacerbating this problem is co-infection by HIV, as 25% of TB deaths were of those who were HIV/AIDS positive 1. Contributing to this high death toll is that the disease afflicts primarily under-developed countries where

access to treatment is difficult to find. Unsurprisingly, the greatest incidence of both TB deaths and new cases are found predominately in South-East Asia, China, India, and the

African continent 1.

1.2 Characterization of Mycobacterium tuberculosis

1.2.1 The pathogenic Mycobacterium

Robert Koch identified the infectious nature of TB in 1882 2. His work characterized the etiological agent of TB as a bacterium: Mycobacterium tuberculosis, also called tubercle bacillus, or Koch’s bacillus. Robert Koch was awarded the Nobel Prize in Physiology or

Medicine in 1905 for his pioneering work in bacteriology and, especially his investigations and discoveries in relation to Tuberculosis 2.

Mycobacterium tuberculosis is a pathogenic bacterium belonging to the

Mycobacteriaceae family. M. tb is the etiological agent of TB for which humans are the sole carrier. M. tb in addition to several other species of Mycobacterium, are known to cause

Tuberculosis in humans and other organisms. Collectively these organisms are known as the

Mycobacterium tuberculosis complex 3. Several other species from the Mycobacterium genus are also known to cause diseases distinct from Tuberculosis. Mycobacterium leprae is the causative agent of leprosy, while the Mycobacterium avium complex and Mycobacterium kansasii are opportunistic mycobacterial infections that cause a pulmonary diseases that is clinically distinct from TB 4, 5.

1.2.2 Description of Mycobacterium tuberculosis

Mycobacterium tuberculosis is a highly aerobic bacterium requiring large amounts of

oxygen for proper growth. It is a slow dividing bacteria, requiring 12-24 hours per division 6.

M. tb is considered a gram-positive bacterium despite the fact that it does not retain crystal violet stain. M. tb lacks a true outer membrane, but has a thick waxy coating on its surface which makes it resistant to staining 7. This waxy coating, which is characteristic of all species of Mycobacteria, is composed primarily of the long chain fatty acids, mycolic acids. This thick, hydrophobic cell wall is believed to confer the bacteria passive resistance to many drug molecules, protection from immune cells, and is a source for many virulence factors 5, 8, 9.

1.2.2 Infection and pathogenesis of the disease

Infection by the bacterium occurs through the transfer of aerosolized drops from an actively infected person to an un-infected individual. The bacilli enter the upper respiratory tract of the host at which point the various stages of the disease progression can begin 10.

Once in the lungs, the bacteria are recognized my alveolar macrophages and internalized through a process called phagocytosis 11. Following this, the bacterium is encapsulated by the phagosome. Normally following phagocytosis, bacteria or other foreign objects are ultimately destroyed following maturation of the phagosome. Maturation of the phagosome involves many different processes, including the fusion of the phagosome to the lysosome, acidification of the phagosome, and the production of reactive oxygen and nitrogen species 12.

However, this maturation process is arrested by M. tb, and allows the bacterium to thrive and progress through the remainder of infection process following macrophage uptake 13, 14, 15, 16.

How this phagosomal maturation is arrested by M. tb is not entirely understood, but many of the underlying mechanisms have been studied 10, 15, 16, 17. In the majority of cases, the bacteria are contained following macrophage infection. Recruitment of other immune cells to the site of infection results in the formation of granulomas, which allow the immune system 26

to contain the infection and promote the latent phase 18. An active infection can occur when the ability of the immune system is compromised, and the granulomas can no longer be maintained, such as in the case of co-infection with HIV 6, 19. Granuloma failure results in the formation of a nutrient rich environment where the M. tb bacilli can replicate uncontrollably.

90% of M. tb infections are confined to the lungs (pulmonary TB), while some infections can occur in other bodily systems (extra pulmonary TB) if the person infected is immunocompromised 6.

1.2.3 Treatments

Prior to the dawn of modern medical techniques and antibiotics, those suspected of having TB infections were sent to sanatoriums to prevent the spread of the disease 20.

Following the discovery that TB was caused by the M. tb bacterium by Dr. Robert Koch in

1882, various treatments have come and gone with varying levels of success 20. Beginning in

1921, a vaccine was developed to prevent infections. The Bacille Calmette Guerin vaccine

(BCG) was based on the non-pathogenic Mycobacterium bovis 21, 22. This vaccine is still widely used today as a preventative measure in infants 21. However, it cannot treat the disease once the person has been infected. It is still the only vaccine that has been approved for use in humans against M. tb 22.

The first antibiotic used in the treatment of TB was streptomycin. However, strains that were resistant against streptomycin quickly emerged and the necessity for new antibiotics quickly increased 23, 24. Current antitubercular drugs are divided into three different groups: first, second, and third-line drugs (Table 1). First-line drugs, as their name suggest, are given during the initial stages of infection. These include isoniazid (INH), rifampicin (RMP), ethambutol (EMB), and pyrazinamide (PZA) 1. Use of second-line drugs 27

is restricted to infections that do not appear to respond to any first-line drugs 25. Additionally, they typically have more side effects, increased toxicity, and are less effective than the first- line drugs 1. Second-line drugs include the aminoglycosides, polypeptides such as capreomycin and viomycin, thioamides, and the fluoroquinones. The third-line drugs are often administered as a last resort, as their efficacy has not been demonstrated. These include thioacetazone, macrolides, and clofazamine 26, 27, 28.

Table 1: Common first and second line drugs that are currently in clinical use as part of the TB treatment regimens.The table has been adapted from Zumla et al. Nat. Rev. Drug Discov., 2013 29.

Drug Name Year Target Mode of Inhibition Discovered First-line drugs

INH 1952 Enoyl-acyl carrierprotein Mycolic acid reductase biosynthesis RMP 1963 RNA polymerase β-subunit Transcription PZA 1954 S1 component of the 30S Translation and ribosomal subunit trans-translation, acidifies cytoplasm EMB 1961 Arabinosyl transferase Arabinogalactan biosynthesis Second-line drugs p-amino salicylic 1948 Dihydropteroate synthase Folate biosynthesis acid Streptomycin 1944 S12 and 16S rRNA Protein synthesis components of the 30S ribosomal subunit Ethionamide 1961 Enoyl-[acyl-carrier-protein] Mycolic acid reductase biosynthesis Ofloxacin 1980 DNA gyrase and DNA DNA supercoiling topoisomerase Capreomycin 1963 Interbridge B2a between Protein synthesis 30S and 50S ribosomal subunits Kanamycin 1957 30S ribosomal subunit Protein synthesis

Amikacin 1972 30S ribosomal subunit Protein synthesis

Cycloserine 1955 D-alanine racemase and Peptidoglycan ligase synthesis

1.2.4 Challenge of drug resistance and co-infection with HIV

Treatment with of M. tb infections is a long duration process, but if managed properly, can cure the disease 30. Typical therapy starts with an initial two-month period using the first-

line drugs: INH, RMP, PZA, and EMB 1. This is followed by a 4-7 month continuation treatment phase using INH and RMP 30, 31. Taken together, the total treatment requires 6-9 months and requires up to 192 doses of medication 1. Because of the total drug load and the long duration of treatment, strict adherence to the regimen is required to completely eradicate the infection. In under-developed areas of the world such as China, South-East Asia, India, and Africa where TB is most prevalent, access to the necessary health care and medication is often a problem 1, 30. Therefore, doses of the drugs are missed and the appropriate follow-up monitoring is not provided. This has, in part, led to the rise of drug resistant strains of M. tb

30, 31.

Drug-resistant strains of M. tb can be classified as either multiple-drug resistant

(MDR-TB), or extensively-drug resistant (XDR-TB) 1. The WHO estimates that 450,000 cases were MDR-TB, and that 170,000 of those died from it in 2012 1. MDR-TB strains are resistant to several of the first-line drugs, most commonly INH and RMP 1. XDR-TB strains are resistant to all first-line drugs, in addition to one or more of the second-line drugs. The

WHO estimated that nearly 10% of the MDR-TB cases were XDR 1. More recently, cases of variant strains that are totally-drug resistant (TDR-TB) to all known anti-tubercular drugs have emerged 32.

Another difficulty in treating M. tb infections concerns those individuals who are also infected with HIV. Among all active TB cases reported, 13% of those patients were also HIV positive 1. Most strikingly, was the fact that of all deaths by TB, 25% were from those co- infected with HIV 1. Part of the problem arises from the fact that immunocompromised individuals are at the most risk to develop an active infection due to their weakened immune

systems 33. Additionally, treatment with anti-tubercular drugs is difficult due to the fact that they are contraindicative with many of the anti-viral drugs prescribed for HIV infections 24, 34.

1.3 Overview and significance of the different projects

1.3.1 GlgE

GlgE is part of the TreS/Pep2/GlgE/GlgB pathway which synthesizes an α-1,4-linked,

α-1,6-branced glucan from trehalose. GlgE is a maltosyltransferase which utilizes the phospho-sugar, maltose-1-phosphate (M1P), as the maltosyl-donor substrate to extend the linear portion of the α-1,4 glucans which are ultimately branched by the glycogen branching enzyme, GlgB. The GlgE pathway is dispensable in M. tb owing to the fact that it shares some functional redundancy with another glucan producing pathway, the Rv3032 pathway.

However, genetic studies have shown that bacteria lacking both pathways (ΔtreSΔRv3032) are not viable. More interesting was the observation that inactivation of glgE alone resulted in the rapid death of the bacteria. This was not the result of the absence of the α-1,4-glucan, but rather the toxic buildup of the substrate, M1P. This suggests that targeted inhibition of

GlgE could lead to new therapeutics that could rapidly kill the bacterium. This rapid killing could side step the issues of patient non-compliance, which can lead to drug resistance.

Chapter 2 of this research looks at the efforts to determine the M. tb GlgE structure using X-ray crystallographic techniques. This would allow for structural based design of inhibitors specifically for GlgE. Secondly, we looked at developing an assay to test the activity of GlgE, as well as mutant forms of the enzyme in order to ascertain the importance of individual residues in the active site. Chapter 3 describes the effort to utilize a variant

GlgE homologue from Streptomyces coelicolor (GlgEI-V279S) to produce high-resolution

structures in complex with various inhibitors designed to target GlgE. Additionally, we developed a continuous, high-throughput assay to monitor GlgE enzymatic activity.

1.3.2 TPP2

TPP2, trehalose-6-phosphate phosphatase, is part of the ostA/otsB2 (TPS/TPP2) pathway in M. tb and is the sole de novo biosynthetic pathway for trehalose in the bacterium.

TPS produces trehalose-6-phosphate from UDP-glucose and glucose-6-phosphate. This is subsequently dephosphorylated by the action of TPP2. TPP2 has been shown to be essential for the viability of the organism. This is potentially due to its importance in the biosynthesis of trehalose, or due to the potential of T6P being toxic in a similar manner as M1P described in glgE inactivation. Therefore, designing inhibitors to block the activity of TPP2 may be of therapeutic value in developing new treatments for the disease.

Chapter 4 of this research looks at the development of a novel, high-throughput screening assay, to assess the kinetics of TPP2, in addition to its use in the screening of compounds to target TPP2. This assay was also utilized to test the activity of a mutant form of TPP2. This chapter also describes the efforts to obtain the crystal structure of the TPP2 enzyme from M. tb.

1.3.3 Mycothiol-S-transferase

Mycobacterium tuberculosis, in addition to many other Actinobacteria, produces large amounts of mycothiol (MSH). Mycothiol is a low-mass thiol, which serves to maintain redox homeostasis within the bacterium. In addition to this role, mycothiol serves as a protectant to neutralize xenobiotics, electrophilic toxins, and other oxidative stresses. The utilization of MSH by the bacterium has also been implicated in the neutralization of many

anti-tubercular drugs, as well as playing a possible role in the activation of several pro-drugs used to treat the disease. MSH is produced in the MshA/B/C/D pathway and is ultimately utilized by the mycothiol-S-transferase, MST. The production and utilization of MSH in M. tb is essential for viability. Therefore, discovering drugs that may target and inhibit MST may provide an unexplored avenue for combating the disease.

Chapter 5 of this research looks at determining the crystal structure of the M. tb MST.

In addition to this, docking software was used to probe the surface of the MtMST to identify the potential active site of MtMST in the absence of a structure bound to MSH. The ability of

MtMST to modify different antibiotics was also determined using a fluorescence based assay.

Chapter 2

Kinetic and structural characterization of the

Mycobacterium tuberculosis maltosyltransferase, GlgE

2.1 Background

2.1.1 The many roles of glycogen

Glycogen is a highly soluble α-1,4-linked-α-1,6-branced polymer of glucose, which can have anywhere from a few hundred, to many thousands of glucose units throughout its chains and branches 35. In animals, it is primarily utilized as an energy source, as it can be quickly converted to glucose to power the metabolic needs of the organism. In bacteria, one purpose of glycogen is to act as a carbon sink for when there is an excess of carbon available to the bacterium as a means to protect against times of starvation 36, 37. Besides this role, glycogen storage can also lead to many profound physiological changes in bacteria, and is required for a variety of functions. In bacteria such as Bacillus subtilis and Streptomyces coelicolor, glycogen accumulation is necessary for sporulation events 36. In the human pathogen Salmonella enteritidis, glycogen synthesis is required for biofilm formation and virulence 36. For this reason, the synthesis and recycling of glycogen and other branched

glucans within cells are highly redundant and controlled. Though structurally similar, eukaryotic glycogen synthases requires the enzyme glycogenin, to produce small, covalently attached glucan primers for the initiation of glycogen production. Bacterial glycogen can be produced de novo in the absence of a primer (Figure 1) 38.

2.1.2 The classical pathway of glucan synthesis: the GlgA pathway

M. tb along with virtually every other bacterial species, possess the GlgA pathway as the primary biosynthetic route for the synthesis of branched α-1,4 glucans 39. As such, the

GlgA pathway has been denoted as the classical pathway for glycogen synthesis. This pathway generates glycogen from the action of three different enzymes: GlgC, GlgA, and

GlgB (Figure 2).

Figure 1: Bacterial and eukaryotic glycogen.Both are composed primarily of α-1,4-linked (blue circles) glucose with α-1,6-branches (red circles). However, eukaryotic glycogen requires a core molecule of glycogenin to serve as a primer, whereas the bacterial form can be made without one.

Extracellular glucose is imported into the cell and phosphorylated to produce Glucose-6- phosphate (G6P) 38. G6P is converted to Glucose-1-phosphate (G1P) by the action of the

phosphoglucomutase PgmA. GlgC, a nucleotide diphosphoglucose pyrophosphorylase, generates ADP-glucose from the G1P generated from PgmA. ADP-glucose is then polymerized by the glycogen synthase, GlgA, to form a linear glucan.

Figure 2: Mycobacterium tuberculosis classical and non-classical glucan biosynthetic pathways.Text in blue represents enzymes, while those in red are products or substrates within the pathways. Black arrows lead to glycogen or glucan synthesis, while green arrows represent the breakdown. This figure was adapted from Chandra et al. Microbiology, 2011 39. Subsequent α-1,6 branches are added to the glucan to yield a branched glucan by the glycogen branching enzyme, GlgB. GlgA can then further elongate these branches, and these extended branches can be further modified by GlgB to produce more branches (Figure 3A).

The end result yields the highly branched glycogen macromolecule. The glycogen and glucan produced from this pathway can be broken down and recycled by several different pathways.

The glycogen phosphorylase, GlgP, can break down single glucose units of glycogen to yield

G1P. This G1P can be recycled to produce more glycogen, or enter the into glucose

metabolism pathways to produce energy. Another pathway, the TreXYZ pathway, is a series of enzymes that remove two glucosyl units, ultimately producing trehalose from the glycogen

40.

In Mycobacteria, the glycogen produced from the GlgA pathway can serve a variety of roles. As previously stated, it can simply act as a carbon storage device that can be synthesized or broken down as the metabolic needs of the bacterium change. Glycogen recycling has also been implicated for proper bacterial growth. The non-pathogenic

Mycobacterium smegmatis requires glycogen recycling during exponential phase growth in order to remain viable 37. Additionally, many pathogenic Mycobacteria have a loosely joined outer layer known as the capsule. The capsule is composed mostly of carbohydrates, in addition to small amounts of proteins and lipids. A recent study by Sani et al. of the capsular layer of various Mycobacterial species showed that the vast majority of the carbohydrate portion of the capsule (~80-90%) is comprised of glycogen, while the remaining carbohydrate portion is composed of a non-acylated arabinomannan (~10-20%) 41.

Additionally, secreted proteins from the ESX-1/2 secretions were also present 41. However, the composition of the capsular layer can vary widely depending on growth conditions, species, and growth phase of the bacterium. It is also easily disrupted in the presence of detergents and agitation during culture 41. This capsular glucan has been demonstrated to be very important in immune system modulation, phagocytosis, and persistence inside the host

42. Despite this importance, little is known about the mechanism by which the glucan is exported to the capsular layer.

2.1.2 The non-classical pathway of glucan synthesis: the Rv3032 pathway

In addition to the classical GlgA pathway of glucan synthesis, M. tb produces unique polymethylated polysaccharides (PMPS). This non-classical glucan-producing pathway, the

Rv3032 pathway, produces 6-O-methylglucose lipopolysaccharides (MGLP) (Figure 3B).

MGLPs are produced ubiquitously in Mycobacteria, being found in both fast and slow growing species 43. An additional PMPS, 3-O-methylmannose lipopolysaccharides (MMP), is also found in fast growing Mycobacteria such as M. smegmatis and M. phlei, but is absent in slow growing Mycobacteria such as M. tuberculosis (Figure 3C) 43. As a result, it is thought that MGLPs play a significant role in the pathogenicity of the bacterium.

Figure 3: M. tb glucans Examples of the various glucans produced in Mycobacteria. A) α-1,4-linked α-1,6-branced glycogen; R = additional glucosyl units. B) MGLP from M. bovis, R1 = octanoate, R2 = succinate, R3 = acetate, propionate, or isobutyrate. C) MMP from M. smegmatis. This figure was adapted from Mendes et al. Nat. Prod. Rep, 2012 43

MGLPs are α-1,4-linked glucans of approximately 20 glucose units in length with α-

1,6-branches. The unique feature of the MGLPs is that they are heavily acylated along the glucosyl units in the chain 44. Despite being described since the 1960s, the physiological role

MGLPs are playing in M. tb is still poorly understood. There is evidence that they may regulate fatty-acid biosynthesis, or the chaperoning of these fatty acids by complexing with them. Many MGLPs have been shown to form stable 1:1 complexes with several different fatty acid substrates in vitro 44. The Rv3032 pathway has been recently described and proposed to be the source of the MGLPs generated in Mycobacterium tuberculosis. Rv3032, unlike GlgA, has been proposed to utilize both ADP-Glu and UDP-Glu as the glucosyl donor for linear glucan synthesis. Subsequent branching of the glucan is achieved by another enzyme, Rv3031 45. This branched glucan is then is acylated by a variety of different enzymes in the pathway. Some of the proposed acetyltransferases include Rv3030, Rv3034, and Rv1217, however their roles are still unconfirmed (Figure 2) 46.

2.1.3 The new kid on the block: the GlgE pathway and glucan biosynthesis

In addition to the classical GlgA pathway, and the non-classical MGLP producing

Rv3032 pathway, a new, non-classical pathway has been recently identified in M. tb, which catalyzes the conversion of trehalose into linear glucans 47. This GlgE pathway is novel in that it utilizes a disaccharide phosphate as a donor molecule rather than a monosaccharide nucleotide diphosphate, which is observed in both the GlgA and Rv3032 pathways 47. The

GlgE pathway begins through the isomerization of trehalose to maltose by the trehalose synthase, TreS. Maltose is then phosphorylated by the maltokinase (Pep2/Mak) using ATP, to yield maltose-1-phosphate (M1P). The M1P generated by Pep2 is then utilized by the maltosyltransferase GlgE. GlgE catalyzes the elongation of linear α-1,4 glucans by the addition of a single maltosyl unit from M1P. The glucans generated from GlgE can then be branched by a glycogen branching enzyme, GlgB (Figure 2).

The identification of the GlgE pathway also identifies a means for which the bacterium is able to store excess trehalose. Trehalose production is up-regulated in times of stress, and Mycobacteria have specific transporters to recover trehalose from the outer membrane and extracellular space 48. Additionally, trehalose is used as the primary carbon source in Mycobacterium making its storage and utilization even more important 49. However, storage of the sugar afterwards had been previously unknown. Experiments with TreS showed a net flux from trehalose to maltose at a ratio of about 4:1 49. The identification of the

GlgE pathway and the committed step of phosphorylation of maltose by Mak, presents a means for which the bacterium can store excess trehalose. GlgE and the other enzymes of the pathway had been implicated since the 1990’s as being essential for trehalose cycling in vivo, but until their identification recently, this was only conjecture 49.

2.1.3 Interplay of the glucan producing pathways

Despite the difference in enzymes and substrates, little is known about the interplay and cross talk between the three pathways present in M. tb. It is highly likely there is some functional redundancy among the different pathways, as well as distinct functionalities for each. Indeed, disruption of enzymes within each pathway produces a diverse set of phenotypes within the M. tb bacterium. Knockout of either the Rv3032 or GlgE pathways

(Δrv3032 or ΔtreS) produces a bacterium that is viable and has little to no effect on growth 39.

However knock out of both simultaneously showed these pathways to be synthetically lethal.

This suggests that there is some redundancy between the Rv3032 and GlgE pathways 39. This is not observed in the knock out of the GlgA (ΔglgC) and Rv3032 pathways, as the organism is viable 39.

Looking at levels of the capsular glycogen produced from gene deletions produces a variety of phenotypes. Bacteria lacking GlgC or GlgA (ΔglgC or ΔglgA) show a distinct lack of the intracellular glycogen. However, only the ΔglgA mutant showed a marked decrease in capsular glycogen levels, suggesting that the classical pathway is the primary producer of capsular glycogen 39. This is evidenced by the fact that Δrv3032 showed no change in either intracellular or capsular glycogen levels. Bacterium lacking GlgE (ΔglgE) produced a bacterium that was non-viable and quickly died. This interesting phenomenon identified

GlgE as an intriguing target for therapies designed to treat M. tb infections. Despite all this, little is known about how much the GlgE pathway contributes to the other glucan producing pathways, and what the ultimate destination for the glucan produced by GlgE is.

2.1.4 The GlgE pathway and its significance for drug discovery

Unlike the effects observed in Δrv3032 or ΔglgA deletions, deletion of the glgE gene in M. tb results in the rapid killing of the bacterium. This is not the result of lack of glucan produced by the pathway; rather, it is due to the accumulation of the M1P substrate 47. Once

M1P concentrations reach a critical level, it becomes toxic, and rapidly causes cellular death.

Additionally, the increase of M1P concentration elicits an apparent stress response by the bacterium via the stimulation and the over expression of biosynthetic enzymes necessary for the production of trehalose which acts as a stress protectant. This ultimately results in the increased production of more M1P 47. This positive feedback loop and overproduction of

M1P causes pleiotropic effects that cause rapid bacterial death 47. This effect is novel in that killing is the result of an over production of a toxic metabolite rather than the absence of an important metabolite. Because of this rapid and novel mechanism of killing, efforts to discover GlgE inhibitors may afford the development of potent compounds that rapidly kill 41

both active and dormant M. tb. As such, designing inhibitors against GlgE could help forestall the emergence of antibiotic resistance by the efficient and rapid killing of the bacterium.

2.2 Material and Methods

2.2.1 Molecular cloning

H37Rv M. tb genomic DNA was used to PCR amplify the wild-type glgE (Rv1327c) gene using the primers in Table 2. The resulting product was placed between the NdeI and

BamHI cut sites of a modified pET-28 plasmid. The resulting pDR28-glgE encodes a recombinant GlgE enzyme possessing an N-terminal histidine tag. The sequence of the glgE expression plasmid was confirmed by DNA sequencing. Several mutants of glgE as well as a truncated form of the enzyme were made. Mutant forms of glgE, glgE-D383N and glgE-

D418N, were cloned using site-directed mutagenesis using the pDR28-glgE as a template and the primers in Table 2.

Table 2: Primers used for PCR amplification of the various glgE mutants glgE-5’ CAC CAT ATG AGT GGC CGG GCA AT glgE-3’ AAA GGA TCC TCA CCT GCG CAG CA glgE-D383N CAC CGA AGA AGT ACC AGA ACA TCT AT glgE-D418N CAA GTT CTT TCG CGT CAA CAA TCC CC

2.2.2 Expression and purification

The pDR28-glgE plasmid was used to transform T7 Rosetta cells. The bacterial cells were cultured at 37 °C in Luria Broth to an O.D. of 0.6 at 600 nm. Protein expression was 42

induced by the addition of IPTG to a final concentration of 1 mM. Cells were harvested by centrifugation after incubating for 24 hours at 16 °C. Pelleted cells were re-suspended in buffer A containing 20 mM Tris pH 7.5, 5 mM imidazole, 500 mM NaCl, 10% glycerol, and

5 mM -mercaptoethanol. Lysozyme (10 μM) and DNaseI (100 μM) were added to the cell re-suspension and incubated for one hour on ice prior to lysis by sonication. The resulting suspension was centrifuged at 15,000 g for 30 minutes. The supernatant was applied to a 5 mL HiTrap Talon Crude column (GE Healthcare) that had been equilibrated with buffer A.

Proteins were eluted from the column via a linear gradient of imidazole from 5 – 150 mM over 20 column volumes. Fractions containing GlgE were pooled, concentrated by ultrafiltration and applied to a Hi-Load Superdex 200 size exclusion column (GE Healthcare) equilibrated with a buffer containing 20 mM Tris pH 7.5, 150 mM NaCl, 1 mM MgCl2, and

0.3 mM TCEP. Fractions containing GlgE were subsequently pooled. The purity of the protein was confirmed using SDS-PAGE. Mutant forms of GlgE were expressed and purified using the same protocol as the wild-type GlgE.

2.2.3 Quinaldine Red assay for GlgE activity

Stock solutions of 0.05% w/v Quinaldine Red in water, 2 % w/v polyvinyl alcohol in water, 6% w/v ammonium molybdate tetrahydrate in 6 M hydrochloric acid were prepared.

The solutions were mixed 2:1:1:2 ratios with deionized water to produce the QRed working solution, which would be used for the phosphate detection in the assay. Fresh working solution was always prepared prior to performing assays.

Enzymatic reactions were carried out in 50 µL volumes at 37 °C on a synergy H4

Hybrid Reader (BioTek) using 50 nM of enzyme. Buffer conditions of the assay were as

follows: 50 mM Bis-Tris propane pH 7.5, and 50 mM NaCl. Oyster glycogen (Sigma-

Aldrich) was used as an acceptor molecule. A stock solution of glycogen was prepared to a final concentration of 200 mg/mL. This solution was extensively dialyzed against deionized water to remove the residual phosphate that was present in the solid. The glycogen solution was added to the reaction mixture to a final concentration of 10 mg/mL. Kinetic parameters were determined by varying the concentration of M1P from 0-5 mM. All reactions were performed in triplicate. Endpoint reads were carried out from 0 to 10 minutes. Aliquots of each reaction were quenched with 140 µL of the QRed working solution every minute.

Quenched reactions were incubated for 2 minutes at 37 °C after which 14 µL of a 32% w/v sodium citrate solution was added. This final solution was left to incubate for an additional

15 minutes at 37 °C. Final reactions were placed in a white 384-well plate. Phosphate release was measure by fluorescence using excitation and emission wavelengths of 430 nm and 530 nm respectively. The Prism 5 software was used to perform non-linear regression analysis of data.

The NIH Clinical Collection screening was carried out in the same manner as described above. 0.5 µL of 10 mM stocks of each drug were added to each reaction to give a final concentration of 100 µM of compound. 100% DMSO was added to both negative and positive control reactions to account for the added DMSO in the compound samples.

Reactions were carried out over 4 minutes and quenched. Fluorescence values were measured as described previously.

2.2.4 Crystallization and structural determination of GlgE complexes

Crystals of GlgE were grown using the hanging-drop vapor diffusion method. The

Index Screen (Hampton Research) or The Classics Lite Suite (Qiagen) were set up using a range of protein concentrations. Additionally, the poly-Histidine tag was either removed or left on the purified protein. Once a preliminary hit was obtained, further optimization was attempted by varying the concentrations of the compounds in the condition, or by the use of the Detergent and Additive screens (Hampton Research).

Diffraction quality crystals of GlgE were obtained using the conditions below. GlgE at 5.3 mg/mL using a 0.1% Abs value of 1.908 was used for crystallization experiments.

Maltose was added to the concentrated GlgE to a final concentration of 10 mM. Crystals of

GlgE were grown by the hanging-drop vapor diffusion method. Crystallization drops containing 2 μL of GlgE solution and 2 μL of well solution were equilibrated with 100 μL of well solution. Initial crystals were found in a condition containing 0.1 M HEPES pH 7.5 and

1.0 M ammonium formate after roughly 1 month. Improved crystals were produced by the addition of 0.4 μL of either 10% 1,2 butanediol or 40% v/v 2,2,2, trifluoroethanol to the crystallization drop. 1 μL of glycerol was added to the crystallization drop prior to flash- cooling the crystals in liquid nitrogen.

To produce the GlgE-M6 complex, crystals of the GlgE-MAL complex were produced as previously described. These GlgE crystals were soaked with maltohexaose (10 mM) for 1 hour prior to flash-cooling the crystals in liquid nitrogen in preparation for diffraction experiments. One μL of glycerol was added to the crystallization drop prior to flash-cooling of crystals in liquid nitrogen.

X-ray diffraction experiments were carried out at the LS-CAT beamline at the

Advanced Photon Source of Argonne National Labs, Argonne, IL. M. tb GlgE structures 45

were solved using diffraction data collected at a wavelength of 0.97856 Å. Diffraction data were integrated and scaled using HKL3000 50. Initial indexing of the Mtb GlgE diffraction data was not unambiguous. These data were integrated using all possible Bravais lattices and subsequently scaled. The highest symmetry space group that produced Rsym values less than

50 % was of the C2 space group. Following scaling, the C2 diffraction data were subjected to analysis using the Xtriage tool in PHENIX 51. This analysis indicated no evidence of twinning.

The structure of the Mtb GlgE-MAL was phased using molecular replacement in

PHENIX 51, 52. The Mtb GlgE-MAL used the monomer of the Sco GlgEI (3ZT5) as a search molecule. The GlgE-M6 structure used the GlgE-MAL structure as a model for rigid body refinement. Rigid body refinement, simulated annealing, positional and B-factor refinements were carried out using PHENIX 53. Manual refinement of all structures was performed using

COOT 54. Ligand and ligand restraints were generated using eLBOW software in PHENIX.

Structural validation was performed using Molprobity 55, 56.

2.3 Results and Discussion

2.3.1 Optimization of the fluorescence based Quinaldine Red assay for GlgE activity assessment

2.3.1.1 Kinetic characterization of GlgE using Quinaldine Red assay for phosphate detection

GlgE is a member of the GH13 α-amylase family; specifically it is a maltosyltransferase that utilizes M1P as a donor molecule 47, 57. This activated sugar is used as a maltosyl-donor in the transfer to an α-glucan, thereby extending it. In addition to the

extended glucan, inorganic phosphate is also produced (Figure 4A). This leads to two avenues of assessing GlgE activity: qualitatively by measuring increasing glucan size as a function of time using mass spectrometric techniques, or quantitatively by measuring the release of inorganic phosphate as a function of time. However, to determine steady-state

GlgE kinetics, assessing the activity of GlgE mutants, or for inhibition studies, quantitative data is needed. Therefore, detecting inorganic phosphate release was determined to be the method of choice.

Numerous methodologies exist to detect phosphate release. We employed the use of the fluorescence based, Quinaldine Red (QRed) assay to detect inorganic phosphate. This assay uses an acidified solution of ammonium molybdate and the Quinaldine Red dye

(Figure 4C). In the presence of inorganic phosphate, the dye solution forms a molybdate- phosphate-Quinaldine Red complex (Figure 4D). The solution is colorless initially, but

58 changes to a red color (λmax = 530 nm) in the presence of phosphate . At this point, absorbance could be used to measure free phosphate, as the production of the complex is concentration dependent. However, using white plates allows for the use of fluorescence

(Figure 4B). When the white plates are exposed to light, (λ = 430 nm) they re-emit at λmax =

530 nm, which is identical to the λmax of the chromophore complex. As a result, we can measure the fluorescence quenched by the chromophore, making this assay much more sensitive than using absorbance 58.

This assay was adapted for use in determining the steady-state kinetics of GlgE.

Previous studies determined the kinetics of the Mtb GlgE, as well as the kinetics of both the

Sco GlgEI and GlgEII 47, 57. In these studies the malachite green assay was employed, and

maltohexaose as an acceptor substrate. For our determination, we wanted to determine if using the Quinaldine Red assay would provide a more accurate assessment of GlgE activity.

In addition to this, we used oyster glycogen, which is a more analogous substrate to the M. tb

α-glucan, to determine how this would alter the enzyme kinetics.

Figure 4: Quinaldine Red based assay for phosphate detection.(A) GlgE releases inorganic phosphate as a product (B) which can be detected using the fluorescence based Quinaldine Red white plate method. (C) Structure of Quinaldine Red. (D) Quinaldine Red reaction: Inorganic phosphate complexes with ammonium molybdate under acidic conditions. The ammonium molybdate-phosphate complexes with Quinaldine Red forming a red colored solution product.

Reactions for the QRed assay were optimized using 50 nM of enzyme in a buffer containing 50 mM Bis-Tris propane pH 7.5 and 50 mM NaCl. Oyster glycogen was used at a final concentration of 10 mg/mL. The glycogen needed to be extensively dialyzed against 48

deionized water, as there was a substantial amount of phosphate contamination present.

Initial assay validation used 1 mM of M1P as the substrate concentration. Progress curves were obtained by quenching aliquots of the reaction every 60 seconds for 7 minutes, and the emission at 530 nm was measured (Figure 5). A linear decrease in fluorescence was observed for the first four minutes of the reactions indicating the production of phosphate. In the absence of GlgE, slight quenching was observed, most likely due to slow background hydrolysis of the M1P.

30000

25000

20000

15000 RFU 10000

5000

0 0 1 2 3 4 5 6 7 8 Time (minutes)

Figure 5: GlgE Quinaldine Red progress curve. Initial test of GlgE activity using the Quinaldine red assay. The total RFU decreases overtime due to the quenching effect of the chromophore formed by the release of inorganic phosphate.

Once confirming enzymatic activity using the QRed assay we moved on to determining the steady-state kinetics for GlgE. For this procedure, the concentration of glycogen was held constant at 10 mg/mL, while the concentration of M1P was varied from

0–5 mM. Reactions were performed in triplicate and progress curves for each reaction were determined as previously described, except that reactions were quenched every minute for 10

minutes. Initial rates were determined with the Beer-Lambert law using a prepared standard curve of phosphate quenched with the QRed. Fitting the initial rates of each reaction against substrate concentration to the Michaelis-Menten model was done using the Prism 5 software

(Figure 6).

Vmax and Km were calculated to be 132.7 ± 4.6 µM phosphate/min and 744.8 ± 1.4

2 -1 µM M1P respectively (R = 0.9794). The kcat was calculated to be 44.22 ± 1.45 s , while the

-1 -1 second-order rate constant of kcat/Km was determined to be 59371 ± 5962 M s . The determined Km for M1P is one third the value previously determined for GlgE using the

Malachite Green assay (Table 3).

Figure 6: Michaelis-Menten curve for M1P. The M1P concentration was varied from 0-5 mM, while the glycogen concentration was held constant at 10 mg/mL. Error bars (standard deviation) were calculated from triplicate reactions.

One interesting observation was the nearly 10 fold increase in kcat and kcat/Km using the glycogen substrate. This may be attributed to the fact that glycogen has orders of magnitude 50

more free 4-OH moieties due to the branched nature of the substrate. This significantly increases the effective concentration of maltosyl acceptor in the reaction and thereby increases the rate of enzyme turnover. Additionally, once linear glucans reach sizes greater than 14 glucosyl units, the ability of GlgE to transfer to these longer glucans is greatly diminished 57.

Table 3: Comparison of the kinetic parameters of GlgE enzymes Comparison of the kinetic parameters of GlgE enzymes determined by the Quinaldine red (QRed) and malachite green (MG) based assays from Mycobacterium tuberculosis and Streptomyces coelicolor. Km, enzyme turnover, and catalytic efficiency calculated from the Quinaldine red assay all compare favorably with the values determined for the corresponding protein in each organism with the malachite green assay.

-1 -1 -1 Enzyme Acceptor M1P Km (µM) kcat (s ) kcat/Km (M s ) Assay

M. tb GlgE Glycogen 744 ± 71 22.4 ±1.5 59371 ±5963 QRed

M. tb GlgE Maltohexaose 250 ± 50 1.26 5000 MG

S. co GlgEI Maltohexaose 300 ± 60 12.3 41000 MG

S. co GlgEII Maltohexaose 1200 ± 200 10.0 8000 MG

2.3.1.2 Assessing the quality of the Quinaldine Red assay for high-throughput screening applications

One common methodology to determine if an assay is suitable for high-throughput screening applications is to determine a Z’ score. Simply put, a Z’ uses a large set of reactions with enzyme (c+) and those without (c-) enzyme with substrate concentrations near

59 the determined Km . The Z’ can be calculated with Equation 1:

3휎 + 3휎 푍′ = 1 − 푐+ 푐− |휇푐+ − 휇푐−|

Equation 1: Z' calculation

In the equation σc+ and μc+ represent the standard deviation and average signal of a control reaction containing enzyme, while σc- and μc- are the standard deviation and average of the negative control without enzyme. From the equation, a perfect assay would have a Z’ equal to one, however any assay which has a Z’ greater than 0.5 is considered suitable for screening purposes 59. Therefore, any assay optimization scheme aims to minimize the standard deviation of its positive and negative control reactions, while maximizing the signal to noise between the two.

Using the assay conditions obtained from the optimized kinetics determination, a calculation of the Z’ value for the Quinaldine Red assay was performed. Using 750 μM of

M1P, 48 positive reactions and 48 negative reactions were performed. For this set of experiments, we measured the RFU at the endpoint of the reaction after four minutes. From these reactions the calculated Z’ was determined to be 0.672 (Table 4). From this determined

Z’, the Quinaldine Red assay appears to be suitable for HTS applications.

Table 4: Z’ Determination of the GlgE Quinaldine Red assay

Standard Deviation Average Signal (RFU)

(+) Control 278.8 1744.13

(-) Control 1508.6 18118.3

Z’ = 0.673

2.3.1.3 NIH Clinical Collection Screening of GlgE

The NIH Clinical Collection (NCC) is a collection of 450 drug compounds that have been in phase I-III clinical trials and have shown some efficacy, bioavailability, and are considered drug-like, which can be purchased from Biofocus.

The NCC screen was used to identify potential lead compounds that may inhibit the activity of GlgE. Initial NCC screening used the Quinaldine Red assay and the reaction conditions established from during the Z’ determination. Compounds in the screen are concentrated to 10 mM and dissolved in 100% DMSO. These were diluted to 100 μM in each reaction for the initial screening. NCC compounds were screened in groups of 20 in addition to two positive, and two negative controls to give a total of 24 reactions per set. This was done in order to optimize the number of reactions performed, while making it manageable to quench all reactions and read the RFU at the endpoint.

All 450 compounds were screened for inhibitory activity towards GlgE; however none appeared to inhibit the enzyme from a cursory screen. Despite this fact, there may be some inhibitors possessing some efficacy toward inhibiting GlgE. One potential issue with the Quinaldine Red assay is the use of a molybdate-phosphate complex quenching. Despite 53

the sensitivity of the assay, interfering compounds which absorb near or at the complex will cause issues. Indeed, there are many compounds in the NCC screen which contain chromophores, some of which are red in color. These would interfere with the assay leading to incorrect data, most likely false negatives. It is possible that these compounds could have been potential inhibitors, but they are not easily tested within the scope of this assay.

Therefore, testing them using an orthogonal assay, such as measuring elongation of acceptors using mass spectrometry could be employed. However, this is non-quantitative, and does not lend itself to HTS. Therefore another assay would need to be developed to potentially test these compounds. We later developed a new assay for GlgE activity, which will be discussed in Chapter 3.

2.3.3 Determination of the wild-type M. tb GlgE structure in a binary complex with maltose (Text and figures reproduced from Lindenberger et. al., Sci. Rep., 2015)

2.3.3.1 Crystallization strategies for the Mtb GlgE-maltose

To gain greater insight into the biochemical workings of the M. tb GlgE, we attempted to crystallize, and determine the structure of the enzyme. Because the mechanism of GlgE had yet to be determined, it was hypothesized that the enzyme may require a divalent metal to coordinate the phosphate moiety of the M1P substrate. Therefore, the buffer used for GlgE crystallization contained MgCl2. This was later shown to be unnecessary.

However, the addition of the MgCl2 was necessary for the formation of diffraction quality crystals. Additionally, in the absence of any substrate analogues, or the identification of inactive mutants, maltose was added to the enzyme to mimic the maltosyl moiety of M1P and potentially enhance the likelihood of crystallization by stabilizing the protein structure.

Initial crystals of the protein were obtained from the Index Screen in a condition containing 0.5 M magnesium formate and 0.1 M HEPES pH 7.5. These crystals were very small and had a cubic morphology (Figure 7A). However, these crystals diffracted to very low resolutions (> 6 Å), and exhibited long-range disorder as evidenced by severely streaked reflections. Attempts to improve these crystals were performed using different formulations of the described condition in addition to using the Additive and Detergent screens. One detergent identified from the Detergent screen, a dodecyl-maltoside, improved the size of the crystals; however, the resolution range was still insufficient to determine the structure

(Figure 7B).

Figure 7: Initial GlgE crystals (A) Initial hit was found in a condition containing 0.5 M magnesium formate and 0.1 M HEPES pH 7.5. (B) Crystals were improved by the addition of dodecyl-maltoside; however, diffraction data obtained from them was of very low quality.

Therefore, a secondary screen was performed using the Classics Lite screen. One condition was found that produced crystals. These crystals were of a similar morphology to those initially obtained from the Index Screen. The crystallization condition contained 1 M ammonium formate and 0.1 M HEPES pH 7.5 (Figure 8A). Different formulations of this 55

condition were attempted, however, the initial condition was deemed to be the most optimal.

Both additive and detergent screens were performed again to improve crystals. Two additives,

1,2 butanediol, and 2,2,2, trifluoroethanol, when added to the crystallization drops, produced much larger crystals than those previously obtained (Figure 8B). Both additives yielded very similar crystals in terms of quality and resolution. These crystals were used to solve the initial GlgE structure.

Figure 8: Improved GlgE crystals (A) Crystals were obtained from a solution containing 1 M ammonium formate and 0.1 M HEPES pH 7.5 (B) Crystals were further improved by the addition of either 1,2 butanediol or 2,2,2, trifluoroethanol. The crystal in panel B was the largest one obtained and denoted “Big Daddy”.

2.3.3.2 Data collection and structural refinement of the GlgE-maltose structure

The crystals of the M. tb GlgE were analyzed at Argonne National Labs and diffraction data extended out to 3.3 Å. The initial integration and scaling of the GlgE data was difficult. Initial indexing showed that the protein crystallized in a cubic space group.

However, no solutions could be obtained. The data were then scaled and integrated using all possible Bravais lattices. The highest symmetry space group which produced an Rsym value of less than 50 % after integration and scaling was C2. Subsequent analysis of the data 56

showed no evidence of twinning. The C2 space group had the following unit cell parameters: a = 342.2 Å, b = 242.6 Å, c = 243.7 Å; α = γ = 90 °, β = 135.1 (Table 5). It was known that

GlgE likely had a TIM barrel architecture based on homology to the glycoside hydrolases.

Therefore, molecular replacement using a TIM barrel from an α-glucosidase may have been successful for phasing. However, the structure of a GlgE ortholog had just been published and deposited 57. Instead, the structure of the Streptomyces coelicolor GlgEI (PDB accession code: 3ZT5) was used as the search model.

Table 5: Data Collection and Refinement statistics for GlgE-MAL and GlgE-M6

Data Collection Mtb GlgE-MAL Mtb GlgE-M6 Space Group C2 C2 Unit Cell Dimensions a, b, c (Å) 343.2, 242.6, 243.7 338.4, 239.4, 239.4 α, β, γ (°) 90, 135.1, 90 90, 134.1, 90 Resolution (Å) 50-3.3 50-3.9

Rmerge 11.7 (77.6) 18.2 (94.7) I/σI 15.0 (2.5) 13.8 (2.2) Completeness (%) 99.9 (100) 99.7 (94.7) Redundancy 6.6 (6.3) 7.2 (6.5) Refinement Resolution (Å) 47.6-3.3 43.7-3.9 No. Unique Reflections 210,465 122,284

Rwork/Rfree 0.1937/0.2220 0.2235/0.2564 No. Atoms Protein 31,458 31,398 Water n/a n/a Ligand 138 540 B-factors (Å2) Protein 72.3 117.4 Water n/a n/a Ligand 83.8 125.2 R.m.s deviations Bond lengths (Å) 0.016 0.008 Bond angles (°) 1.41 1.65 Ramachandran Favored (%) 96.2 95.7 Outliers 0.9 1.1

2.3.3.3 Analysis of the Mtb GlgE-maltose structure

The crystal structure of the wild type M. tb GlgE bound to maltose (Mtb GlgE-MAL)

was refined to 3.3 Å resolution. GlgE exists as a homodimer, with the final refined model having six molecules in the asymmetric unit: 2 sets of dimers and two monomers. Both of the monomers formed dimers with the monomers in the adjacent unit cells. This dimer appears to be biologically relevant, as size exclusion chromatography and gel based analysis indicate a dimer. Most of the protein was resolved in the structure except for a few select regions. The

N-terminus of the protein, which contains the uncleaved poly-Histidine tag and PreScission

Protease cleavage sequence, is mostly unresolved and density for any residues does not appear until P15. The C-terminus of the protein is almost completely resolved with only the final two residues, R700 and R701, unresolved. Specific internal loop regions of the protein are also unresolved, and these will be discussed in further detail later. The overall homodimeric structure of the Mtb GlgE is very similar to the previously reported Sco GlgEI enzyme. Both share five-domain architecture, with two unique insert regions. The five domains: domain A, domain B, domain N, domain S, and domain C form the overall structure, and inserts 1 and 2 of domain A complete the overall structure (Figure 9A) 57.

Domain A (residues 232-276, 323-345, 391-535, 576-596) of the Mtb GlgE forms the catalytic TIM barrel, as well as forming part of the dimer interface. This TIM barrel domain is common for the GH13 glycoside hydrolase family to which GlgE belongs 60. Domain A also contains two insert regions (insert 1, residues 277-322; insert 2, residues 536-575). Insert

1 follows the second β-strand of domain A, while Insert 2 appears after the eighth β-strand.

Both inserts are adjacent to each other in three-dimensional space.

Figure 9: Mtb GlgE overall structure. (A) Mtb GlgE labeled by domain. Domain N (residues 1-126, 215-231; magenta), Domain S (residues 127-214; yellow), Domain A (residues 232-276, 323-345,391-535, 576-596; green), Insert 1 (277-322; blue), Domain B (residues346-390; red), Insert 2 (residues 536-575), and Domain C (residues 597-702; pink). (B) Superposition of the Mtb GlgE-MAL (green) and Sco GlgEI (cyan) (3ZT5) gives an R.M.S. displacement value of 2.6 Å for the C atoms.

Domain B (residues 346-390) appears as an insertion after the third β-strand of

Domain A, and is comprised of a set of anti-parallel β-strands, followed by a single, short α- helix. Domain B is again, very typical of the GH13 α-amylase family 61. In some cases,

Domain B is utilized to bind calcium ions by some α-amylases 61. However, no evidence of a divalent metal ion in either our structures or those structures published by Syson et al. show density for a metal in this domain. Additionally, Syson et. al. showed that the enzymatic activity of GlgE is independent of any addition of divalent metals (MgCl2, CaCl2), or the addition of metal chelators such as EDTA or EGTA 57. Domain A and domain B, as well as

the insertions of domain A, define the overall catalytic domain of GlgE.

Domain N (residues 1-126, 215-231) is a β-sandwich structure and is at the core of the dimer interface between subunits. C29 of domain N also forms an intermolecular disulfide bridge with the C29 of the adjacent monomer 62. Domain C (residues 597-702), also possessing a β-sandwich structure, appears to play a role in stabilizing domain A of the adjacent monomer and binding of the maltosyl acceptor substrate 57. Domain S (residues 127-

214), which is unique to the GlgE family of enzymes, is an insertion into Domain N, and is composed of a 4 helical bundle 57. Domain S plays a role in forming the dimer interface with the neighboring domain B of the second subunit, and may play a role in acceptor substrate binding. Because of this highly conserved architecture, superimposing the Sco GlgEI and

Mtb GlgE-MAL using t

(Figure 9B).

In contrast to these similarities, there are some minor differences observed between the structures. First, there are three insertions of 2-4 residues and a much larger insertion seen in the Mtb GlgE that are lacking in Sco GlgEI. The largest insertion in the Mtb GlgE is a 21- residue loop encompassing residues 71-92 in domain N (Figure 10A). This loop is dynamic, and as a result, almost no density is observed for this loop in the Mtb GlgE crystal structure.

These loops are not found within the active site and are therefore unlikely to directly affect enzymatic activity. However, phosphorylation by the essential Mycobacterium tuberculosis kinase PknB at S82 within this structurally dynamic 21- residue loop has been shown to decrease GlgE enzymatic activity 63. Determination of whether this inactivation mechanism is due to structural rearrangement in the active site or a decrease in substrate binding affinity

will require additional experiments. The latter seems highly unlikely since S82 is located on the side of the GlgE dimer opposite to that of likely glucan binding region.

Figure 10: Dynamic loop and active site of the Mtb GlgE. (A) Loop region from the Mtb GlgE (residues 71-92 colored in dark blue) indicated by arrows, is one of the distinct features not seen in the Sco GlgEI. (B) 4-helical bundle S domain. This missing loop region (residues 144-149) is indicated by arrows.

Another difference between the Mtb and Sco GlgE structures is found in the four helical bundle forming the S domain. This region in the Mtb structure appears to be much more dynamic in the crystal structures of the Mtb GlgE than the Sco GlgE as evidenced by the significantly elevated B-factors. This is very evident by the absence of a small loop region in this domain. This turn between α1-α2 (residues 144-149) is highly dynamic and as a result, no difference density for it is observed (Figure 10B). The dynamics observed in this turn afford phosphorylation of T148 by PknB and this post-translational modification negatively regulates the activity of GlgE both in vitro and in vivo 63. Once again, the mechanism of this negative regulation is unknown; however, phosphorylation of Thr148 may modulate binding of the glucan since the S Domain that harbors residue T148 abuts the M1P binding site in domain A. 62

One additional minor difference between the Sco and Mtb GlgE orthologs is that the

Mtb GlgE crystal structures possess an intermolecular disulfide bridge between residue C29 of the two molecules in the GlgE dimer. The C29 residue is not conserved in the majority of the deposited GlgE protein sequences and GlgE orthologs form a stable homodimer when lacking a disulfide bond. Therefore, it is unlikely that the observed disulfide bond plays any important physiological role.

In contrast to these differences, the maltosyl donor sites of both enzymes are highly conserved in sequence and structure. This site is divided into two parts: the -1 subsite where the reducing end of maltose or the phosphoglucosyl moiety of the M1P substrate binds, and the -2 subsite where the non-reducing glucosyl moiety is bound 64. The enzymes exhibit a similar set of hydrogen bonding interactions that coordinate the maltose and position it within the maltosyl-donor site (Figure 11A).

Figure 11: Active sites of the Mtb GlgE and Sco GlgEI. (A) Superimposed active sites show the highly conserved architecture. (B) The only significant change to the active site residues is Ser303 in the Mtb GlgE while in the Sco GlgEI it is a valine (Val279)

The only residue that differs between the Mtb GlgE and Sco GlgEI maltosyl donor sites is

S303 (using Mtb GlgE numbering), which is positioned near the 6-OH of the -2 sugar in the

Mtb GlgE-MAL complex and forms a hydrogen bonding interaction with the endocyclic O5 oxygen atom. In the Sco GlgEI, this serine has been replaced by a valine (V279), but is positioned in the same orientation as the Mtb GlgE S303 residue (Figure 11B). Rather than forming a hydrogen bond with substrate, V279 of Sco GlgEI forms a van der Waals interaction with two carbon atoms in the ring bound in the -2 subsite. This difference affects neither the active site structure nor the apparent steady-state kinetic parameters observed for

Mtb GlgE, Sco GlgEI, or Sco GlgEI-V279S 57, 65, 66.

2.3.4 Identification of a high-affinity maltohexaose-binding site in the M. tb GlgE (Text and figures reproduced from Lindenberger et. al., Sci. Rep., 2015)

2.3.4.1 Crystallization strategies for the Mtb GlgE-maltohexaose

GlgE utilizes an -1,6-branched, -1,4 glucan as a maltosyl-acceptor molecule and thereby extends the linear portion of the glucan using M1P as the maltosyl donor 47. To further characterize the maltosyl acceptor binding site and the mode of GlgE binding to linear polysaccharides we attempted to crystallize GlgE in the presence of maltohexaose (M6), a multimer of six, -1,4 linked glucoses. Initial attempts to co-crystalize the compound with

GlgE and maltose resulted in either poorly diffracting crystals, or no observable density for the maltohexaose. Syson et. al. also encountered this problem when attempting to co- crystallize maltohexaose with the Sco GlgEI. One potential issue is that GlgE in the presence of linear glucan acceptor, but no donor, can perform a transglycosylation reaction, whereby it rearranges larger glucans into both smaller, and larger ones 57. They rationalized that co- crystals of GlgEI and maltohexaose over the duration of the crystallization process led to a heterogeneous mixture of small and large glucans, thereby preventing crystallization outright or leading to no observable density for any glucan 57. Therefore, in lieu of co-crystals of

GlgE and maltohexaose, we performed soaking experiments. Here, we soaked Mtb GlgE-

MAL co-crystals with maltohexaose immediately prior to performing X-ray diffraction experiments. This ultimately proved to be much more successful.

2.3.4.2 Data collection and structural refinement of the Mtb GlgE-maltohexaose structure

A crystal structure of a ternary complex with Mtb GlgE bound to maltose and maltohexaose (GlgE-M6) diffracted to a much lower, 4.0 Å resolution, than was previously obtained. The protein, which crystallized in the C2 space group, had the following unit cell

parameters: a = 338.4 Å, b = 239.4 Å, c = 239.4 Å; α = γ = 90 °, β = 134.1 (Table 5).

Indexing and scaling of the GlgE-M6 structure was performed in the same manner the Mtb

GlgE-MAL structure. Initial rigid body refinement was performed utilizing the previously refined GlgE-MAL structure. Initial inspection of the Fo-Fc maps calculated following rigid body refinement revealed a horseshoe-shaped density that was more than six times above background 67. This density, when fit with the maltohexaose ligand, corresponded to a single molecule of maltohexaose (Figure 12).

Figure 12: M6 docking site of GlgE. The M6 binds to loop regions primarily made up by domains A and C, which form part of the active site.

2.3.4.3 Structural analysis of the Mtb GlgE-maltohexaose structure

The maltohexaose binds to multiple loop regions of domains A and C that define a portion of the acceptor site; specifically, maltohexaose binds the short loop regions that connect secondary structural elements encompassing residues 465-481, 504-514, and 622-

632. Specific binding interactions between the Mtb GlgE and maltohexaose are mediated primarily through hydrogen bonding interactions with the backbone atoms of residues that encompass the donor site, including the backbone carbonyls of T474, N512, and L628. Two amino acid side chains also contribute to maltohexaose binding. Specifically, residue N629 in domain A forms a hydrogen bond through the carbonyl of the carboxamide moiety, while the side chain of F631 forms van der Waals interactions with the fourth sugar of the maltohexaose (Figure 13).

Figure 13: A high affinity docking site for M6. Fo-Fc omit map calculated while omitting maltohexaose is contoured at 3 showing the binding of maltohexaose (M6) to the Mtb GlgE. Binding is mediated by backbone carbonyls of T474 of the B domain, N512 and L628 of the A domain, and the carbonyl of the sidechain carboxamide group of N629 of the A domain. F631 of the A domain interacts with the fourth sugar via van der Waals interactions.

The dearth of specific side chain interactions with maltohexaose suggests that conservation of the structure of the protein backbone atoms at this site is required for glucan binding, but amino acid sequence conservation may not be as important. Comparing the residues of the maltohexaose binding sites in the Sco GlgEI and Mtb GlgE structures exhibits approximately 74% sequence identity between the two proteins. The amino acid differences 68

between the two enzymes within this site include residues T474, N512, and F631 of the Mtb

GlgE, which are represented by residues N450, G489, and R606 in the Sco GlgEI enzyme, respectively. Based on the interactions observed in the Mtb GlgE-M6 structure, the variation in side chain identity for these three residues should not significantly impact maltohexaose binding since the interactions are mediated by backbone atoms (T474 and N512), or through non-specific van der Waals interactions (F631) (Figure 14) 68.

Figure 14: Maltohexaose binding site sequence alignment between the Sco GlgEI and Mtb GlgE. Areas defined by the magenta boxes and the M6 symbol indicate the residues forming the maltohexaose binding site observed in Mtb GlgE, while the teal circle indicates the cyclodextrin binding surface observed in Sco GlgEI. The orange box indicates residues important for binding cyclodextrin in Sco GlgEI but differ in Mtb GlgE.

Surprisingly, the maltohexaose is situated approximately 26 Å from the M1P binding site, a distance that requires approximately three additional maltohexaose units to span. Therefore,

the high affinity maltohexaose binding site observed in the Mtb GlgE-M6 structure likely represents only a small portion of a much larger -1,6-branched, -1,4 glucan binding site.

While sequence conservation between the Mtb and Sco GlgE enzymes at the maltohexaose site is 74%, the surface leading from the high affinity site into the enzyme active sites are nearly identical in both amino acid sequence and three dimensional structural. This suggests that conservation here is very important for correctly orienting the large glucans representing the maltosyl acceptor substrate.

It had been shown previously that cyclodextrins inhibit the Sco GlgEI weakly, but show no inhibition towards the Mtb GlgE. Structures of the Sco GlgEI bound to both - cyclodextrin (IC50 = 19 mM) and -cyclodextrin (IC50 = 6 mM) have been solved (RCSB accession numbers: 3ZT6 and 3ZT7) 57. In these structures, the cyclodextrins sit at the homodimer interface in a largely hydrophobic site formed by Domain A of one subunit and

Domain N of the second subunit. The site is orthogonal to the M1P binding site and adjacent to the observed maltohexaose binding site as seen in the Mtb GlgE-M6 structure, but the maltohexaose and cyclodextrin binding sites do not overlap (Figure 15). It was hypothesized that cyclodextrin inhibition of the Sco GlgEI is likely due to partial occlusion of the acceptor glucan binding site and prevention of the glucan substrate from accessing the enzyme active site 57.

Figure 15: Surface conservation comparison of the Mtb and Sco GlgE enzymes. Blue surface indicates identical residues, light-blue semi-conserved changes, and white no sequence similarities. The second subunit of the GlgE dimer uses grey to indicate lack of sequence similarity. The orange surface represents changes observed at the cyclodextrin binding pocket. Maltohexaose is shown as yellow spheres and -cyclodextrin is shown with cyan bonds. Sequence alignment was performed using ESPript 3.0 (http://espript.ibcp.fr) 69

The Mtb GlgE structures presented here offer further insight to the differences between the Sco and Mtb GlgE orthologs that likely preclude binding of the cyclodextrins to the Mtb GlgE. The observed cyclodextrin binding site in Sco GlgEI differs from the Mtb

GlgE ortholog at two amino acid positions: residues R427 in one protein subunit and G84 from the neighboring subunit of the Sco GlgEI are both replaced by proline residues in the

Mtb GlgE structure. The resulting structural changes within this region likely abolish cyclodextrin binding to Mtb GlgE due to a potential steric clash between the proline side chains and the cyclodextrin. Although the Mtb GlgE is not inhibited by cyclodextrins and is

unlikely capable of high affinity cyclodextrin binding at the same site observed in the Sco

GlgEI ortholog, the cyclodextrin binding site observed in the Sco GlgE does indeed appear to represent a small portion of the conserved path leading from the maltohexaose binding site observed in the Mtb GlgE-M6 structure to the M1P binding site.

Given the modest resolution from the data collected, the orientation of the maltohexaose cannot be ascertained for certain. However, from this structure, we have proposed a model whereby the linear -1,4 glucan binds across the dimer of GlgE via interactions with the A and C domains of one molecule of the Mtb GlgE dimer, and the B and S domains of the other molecule in the dimer (Figure 16). The -1,6 branched portions of the glucan are then able to enter the active site of the enzyme to function as the maltosyl acceptor substrate. Previous studies show length-dependent product formation in the Sco

GlgEI where product lengths between 8 and 12-mer oligosaccharides are preferred 57.

Combined with our model, the preferred product length data suggests either a conformational hindrance imparted by longer oligosaccharides in a processive enzyme or the inability of longer oligosaccharides to access the active site due to steric hindrance in a non-processive enzyme. The length-dependent addition of maltose and the model proposed here for binding larger oligosaccharides suggest that maltosyl addition is processive and conformational rigidity in the newly synthesized branch may promote rotation and subsequent translation of the linear polysaccharide through the high-affinity docking area where the maltohexaose is binding in the Mtb GlgE-M6 complex.

Figure 16: GlgE bound to model linear glucan. A speculative model of the binding mode of a large linear glucan to GlgE. The M6 molecule (yellow) has several other glucans modeled at its end. This binding mode would suggest that the dimer of GlgE is required for binding of the natural, glycogen substrate.

2.3.2 Identification of important catalytic residues in GlgE

2.3.2.1 Rationale for choosing Asp383 and Asp418

Prior to obtaining structural knowledge of a substrate or analogue bound enzyme complex, the identity of the catalytic nucleophile of GlgE was still unknown. It was hypothesized that the most likely residue was D418. This was based on sequence and similarity to other members of the GH13 family of α-amylases, as well as the Mtb GlgE and

Sco GlgEI maltose bound structures orienting the 1’-OH near D418 57, 70. GlgE also contains the RXXD motif where the catalytic nucleophile is preceded by two residues and then an

arginine, which is highly conserved in this family 71. In addition to this, it was also speculated that E448 was likely the general acid of the reaction based sequence similarity and its location within the active site to other GH13 glycoside hydrolases.

Because GlgE is a transferase rather than a hydrolase, and utilizes a phospho-sugar substrate, we speculated that the binding and catalytic residues may be different. Analyzing the electrostatics of the GlgE active site showed that the -1 subsite is relatively negatively charged, while the -2 subsite is mostly positive (Figure 17A). Because GlgE utilizes a phosphorylated sugar, it seemed more favorable for the phosphate moiety of the M1P to bind in the -2 subsite, the opposite pocket than what is expected. The -2 subsite also contains both an aspartic acid (D383) and glutamic acid (E384). Therefore, we speculated that it might be possible for the -2 subsite to contain the catalytic portion of the enzyme rather than the -1 subsite.

2.3.2.2 Docking and enzymatic studies on identified residues

To determine this, we initially used the Glide docking software to determine how the energetics of the active site would accommodate the M1P substrate. Using Glide, a grid was built around the M1P binding site, and the software used to fit the M1P. Surprisingly, the

M1P did not bind in the orientation that was predicted. Glide predicted that the M1P would likely bind in the reverse orientation, with the phosphate moiety occupying the positive site of the -2 subsite (Figure 17B). The docking study added levity to our prediction, however further experiments were needed to confirm this.

Figure 17: Structural docking studies of M1P and GlgE. (A) Crystal structure of the Mtb GlgE bound to maltose with the potential nucleophiles indicated. (B) Docking of M1P using Glide shows the phosphate moiety of the M1P binding in the alternative orientation near D383.

To accomplish this, variant forms of GlgE were made in order to determine their effect on catalysis. D418 (GlgE-D418N) and D383 (GlgE-D383N) were mutated to asparagine and the effects of these mutations on activity were tested. It was anticipated that one of the mutations would lead to a completely inactive mutant, while the other would retain some or most of its activity. We utilized the Quinaldine Red assay that was previously used to determine the kinetics of the wild-type enzyme. Surprisingly both mutants had substantial losses in activity. GlgE-D383N had roughly 10% of the activity of the WT, while the D418N variant had approximately 20% of the activity (Figure 18). Surprising as this result was, it still did not definitively identify which residue was the catalytic nucleophile.

Figure 18: Relative activities of GlgE and mutants. The relative activities of the two GlgE mutants were compared to that of the wild-type enzyme.

However, structures from both Syson et al. and us showed unequivocally, that D418 was indeed the nucleophile 72. It is likely that D383 plays a very important role in binding substrate, and that mutations to this residue, affect the binding of substrate, rather than actual catalysis. The GlgE-D418N mutant was able to stabilize the transition state of the substrate and promote phosphate release; however, subsequent transfer of the donor to acceptor would not occur. Because the QRed assay indirectly measures maltosyltransfer by quantifying phosphate release, the D418N mutant likely had no ability to transfer the maltose even though it appeared to have higher activity than the D383N mutant.

2.5 Conclusions and future works

From this study we were able to obtain valuable information concerning the activity and structure of the Mtb GlgE. Firstly, we were able to obtain the structure of the wild-type

GlgE in complex with maltose. Despite the sequence and kinetic similarities between the Sco

GlgEI and the Mtb GlgE, elucidating the structure of the Mtb GlgE afforded the full assessment of the similarities and differences between these orthologs. As expected, the Mtb

GlgE and Sco GlgEI structures are very similar overall with the Mtb GlgE possessing the same 5-domain architecture observed in the Sco GlgEI. One important difference between

Sco GlgEI and Mtb GlgE that potentially affects inhibitor development is the single amino acid change in the M1P binding site. While Sco GlgEI possesses a valine at position 279, the

Mtb GlgE has a serine at the corresponding position (Ser303). Because the hydrophobicity of the active site is important for inhibitor design, the mutation to produce the V279S variant of the Sco GlgEI is important for future drug discovery efforts 73, 74. However, the identification of a maltooligosaccharide binding site in Mtb GlgE distinct from the previously identified cyclodextrin binding site for Sco GlgEI was unexpected and highlights some of the inherent challenges in identifying or targeting the binding sites for large biopolymers such as glucans, nucleic acids, and proteins 75.

Using the developed Quinaldine Red assay, we were able to elucidate the Michalis-

Menten kinetics of the Mtb GlgE using glycogen as a substrate. The determined kinetics as well as previously determined kinetic parameters were in agreement. However, the second order rate constant (kcat/Km) that we determined was substantially higher than previous reports. This is likely due to utilizing glycogen rather than maltohexaose, as the glycogen is a better analogue of the natural substrate, in addition to have exponentially more sites for elongation. We were also able to utilize the assay to understand mutations within the active site. Structural information as well as our enzymatic data shows that D418 is the nucleophile of the GlgE catalyzed reaction. While the D383N mutant also exhibited a substantial loss in activity, this was likely a consequence of its involvement in substrate binding and not catalysis.

However, despite our best attempts, we were unsuccessful in improving the crystallization and ultimate diffraction quality of the Mtb GlgE crystals. It is likely that the dynamic nature of the Mtb GlgE domains, in addition to the extra loop regions not found in the Sco GlgE are playing a role in preventing favorable packing in the crystal and lowering the overall diffraction quality. Because the Mtb GlgE and Sco GlgEI are highly conserved in sequence, structure, and enzymatically, future works will almost exclusively rely on the use of the Sco GlgEI as a surrogate for GlgE structural studies. Discovering that the only change in the active site was a point mutation of a serine in the Mtb GlgE, while the Sco GlgE contained a valine was important. This information allowed us to design a variant form of the

Sco GlgE that allowed for high resolution structural data while keeping a completely intact

Mtb GlgE active site (Chapter 3). Despite this, the Mtb GlgE can still be used for enzymatic assays for the study of inhibitors targeted towards it.

Chapter 3

Structural Characterization of the Streptomyces coelicolor

GlgEI-V279S in complexes with non-covalent inhibitors

3.1 Rationale for using the S. co GlgEI as a surrogate for the M. tb GlgE

Despite our best attempts, we were unable to improve the diffraction quality of the

Mtb GlgE crystals. In lieu of using the Mtb GlgE for future structural studies, it was decided that utilizing the Sco GlgEI would be more beneficial. The conserved domain structure of the

Mtb GlgE and Sco GlgEI enzymes and the similar kinetics of the maltosyltransfer reaction catalyzed by both enzymes, suggest that Sco GlgEI is a reasonable surrogate for developing and testing potential Mtb GlgE inhibitors (Fig 19). Additionally, crystals of the Sco GlgEI are known to diffract to much higher resolution than the crystals obtained with the Mtb GlgE.

This is essential to elucidate the detailed interactions between proteins and bound inhibitors.

Therefore, an Sco GlgEI variant was constructed that would afford high-resolution X- ray diffraction data and possess an M1P binding site identical to that of the Mtb GlgE. This is particularly important for future GlgE inhibitor development, since the hydrophobicity of a region targeted for inhibition is an important characteristic that defines the druggability of an

enzyme 74. Specifically, the presence of V279 in the Sco GlgEI M1P binding site significantly increases the hydrophobicity of this site when compared to the Mtb GlgE enzyme, which possesses a serine residue at the analogous site (Figure 19). Therefore, a mutation was made to the S. co glgEI gene to encode a V279S variant of Sco GlgEI. This variant perfectly matches the M1P binding site observed in the Mtb GlgE structure and has already been shown to retain its maltosyltransferase activity and is inhibited by compounds shown to affect Mtb GlgE enzymatic activity 66.

Figure 19: GlgE dimers with active sites. Left: Mtb GlgE dimer with inset active site structure. Right Sco GlgE dimer with inset active bound to maltose. Both dimers are highly conserved with the only active site residue varied between the two being S303 in the Mtb GlgE is a serine in the Sco GlgE.

3.2 Material and Methods

3.2.1 Molecular cloning of the Sco glgEI

A gene encoding the Streptomyces coelicolor GlgEI-V279S variant protein was purchased from GeneArt® (Life Technologies). The gene was subsequently PCR amplified to produce sufficient concentration of DNA for subsequent cloning using the primers below. 81

The product was placed between the NdeI and XhoI cut sites of a modified pET-32 plasmid.

The final pDR32-glgEI-V279S plasmid encodes the S. co glgEI with a C-terminal histidine tag. The sequence of this construct was verified using DNA sequencing.

3.2.2 Expression and purification

The pDR32-glgEI-V279S plasmid was used to transform BL21* cells. The resulting protein expression and purification procedures are identical to those described for the M. tb

GlgE except for one change; the buffer used during size exclusion chromatography contained

20 mM Tris pH 7.5, 150 mM NaCl, and 0.3 mM TCEP.

3.2.3 Crystallization of Sco GlgEI-V279S-MCP, GlgEI-V279S- αMTF, and GlgEI-V279S-

DDGIM, and GlgEI-V279S-DDGMP complexes

Sco GlgEI-V279S at 8.0 mg/mL using a 0.1% Abs value of 1.526 was used for crystallization experiments. MCP, αMTF, and the DDGIM were added to the concentrated

GlgE to final concentrations of 18 mM, 17 mM, and 36 mM respectively. Crystals of Sco

GlgEI-V279S were grown by the hanging-drop vapor diffusion method. Crystallization drops containing 2 μL of protein solution and 2 μL of well solution were equilibrated with 100 μL of well solution containing 0.2 M sodium citrate pH 7.0 and 10 % PEG 3350. Ethylene glycol was added to a final concentration of 25 % to the drop prior to flash-cooling. DDGMP containing crystals were produced by soaking pre-formed apo-crystals of the Sco-GlgEI-

V279S prepared using the above method. These crystals were soaked with 5 mM of the compound for 48 hours prior to the addition of the ethylene glycol and flash-cooling.

3.2.4 Diffraction experiments

X-ray diffraction experiments were carried out at the LS-CAT beamline at the 82

Advanced Photon Source of Argonne National Labs, Argonne, IL. Structures were solved using diffraction data collected at a wavelength of 1.07819 Å. Diffraction data were integrated and scaled using HKL3000 50. Following scaling, the diffraction data were subjected to analysis using the Xtriage tool in PHENIX 51. This analysis indicated no evidence of twinning.

3.2.5 Structure Determination of the Sco GlgE-V279S complexes.

The structure of the Sco GlgEI-V279S-MCP was phased using molecular replacement in PHENIX 52, 53. The Sco GlgEI-V279S-MCP structure used the monomer of the Sco GlgEI

(PDB accession code: 3ZT5) as a search molecule. The Sco GlgEI-V279S-DDGIM, Sco

GlgEI-V279S-α-MTF, and Sco GlgEI-V279S-DDGMP structures used the Sco GlgEI-

V279S-MCP structure as a model for rigid body refinement. Rigid body refinement, simulated annealing, positional and B-factor refinements were carried out using PHENIX 53,

55. Manual refinement of all structures was performed using COOT 54. Ligand and ligand restraints were generated using eLBOW software in PHENIX 56. Structural validation was performed using Molprobity 55.

3.2.4 Mass Spec labeling studies for covalent modification

A stock solution of 2-deoxy-2,2-difluoro-α-maltosyl fluoride at a concentration of

130 µM was added to a solution of 13 µM Sco GlgEI-V279S (1 mg/mL) in ammonium acetate pH 7.0. The reaction mixture was incubated at room temperature for 2 h. The excess of compound was removed using 3 buffer exchange steps into a 10 mM ammonium acetate pH 7.0 buffer using ultrafiltration. The protein samples were digested with sequencing-grade trypsin using standard protocols (Promega). Determination of peptide mass was carried out

by MALDI-MS using a Bruker UltafleXtreme (Bruker Daltonics). The mass of only the native protein fragment was observed (m/z = 1695.9020).

3.2.5 Sco-GlgE-V279S maltooligosaccharide extension using M1F and α-MTF as substrates

Enzymatic reactions were carried out in 50 μL reactions containing 50 nM of GlgE and 1 mM maltohexaose in a reaction solution containing 20 mM sodium acetate at pH 7.5.

Four reactions were performed: a control containing only maltohexaose (M6), a control reaction of only GlgE and M6, a reaction containing GlgE, M6, and 1 mM of M1F, and a final reaction containing GlgE, M6, and 1 mM of the α-MTF.

3.2.6 Determination of maltooligosaccharide elongation by Sco GlgEI-V279S size using

MALDI-MS

One μL aliquots of each reaction were removed after 10 minutes and the reactions quenched using 19 μL of a saturated DHB (2,5-dihydroxybenzoic acid) matrix solution dissolved in acetonitrile/water (1:1, v/v) with 0.1% trifluoroacetic acid (TFA). Determination of maltooligosaccharide elongation was carried out in positive-ion reflectron mode using an

UltrafleXtreme MALDI-TOF/TOF mass spectrometer (Bruker).

3.2.7 Development and optimization of a real-time coupled assay for GlgE activity and high- throughput screening

Stock solutions of 50 mM 4-Nitrophenyl-α-D-maltopentaoside (PNP-M5) (Sigma-

Aldrich) or 50 mM 4-Nitrophenyl-α-D-glucopyranoside (PNP-Glu), and 1 U of α- glucosidase from Saccharomyces sp. (MP Biomedicals) were prepared using the assay buffer.

The PNP-M5 and PNP-Glu were stored at -20 °C and were stable for several months. The α- glucosidase was stored at 4 °C and remained as active after several months. 84

Enzymatic reactions were carried out in 50 µL volumes at 25 °C on a synergy H4

Hybrid Reader (BioTek) using 50 nM of enzyme. Buffer conditions for the assay were as follows: 100 mM sodium phosphate pH 7.5, and 50 mM NaCl. PNP-M5 and the α- glucosidase were added to final concentrations of 1 mM and 20 mU respectively. All reactions were performed in triplicate. Reactions were performed in transparent 384-well plates and activity monitored using absorbance at a wavelength of 410 nm over the course of an hour. Reaction rates were calculated using the Gen5 software (BioTek). Control reactions for the assay were done using the PNP-Glu and α-glucosidase. All reaction conditions and reads remained the same. The PNP-Glu was used at a final concentration of 500 µM and the concentration of the α-glucosidase was reduced to 20 µU as the previously used concentration resulted in all reactions being complete almost immediately.

The Z’ value for the reaction was determined using 24 positive control reactions and

24 negative control reactions. Positive control reactions contained GlgE, while the negative controls did not, and were substituted with the assay buffer. Reaction conditions were otherwise similar to those already described above.

3.3 Results and Discussion

3.3.1 Crystallization strategies for the Sco GlgEI-V279S

Reproducing the wild-type Sco GlgEI crystals reported by Syson et al. was the first avenue that was taken for these crystallization studies. Their condition contained 0.2 M sodium citrate pH 7.0, 10 % PEG 3350, and 15% ethylene glycol 57. Crystals were produced using this condition, however they were highly etched on the surface, and contained internal

flaws, which would have not been amenable to producing high-resolution diffraction (Figure

20).

Figure 20: Initial Sco GlgEI-V279S crystals. Using the condition reported by Syson et al. yielded crystals with defects that would not have been suitable for producing high- resolution diffraction.

A two dimensional screen was used to adjust the concentrations of PEG3350 and ethylene glycol in order to better optimize crystal growth. This led to the final crystallization condition that was eventually utilized: 0.2 M sodium citrate pH 7.0 and 10% PEG 3350. This condition provided easily reproducible Sco GlgEI-V279S crystals that could be co- crystallized with compounds, or pre-formed crystals, which could be soaked with compounds

(Figure 21). Ethylene glycol was added to the final drops to a final concentration of 25% as a cryoprotectant. An Index Screen was also performed to identify potential new conditions for crystallization, however the citrate condition above provided suitable diffraction quality crystals and as such, no further optimization of Index Screen conditions was continued.

Figure 21: Optimized Sco GlgEI-V279S crystals. These crystals showed no evidence of surface etching or internal flaws and produced very high quality data.

3.3.2 Sco GlgE-V279S in a binary complex with Maltose-C-phosphonate (Text and figures reproduced from Lindenberger et. al., Sci. Rep., 2015 76) Since GlgE uses M1P as a maltosyl donor to extend linear glucans, it was hypothesized that a non-hydrolysable substrate analog could potentially function as a competitive inhibitor of GlgE. This compound, α-maltose-C-phosphonate (MCP), was synthesized and tested for GlgE inhibitory activity 65. MCP inhibited the Mtb GlgE with an

IC50 of 250±24 μM, which is roughly equivalent to the Km of the natural M1P substrate

(Figure 22). This compound was co-crystallized with the Sco GlgEI-V279S to characterize the molecular basis of this inhibition.

Figure 22: Malose-1-phosphate and the first GlgE inhibitor, α-maltose-C-phosphonate. The IC50 value for MCP is shown below the compound.

Crystals of the Sco GlgEI-V279S-MCP were analyzed and diffraction data extended to 1.9 Å. The protein, which crystallized in the P41212 space group, had the following unit cell parameters: a = 113.6 Å, b = 113.6 Å, c = 315.1 Å; α = β = γ = 90 ° (Table 6) and was solved to 1.9 Å resolution. Initial molecular replacement for the structure used the monomer of the Sco GlgEI (PDB accession code: 3ZT5) as a search molecule. The final structure contains a dimer in the unit cell. Most of the protein is resolved in the structure except for portions of the termini. The N-terminus of the protein is mostly unresolved and density for any residues does not appear until P15. The C-terminus of the protein is mostly resolved with

R663 being the last resolved residue out of 675. However, most importantly, the structure exhibits difference density within the M1P binding site that clearly illustrates the binding mode of MCP (Figure 23).

Table 6: Data collection and refinement statistics for Sco-GlgE-V279S structures. Values for the highest shells are in parentheses.

Data Collection MCP DDGIM α-MTF DDGMP PDB Code 4U31 4U2Y 4U2Z n/a

Space Group P41212 P41212 P41212 P41 Unit Cell a, b, c (Å) 113.6,113.6,315.1 113.8,113.8,314.1 113.3,113.3,314.3 113.9,113.9,312.5 Dimensions α, β, γ (°) 90, 90, 90 90, 90, 90 90, 90, 90 90, 90, 90 Resolution (Å) 50.0 – 1.9 50.0 – 2.5 50.0-2.30 50.0-3.20

Rmerge 9.0 (61.4) 8.8 (46.6) 10.0 (51.0) 12.4 (61.7) I/σI 37.6 (4.6) 31.3 (3.8) 26.0 (4.1) 11.5 (2.4) Completeness (%) 100 (100) 94.0 (94.5) 99.9 (100) 99.6 (99.8) Redundancy 13.4 (13.2) 10.3 (10.4) 8.3 (8.3) 4.3 (4.3) Refinement Resolution (Å) 42.7 – 1.9 47.6 – 2.5 42.6-2.30 39.06-3.20 Unique Reflections 175,910 69,481 96,584 64,597

Rwork/Rfree 0.1643/0.1943 0.1769/0.2165 0.1629/0.1993 0.2232/0.2529 No. Atoms Protein 10359 10276 10276 20552 Water 1447 272 891 8 Ligand 100 44 48 108 B-factors (Å2) Protein 24.5 51.8 35.9 57.3 Water 31.8 47.3 38.3 50.1 Ligand 21.8 45.9 37.0 57.9 R.m.s deviations Bond lengths (Å) 0.009 0.010 0.008 0.008 Bond angles (°) 1.08 1.37 1.08 1.21 Ramachandran Favored (%) 98.0 98.1 97.7 97.0 Outliers 0.1 0.1 0.1 0.5

Figure 23: Sco GlgEI-V279S in complex with MCP. Fo-Fc omit map calculated while omitting MCP is contoured at 3 with MCP bound in the enzyme active site.

Throughout the remainder of this chapter, residue numbering is formatted where the number preceding the backward slash represents that of Sco GlgEI-V279S and the number following the backward slash represents the residue number of the Mtb GlgE.

Coordination of the phosphonate moiety is achieved by four residues in the phosphate binding site of the maltosyl donor pocket: N395/419, K355/379, N352/376, and Y357/381

(Figure 24). The sidechain of K355/379 forms an ionic interaction with the negatively charged OP2 of the phosphonate moiety, while the Nδ atom of the N352/376 side chain and the hydroxyl of Y357/381 coordinate the carbonyl of the phosphonate (OP1) via hydrogen 90

bonding interactions. Lastly, the carbonyl oxygen of the N395/419 side chain forms a hydrogen bond with the third oxygen atom of the phosphonate moiety (OP3), suggesting that this oxygen is protonated (Figure 24).

Figure 24: Interactions between the MCP and Sco GlgEI-V279S. Numerous hydrogen bonded interactions coordinate the glucosyl moiety in the -2 site, while the four residues in the phosphate-binding site coordinate the phosphonate moiety of the MCP. The D394/418 nucleophile is present in this structure and is positioned 3.3 Å from the C1’ of the MCP. E423/447 is well positioned to act as a general acid to protonate O1 of the phosphate of M1P at a distance of 3.8 Å Inset below is a cartoon drawing of the MCP.

Since it was not necessary to make an active site mutant to form the stable GlgEI-

V279S-MCP complex, this structure exhibits the direct interactions between the nucleophilic

D394/418 residue and the M1P substrate immediately prior to nucleophilic attack. In this structure, D394/418 is positioned 3.3 Å from the C1’ atom of MCP. Other important interactions of the -1 subsite seen in the GlgEI-V279S-MCP complex have been described in previously published GlgEI structures 57, 72. Briefly, the Nδ atom of N395/419 is hydrogen bonded to the endocyclic oxygen of the -1 sugar. An additional hydrogen bond is formed between Q324/348 and the O6’ hydroxyl group, as well as between D480/504 and O3’ of the

MCP. The oxygen atom of E423/447 was proposed to function as a general acid to protonate the phosphate leaving group during the first step of the reaction and is positioned approximately 3.8 Å from the carbon of the phosphonate moiety that has replaced the bridging oxygen of the M1P substrate. Other interactions seen here are R392/416 and the O2’ hydroxyl as well and hydrogen bonds between the O6’ and D394/418 and N342/366.

Additionally, the glucosyl moiety in the -2 subsite forms an extensive hydrogen bonded network with the side chains of residues K264/288, N268/292, Y535/559, D359/383,

S279/303, and K534/558 as well as the backbone amide of A282/306 (Figure 3-6) as seen and described in previous Sco GlgEI structures 57, 72.

The MCP bound complex structure gives the first complete picture of the fully intact and functional GlgE active site in the presence of a substrate analog. The MCP bound structure described here shows many of the same interactions as those described in the Sco

GlgEI-D394A-M1P (RCSB accession number: 4CN1), but one interesting change was observed. The ionic interaction between the side chain ammonium of K355 and the OP1 of the MCP structure has a bond distance of 2.9 Å, while the equivalent bond in the M1P bound structure is much weaker at a distance of 4.3 Å. This could simply be a function of the slightly improved density for the phosphonate moiety as compared to the phosphate of the 93

Sco GlgEI-D394A-M1P structure. Alternatively, the increase in distance may be the result of slight structural alterations of the active site in the inactive Sco GlgEI-D394A as a result of the mutation of the nucleophile, D394, to an alanine residue. Regardless, the intact active site observed in the MCP structure described here may be slightly more representative of interactions that occur between the enzyme and substrate.

Comparing the structures of the Mtb GlgE-MAL complex and the Sco GlgEI-V279S-

MCP complex show little change within the enzyme active sites. Despite the absence of the phosphate in the phosphate binding site of the GlgE-MAL, the residues responsible for this coordination exhibit only minor positional shifts when compared to the maltose-bound structure. This suggests a structural rigidity in the active site that will improve the probably of designing improved inhibitors.

It has previously been shown that the GlgE enzymes are unable to use α-glucose-1- phosphate as a donor molecule 57. Because the sole difference between the phospho-sugars is the presence of the additional glucose of the M1P, it would appear that the -2 subsite is extremely important for substrate binding and recognition, and subsequent catalysis. Indeed, the -2 subsite contributes many more hydrogen bonding interactions than the -1 subsite.

Additionally, the mutation of the Sco GlgEI to the V279S variant adds an additional hydrogen bond between the serine side chain and the endocyclic oxygen of the second glucosyl moiety in the -2 subsite. Therefore, the design of the next generation of inhibitors should continue to exploit the hydrogen bonded interactions observed in the -2 subsite.

3.3.3 Sco GlgE-V279S in a binary complex with an oxocarbenium mimic (Text and figures reproduced from Lindenberger et. al., Sci. Rep., 2015 76) GlgE belongs to the GH13_3 subfamily of glycoside hydrolase enzymes that proceeds through an oxocarbenium 47, 57. As such, both intermediate and transition state analogues have been used as potent inhibitors of these hydrolytic enzymes 77, 78, 79. Therefore, we synthesized and tested an azasugar, 2,5-dideoxy-3-O--D-glucopyranosyl-2,5-imino-D- mannitol (DDGIM), that would mimic the oxocarbenium formed through a step-wise dissociative mechanism (Figure 25) 80. The DDGIM mimics both the positive charge at the corresponding position, as well as the half-chair conformation formed during catalysis 66, 81, 82.

The DDGIM showed inhibition of both the Mtb GlgE and the Sco GlgE-V279S with IC50 values of 237  27 μM and 102  8 μM, respectively (Figure 25) 65, 66.

Figure 25: The second inhibitor for GlgE DDGIM and the oxocarbenium transition state. Listed are the IC50 values for both the Mtb GlgE and the Sco GlgEI-V279S.

To begin characterizing the inhibitory mechanism, DDGIM was co-crystallized with

Sco GlgEI-V279S. The crystals of the Sco GlgEI-V279S-DDGIM were analyzed and 95

diffraction data extended to 2.5 Å. The protein, which crystallized in the P41212 space group, had the following unit cell parameters: a = 113.8 Å, b = 113.8 Å, c = 315.1 Å; α = β = γ =

90 ° (Table 6) and was solved to 2.5 Å resolution. The crystal was isomorphous with the

MCP bound structure; therefore the dimer was used as a model for rigid body refinement for difference Fourier analysis to generate the initial phases and difference maps. Like the MCP structure, the Sco GlgEI-V279S-DDGIM contains a dimer in the unit cell and is resolved from residues 15-663. Examining the active site reveals clear density for the DDGIM (Figure

26).

Figure 26: Sco GlgEI-V279S in complex with DDGIM. Fo-Fc omit map calculated while omitting DDGIM is contoured at 3 with DDGIM bound in the enzyme active site.

Inspection of the initial Fo-Fc map following rigid body refinement clearly showed the presence of the compound in the enzyme active site. Refinement of the GlgEI-V279S-

DDGIM co-crystal structure shows the five membered imino mannitol moiety occupying the

-1 subsite of the active site with numerous interactions orienting the imino mannitol (Figure

26). The O1’ hydroxyl group is forming hydrogen bonded interactions with the side chains of both R392/416 and E423/447, the general acid in the first step of the catalytic cycle. The nucleophile D394/418, forms an ionic interaction with the secondary ammonium of the imino mannitol moiety at a distance of 2.6 Å, which is significantly shorter than the observed bond distance between the D394/418 and the C1’ of the MCP bound structure (3.3 Å). The glucose moiety of DDGIM bound in the -2 subsite forms an extensive hydrogen bonded interaction network the same as that seen in previous GlgE structures (Figure 27).

Figure 27: Sco GlgE-V279S in complex with DDGIM. The glucose moiety of the DDGIM in the -2 site is coordinated in the same manner as was observed in the previous structures. Interactions of the imino mannitol in the -1 site are numerous. The O01 hydroxyl is hydrogen bonded with Q324/448, the general acid E423/447 is forming a hydrogen bond via the O01 hydroxyl of the imino mannitol, as well as an additional hydrogen bond between D480/504 and O09 hydroxyl. D398/418 is forming an ionic interaction with the secondary ammonium of the DDGIM. Inset between each image is the line structure of each inhibitor.

It is intriguing that, although the five-member imino mannitol ring contains one less hydroxyl group and lacks a phospho-mimicking moiety, DDGIM exhibits an IC50 value (Mtb

GlgE = 237  27 μM and Sco GlgE =102  8 μM) similar to that observed for the MCP substrate analog (Mtb GlgE 250  24 μM). Chemical properties of DDGIM most likely contributing to the comparable inhibition are the shape and the ionic interaction between the imino mannitol and the D394/418 side chain. The DDGIM mimics the semi-planar, half- chair conformation adopted by the oxocarbenium in the -1 subsite during catalysis. Because of this structural mimicry, the secondary ammonium of the imino mannitol is positioned much closer to the D394/418 nucleophile than the analogous C1’ atom in the MCP bound structure. This newly formed ionic interaction observed between the positively charged secondary ammonium and the nucleophile represents a new interaction not previously observed in any of the binary GlgE enzyme ligand-complexes. Similar to what was observed in previous structures and the MCP complex structure; the numerous interactions between the residues composing the -2 subsite and the glucosyl moiety remain fixed despite the different interactions observed between DDGIM and the -1 subsite. Similar to the conclusions from the MCP bound structure, it appears that maintaining the -2 subsite interactions are vital for binding of substrate and inhibitors, while the enzyme is able to accommodate structural variation of inhibitors within the -1 subsite.

3.3.4 Sco GlgE-V279S in a binary complex with an imino-mannitol-N-methyl-C- phosphonate, DDGMP

Despite the functional differences between both the MCP and DDGIM inhibitors, both exhibited roughly similar IC50 values towards GlgE. The MCP mimics the structure of

the natural substrate, M1P, and is able to occupy all sites within the active site simultaneously. The DDGIM, despite missing several functional groups, exploits most of its binding affinity through oxocarbenium mimicry, and through an ionic interaction not present in the MCP. Using the structural data obtained from the previous co-crystals, it was hypothesized that an inhibitor that mimicked the oxocarbenium intermediate, while occupying the phosphate-binding site in the enzyme, could significantly improve upon the initial GlgE inhibitors. Therefore, the DDGIM was modified to produce a 2,5-dideoxy-3-O-

α-D-glucopyranosyl-2,5-imino-D-mannitol-N-methyl-C-phosphonate (DDGMP) (Figure 28).

To characterize the potential interactions of the newly synthesized DDGMP within the enzyme active site, the Sco GlgEI-V279S was crystalized in the presence of the DDGMP.

Figure 28: The third GlgE inhibitor DDGMP. The compound was designed to mimic the transition state of M1P prior to release of the phosphate.

Preformed apo-crystals of the Sco GlgEI-V279S were soaked with the DDGMP compound for 48 hours prior to X-ray experiments. The crystals of the Sco GlgEI-V279S-

DDGMP were analyzed and diffraction data extended to 3.2 Å. The protein, which crystallized in the P4 space group, had the following unit cell parameters: a = 113.8 Å, b =

100

113.8 Å, c = 312.5 Å; α = β = γ = 90 ° (Table 6) and was solved to 3.2 Å resolution. The structure was phased using molecular replacement with the Sco GlgEI-V279S-MCP dimer used as a search model. The Sco GlgEI-V279S-DDGMP contains two dimers in the unit cell and all four molecules are resolved from residues 15-663. Examining the active site reveals clear density for the DDGMP (Figure 29).

Figure 29: Sco GlgEI-V279S in complex with DDGMP. Fo-Fc omit map calculated while omitting DDGMP is contoured at 3 with DDGMP bound in the enzyme active site.

Refinement of the GlgEI-V279S-DDGMP co-crystal structure shows the five membered imino mannitol moiety occupying the -1 subsite with many of the same interactions observed in the DDGIM structure. To briefly restate these, the O1’ hydroxyl group is forming hydrogen bonded interactions with the side chains of both R392/416 and E423/447, the

101

general acid in the first step of the catalytic cycle. The nucleophile D394/418, forms an ionic interaction with the secondary ammonium of the imino mannitol. The glucose moiety of

DDGMP bound in the -2 subsite, forms the extensive hydrogen bonded interaction network; the same as that is seen in previous GlgE structures (Figure 30).

102

Figure 30: Interactions between the Sco GlgE-V279S and DDGMP. The DDGMP maintains many of the same interaction in both the -1 and -2 subsites. The distances observed for the phosphate binding site are different from what was observed in the MCP bound structure.

103

While not as well resolved as the MCP bound structure, the phosphonate moiety of the

DDGMP is clearly visible in the structure. The phosphonate moiety exhibits many of the same interactions that were observed in the MCP structure. Again, to briefly restate, the N- phosphonate moiety coordinated by four residues in the active site: N395/419, K355/379,

N352/376, and Y357/381 (Figure 30). The amino moiety K355/379 forms an ionic interaction with the negatively charged OP2 of the phosphonate moiety, while the Nδ atom of the N352/376 side chain and the hydroxyl of Y357/381 coordinate the carbonyl of the phosphonate (OP1). Lastly, the carbonyl oxygen of the N395/419 side chain forms a hydrogen bond with the third oxygen atom of the phosphonate moiety (OP3). Despite the number of conserved interactions, a few noticeable changes are observed between the MCP,

DDGIM, and DDGMP structures.

The DDGMP unlike the DDGIM, is able to occupy the phosphate binding site, in addition to mimicking the oxocarbenium intermediate. Because of this, the molecule moves in the active site, such that the phosphonate moiety can be coordinated by the phosphate subsite tetrad. This alteration causes the molecule to roll slightly in the active site compared to the MCP and DDGIM. The molecule appears to sit at a position that is the average of the other two molecules. As a consequence of this, the bond distances between several of the key residues in the active site and the molecule itself have changed significantly (Figure 30).

Firstly, the distances between N395/419, Y357/381, and the phosphonate remain relatively constant with N395/419 having a bond distance of 2.5 Å, and the Y357/381, having a bonding distance of 2.6 Å. This is consistent with the MCP bound structure, which has 2.6 Å and 2.6 Å respectively. The first significant alteration is seen in the distance of

104

K355/379. This distance shifts significantly from 2.9 Å in the MCP bound structure to 3.9 Å in the DDGMP structure. This is again seen from the N352/376. Here the distance has also changed significantly from 2.8 Å in MCP structure, to 3.6 Å in the DDGMP one (Table 7).

The shifting of the DDGMP also causes some alterations to the interactions seen in the -2 subsites. However, the changes in bond distances observed here are minimal compared to what is observed in the -1 subsite.

Table 7: Bond distances observed between residues in the Sco GlgEI-V279S active site and the atom in the inhibitors.

Bond distance from residue to compound atom (Å) MCP DDGIM DDGMP

D394 - C1'/N1' 3.3 2.6 3.0

E423 - CP1 3.8 n/a 3.3

N352 - OP1 2.8 n/a 3.6

K355 - OP3 2.9 n/a 3.9

Y357 - OP2 2.6 n/a 2.6

N395 - OP2 2.6 n/a 2.5

One reason this shifting may have occurred is due to the presence of the 5-membered imino-mannitol ring, compared to the 6-membered ring of the MCP. Because of where the tertiary amine of the DDGMP sits compared to the C1’ of the MCP, the distance from the phosphate binding pocket is increased by approximately 0.8 Å. As a result of this, the molecule must move for the atoms of the phosphonate to interact with the sidechains of the residues in the phosphate-binding pocket. This results in a slight tilting and rotation of the molecule, which allows it to interact with Y357/381 and N395/419, but moves it away from 105

the other two residues of the opposite side. This would likely abolish any interactions between the phosphonate and K355/379 and N352/376. This also causes the tertiary ammonium to move approximately 0.4 Å from the analogous position in the DDGIM structure. While this would not completely eliminate the ionic interaction observed here, it would certainly weaken it.

Testing still needs to be performed in order to ascertain an IC50 for the DDGMP.

However, based on the interactions observed in the active site, it is likely to be a more potent inhibitor, but should show similar inhibition to both the DDGIM and MCP inhibitors. From the structure, it was hypothesized that adding an additional methylene group between the tertiary amine of the DDGMP and the phosphonate could potentially increase the efficacy of the inhibitor. The additional carbon would add flexibility to the molecule, potentially allowing it to occupy the -1 and -2 subsites in a similar fashion to the DDGIM and MCP, while being able to interact with all four residues of the phosphate binding site. Two compounds have been made for this purpose: a glycosylated-proline based C-phosphonate

(Inhibitor 4), and a DDGMP based molecule with an additional methylene group, 2,5- dideoxy-3-O-α-D-glucopyranosyl-2,5-imino-D-mannitol-N-ethyl-C-phosphonate (DDGEP

(Figure 31). Both compounds have been co-crystalized with the Sco GlgEI-V279S and preliminary crystals have been obtained for the DDGEP based molecule. Investigations into the IC50 values are ongoing.

106

Figure 31: Potential inhibitory compounds based on the DDGMP structure. The additional methylene group between the tertiary amine and the phosphonate may provide additional flexibility allowing the compounds to occupy all portions of the active site.

3.3.5 Confirming D418/394 as the genuine nucleophile of GlgE enzymatic activity (Text and figures reproduced from Thanna et. al., Org Biomol. Chem., 2015 83)

3.3.5.1 Probing covalent modification of GlgE by α-MTF

Previous studies have shown maltose-1-fluoride (M1F) to be an exceptional substrate for GlgE. Enzymatic reactions performed in vitro show higher levels of activity than when using M1P 57, 72. Precedents have shown that the addition of electron withdrawing fluoro groups to the 2’ and 5’ carbon positions on sugars destabilizes the oxocarbenium intermediate formed during glycosylhydrolase reactions, which can slow the second half- reaction and potentially trap the enzyme in the covalent-bound intermediate form 72, 84, 85. The structure of the Sco GlgEI covalently modified using 1,2-difluoromaltose was very recently published showing that this method is sound 72. However, the determination of the 1,2- 107

difluromaltose covalent complex with Sco GlgEI required the use of a GlgEI variant that lacked the general base, residue E423/447, necessary to activate the maltosyl acceptor in the second half-reaction 72. To further confirm D394/418 as the nucleophile of the enzyme, a

1,2,2-trifluoromaltose (α-MTF) compound was synthesized and complexed with Sco GlgEI-

V279S to trap the maltosyl-enzyme covalent complex (Figure 32) 83. It was anticipated that using the α-MTF would further stabilize the intermediate, allowing us to maintain the complete active site and not mutate the general base. This trapped intermediate could then be used to obtain valuable information regarding the enzyme mechanism and identifying residues important for catalysis.

Figure 32: α-maltose-1-fluoride and the α-MTF. M1F is an exceptional substrate for GlgE, while the α-MTF was designed to trap the maltosyl-enzyme intermediate without the need to mutate the general base in the active site.

3.3.5.2 MALDI-MS experiments

Initial attempts to identify a modification of the D394 nucleophile of Sco-GlgEI-

V279S by the α-MTF focused on using peptide fingerprinting. Specifically, the covalent modification would be observed using trypsinolysis of the modified protein and identification

108

of the resulting peptides using MALDI-MS. This is a very common methodology for the identification of modified proteins. Therefore, the Sco-GlgEI-V279S was reacted with the α-

MTF for 2 hours and a control reaction performed in the absence of α-MTF. The enzymes were digested with trypsin and the resulting peptides were analyzed in order to identify the any modification to the peptide containing the nucleophilic D394. However, no peptides containing D394 were ever observed which had a mass shift of +344 m/z. This would have corresponded to a modification by the α-MTF. The lack of the modification could potentially be due to hydrolysis of the intermediate prior to MS experiments or that the α-MTF was unreactive towards the enzyme. Therefore, further experiments were performed to identify if the α-MTF could potentially be acting as a non-covalent rather than a covalent inhibitor.

3.3.5.3 Inhibition studies of the α-MTF

The lack of modification of the enzyme by α-MTF could be the result of two different scenarios; either the α-MTF-enzyme intermediate was hydrolyzed prior to experiments, or the α-MTF was unreactive. If reactivity was the cause, then the α-MTF should behave as a potential non-covalent inhibitor. Therefore, inhibition studies using the α-MTF were performed.

Using M1P and glycogen as substrates, enzymatic release of phosphate was measured using the EnzChek® Phosphate Detection assay kit. However, no inhibition by α-MTF was observed at concentrations below 10 mM. Because of this, it was hypothesized that α-MTF could potentially be acting as a substrate by the enzyme and be incorporated into the glucan during elongation, and as a result, would indeed show no inhibition. Further experimentation was performed to identify whether or not the α-MTF was being used as a substrate by the

109

Sco-GlgE-V279S.

3.3.5.4 Assessment of α-MTF incorporation into linear glucans (Text and figures reproduced from Thanna et. al., Org Biomol. Chem., 201583) To test the incorporation of α-MTF into glucans, two different approaches could be used. Utilization of α-MTF as a substrate would produce fluoride ions upon nucleophilic attack by D394/418. This could potentially allow for the use of a fluoride electrode to quantitatively determine enzymatic activity by monitoring the increase in fluoride concentration 57. The approach we utilized was to qualitatively measure the increase in glucan size caused by the transfer of the α-MTF to a linear glucan. Because of the presence of the difluoro center of the C2’, any mass shift seen α-MTF as a substrate would be different than using other substrates such as M1F and M1P. Glucan elongation would be measured using MALD-MS.

To accomplish this, 1mM of α-MTF would be reacted with 1 mM of maltohexaose

(M6) acting as the acceptor molecule in the presence of the enzyme. Additional controls were performed: an (A) M6 only control, (B) M6 and Sco-GlgEI-V279S, and (C) M6, M1F and

Sco-GlgEI-V279S as an elongation control (Figure 33).

The spectra produced from the M6 and buffer reaction showed a singular peak representing the mass of a sodiated maltohexaose molecule. The addition of enzyme to the maltohexaose produced a series of larger peaks representing addition of one or more maltose units. This is the result of the known background transglycosylation activity of GlgE. The addition of M1F resulted in the further extension of the M6 producing a wide range of maltooligosaccharides up to maltotetradecaose (M14). When using the α-MTF as a substrate,

110

a spectrum similar to the M6 only spectrum was observed (Figure 33D). While there were larger glucans observed, these were not the result of α-MTF incorporation. Incorporation of

α-MTF would results in a mass increase of +344 m/z rather than the +324 m/z that were observed. Most likely, the maltooligosaccharides produced were the result of the transglycosylation activity of GlgE.

111

Figure 33: Evaluation of α-MTF as a substrate for Sco GlgEI-V279S. (A) MALDI-MS of maltohexaose (M6). (B) MALDI-MS of M6 + Sco GlgEI-V279S. (C) MALDI-MS of M6 + Sco GlgEI-V279S + M1F. (D) MALDI-MS of M6 + Sco GlgEI-V279S + MTF. M4, M6, M8, M10, M12, and M14, represent 4, 6, 8, 10, 12, and 14 maltosyl units in the respective products. R = the continuation of maltosyl repeats.

To further confirm this, fluorine NMR experiments were performed to observe the fluoro groups on the α-MTF before, and after incubation with M6 and GlgE. These experiments show that even after exposure to the enzyme, the α-MTF retains all three fluoro groups on the 112

sugar.

3.3.5.5 Sco GlgEI-V279S in a binary complex with α-MTF

Prior to the knowledge that α-MTF could not be utilized by the Sco GlgEI-V279S, crystallization attempts were performed in order to visualize the α-MTF-enzyme intermediate.

Therefore, the α-MTF was co-crystallized with Sco GlgEI-V279S. The crystals of the Sco

GlgEI-V279S-α-MTF were analyzed and diffraction data extended to 2.3 Å. The protein, which crystallized in the P41212 space group, had the following unit cell parameters: a =

113.3 Å, b = 113.3 Å, c = 314.4 Å; α = β = γ = 90 ° (Table 6) and was solved to 2.3 Å resolution. The crystal was isomorphous with the MCP and DDGIM bound structures; therefore the dimer was used as a model for rigid body refinement for difference Fourier analysis to generate the initial phases and difference maps. Like the MCP and DDGIM structures, the Sco GlgEI-V279S-α-MTF contains a dimer in the unit cell and is resolved from residues 15-663. Examining the active site reveals clear density for the α-MTF (Figure

34).

113

Figure 34: Sco GlgEI-V279S in complex with α-MTF. Fo-Fc omit map calculated while omitting α-MTF is contoured at 3 with α-MTF bound in the enzyme active site.

Inspection of the initial Fo-Fc map following rigid body refinement clearly showed the presence of the α-MTF in the enzyme active site. Confirming what was observed from the previous experiments, no continuous density is observed between the nucleophile,

D394/418, and the α-MTF, indicating the absence of a covalent intermediate. Difference density is observed for the 2,2 difluoro group at C2’ with the final model exhibiting a pair of hydrogen bonds between the difluoro moiety and the side chain of R392/416 (Figure 35).

The nucleophile D394/418 appears to be forming a hydrogen bond with the equatorial fluorine at C2’, suggesting that D394/418 may be protonated in this form.

114

Figure 35: Interactions between Sco GlgE-V279S and α-MTF. No covalently modified residue was observed in the crystal structure due to the non-reactivity of the α-MTF. Density is observed for the 2,2 difluoro group at C2’, which hydrogen bond with R392/416. The nucleophile D394/418, appears to be forming a hydrogen bond with the equatorial fluorine at C2’, suggesting that it may be protonated in this form. E423/447 appears to be forming a halogen bond with the C1’ fluorine. Hydrogen bonding distances from between donor and acceptor are indicated next to the dashed line. Inset below is the cartoon structure of α-MTF.

115

Alternatively, this interaction may represent a halogen bond. Residue E423/447, proposed to function as a general acid to protonate the phosphate leaving group, is forming a 2.8 Å hydrogen bond with atom F1 since it is expected to be protonated during the first step of catalysis. As is the case of the interaction between D394/418 and the fluorine at C2’, this

E423/447 interaction with F1 may represent a halogen bond.

3.3.6 Development of a new, quantitative, real-time assay for GlgE activity

3.3.6.1 Attempts to utilize PNP-maltose and PNP-glucose

Several methods exist to measure GlgE activity. Those include using fluorometric assays to detect phosphate release such as the Quinaldine Red assay used to determine the

Mtb GlgE kinetics in Chapter 2, or monitoring increase in glucan length using mass spectrometry as detailed above. Despite this, both have drawbacks that make the less than useful for HTS techniques. The Quinaldine Red assay, while being more sensitive than traditional phosphate-chromophore assays, still suffers from many of the same issues. Firstly, they are discontinuous and require several quenching and addition steps to produce a final product. The sensitive nature of the Quinaldine Red assay also makes the assay prone to elevated background noise even in the presence of small amounts of phosphate contamination. Using mass spectrometry gives non-quantitative results, which for inhibitor screening, cannot be used to determine values such as IC50.

To remedy these issues we attempted to identify new substrates that could be used by the Sco GlgEI-V279S. One avenue that was considered was the use of glycosides that contained a paranitrophenol (PNP) group at the anomeric carbon. Two were identified that may be useful for this purpose: α-PNP-1-maltose (PNP-Mal) and a α-PNP-1-glucose (PNP- 116

Glu).These compounds have been used in the past to probe the activities of glycoside hydrolase enzymes 86. These enzymes are able to hydrolyze the α-1,4 linkage between the

PNP moiety and the attached glycoside. In the process, this produces a glycoside product and a free PNP group that absorbs strongly at 410 nm. Therefore, hydrolase activity and be monitored by the production of PNP in a continuous matter (Figure 36A).

Figure 36: Attempts to utilize new substrates for GlgE activity assessment. (A) The PNP- glycosides, which were tested. (B) Schematic of how the GlgE PNP assay will function.

It was hypothesized that the Sco GlgEI-V279S could utilize these sugars as donor molecules for catalysis. This would result in the transfer of the sugar component to an acceptor and produce PNP which could be monitored (Figure 36B). Both the PNP-Glu and

PNP-Mal were used at a concentration of 1 mM, with 1 mM M6 as an acceptor molecule.

The production of the PNP was monitored over the course of an hour my measuring the absorbance at 410 nm. However, despite several attempts, no production of PNP was observed. This would indicate that GlgE is incapable of utilizing these molecules as 117

substrates, and again confirms that it is highly specific for M1P. The PNP-Mal resembles the

M1P substrate more than the PNP-Glu. However, the presence of the PNP group, which is substantially larger than a phosphate, might not be accommodated in the enzyme active site and thereby blocking its usage. The PNP-Glu, while smaller, may not be positioned correctly in the enzyme active site, making it non-attackable by the nucleophile. Despite this, a new idea was developed to utilize these PNP conjugated sugars as potential substrates.

3.3.6.2 A real-time assay utilizing the reverse GlgE reaction and PNP-maltopentaose

GlgE has three confirmed activities, two of which have been discussed thus far: the maltosyltransferase activity using M1P as a substrate, and the transglycosylation activity in the presence of glucans and the absence of a donor. Additionally, GlgE is able to catalyze its reverse reaction and produce M1P 57, 87. In the presence of a glucan and high concentrations of phosphate (greater than 25 mM), GlgE can catalyze the production of M1P via a phosphorolysis of the glucan using phosphate as an acceptor. Therefore, we attempted to exploit this as a means for monitoring GlgE activity using the reverse rather than forward reaction.

Again, we would utilize a PNP modified glucan as a substrate. GlgE cannot catalyze the reverse reaction from glucans of less than two glucose units in length. Therefore, those of three or greater could be required. However, as previously discussed, the Sco GlgEI-V279S could not utilize the PNP-glucose or PNP-maltose as substrates. Therefore we coupled the

GlgE catalyzed reaction to the activity of an α-glucosidase which would be capable of hydrolyzing the PNP from the sugars. However, the α-glucosidase is inefficient in hydrolyzing the linkage of maltooligosaccharides greater than two units in length. As a result,

118

we would need to use a maltooligosaccharide that results in the ultimate production of PNP- glucose, meaning an initial chain length that has an odd number of glucose residues.

The final assay that was constructed utilized an α-PNP-1-maltopentaose (PNP-M5).

Through the action of GlgE, 2 molecules of M1P would be produced and the final product would be a single molecule of PNP-glucose. This could then be hydrolyzed by the action of the α-glucosidase yielding glucose and free PNP which can be measured (Figure 3-37). For the reaction 1 mM of PNP-M5, and 20 mU of α-glucosidase, and 50 nM of the Sco GlgEI-

V279S were utilized. The resulting assay showed a very high signal to noise ratio and very low deviation (Figure 38).

119

Figure 37: The PNP-M5 GlgE assay. One molecule of PNP-maltopentaose is converted to two molecules of M1P by the reverse activity of GlgE. The activity of the α- glucosidase cleaves the PNP-glucose to produce glucose and PNP, which can be detected by monitoring the absorbance at 410 nm.

Figure 38: Progress curve of the PNP-M5 assay. Reactions were monitored at 410 nm. Control reactions with enzyme showed significant increase in absorbance over time versus those without enzyme.

120

Once we had shown initial assay success, a control reaction was performed in order to ensure that the rate-limiting step of the reaction was being catalyzed by the GlgE and not the

α-glucosidase. A series of reactions were performed whereby the α-glucosidase concentration was varied, while the others were held constant. It was determined that at the 20 mU concentration of the α-glucosidase was sufficiently high as to not be rate limiting (Figure 39).

Figure 39: α-glucosidase concentration dependence determination. At the chosen concentration of 20 mU, the α-glucosidase does not appear to be rate-limiting. As a final step to verify the quality of the assay, a Z’ was determined (Equation 1). 24 control reactions with enzyme and 24 negative control reactions without enzyme were performed. The Z’ was calculated to be 0.94, showing that this assay is highly reproducible and amenable to HTS (Table 8).

121

Table 8: Z’ determination of the PNP-M5 assay.

Average Rate (OD/sec) Standard Deviation (OD/sec)

+ Control (w/ GlgE) 29.3 0.51

- Control (w/o GlgE) 0.03 0.03

Z’ = 0.94

3.4.5 Screening of potential GlgE inhibitors utilizing the PNP-M5 assay

A collaborator, Natanael Seggretti, synthesized a series of GlgE inhibitors, which he kindly provided to us for testing. These inhibitors were based on β-maltosides with modifications at the anomeric carbon. The inhibitors were modeled after glycoside hydrolase inhibitors 88. To test these inhibitors we utilized the PNP-M5 assay.

Because we are utilizing a coupled assay, tests need to be performed in order to ensure that the inhibitors were not inhibiting the α-glucosidase portion of the reaction. To do this, all 25 inhibitors were tested against the α-glucosidase activity. 100 µM of each inhibitor was added to 20 µU of the α-glucosidase. 1 mM of PNP-glucose was used to initiate the reaction. Four inhibitors were identified that inhibited the α-glucosidase, and therefore could not be tested within the scope of this assay (Figure 40A). An orthogonal assay, such as the

Quinaldine Red would need to be utilized to screen these. The remaining 21 were then screened against activity of the Sco GlgEI-V279S. Again the inhibitors were screened at 100

µM against GlgE (Figure 40B). Three of the screened compounds showed weak inhibition

122

towards GlgE (Figure 40C).

Figure 40: Screening of GlgE inhibitory compounds. (A) Screen of activity towards the α-glucosidase. Compounds below the black line were omitted from futures screens. (B) Screen of compounds against GlgE. Those below the red line (C) were tested again, but failed to show dose-dependent inhibition. These inhibitors were then subjected to a dose dependence test in order to fully verify their inhibitory activity. However, none of the reported inhibitors actual responded in a dose dependent manner. This would suggest that the observed inhibition is the result of an artifact of the assay.

3.5 Conclusions and future works

Using the structural data obtained from the Mtb GlgE and the Sco GlgEI structures, we designed a variant enzyme that would crystallize extremely well, while having an identical active site to the Mtb GlgE. This Sco GlgEI-V279S variant proved to be high useful

123

for both crystal studies and assay development.

The crystal structures of the Sco GlgEI-V279S in complex with four different compounds, provided insight about the enzyme and its interactions with these inhibitors. Due to the ease with which the Sco GlgEI-V279S crystallizes, and that the resulting crystals can be either con-crystallized or soaked with compounds, the limit for which structures can be obtained is dependent on the amount of compounds which can be synthesized or found that inhibit GlgE. Therefore, structural based design of inhibitors for GlgE can be easily done and provide great insight for more second and third generation compounds.

An absorbance-based assay using PNP-maltopentaose and α-glucosidase was developed. This assay allowed us to determine the activity of the enzyme in real-time and provided an excellent Z’ score, lending this assay towards high-through-put screening applications. Using the assay we also screened a very small compound library that contained compounds designed to potentially inhibit GlgE. After screening however, none of the compounds showed inhibition. However, given the ease and reproducibility of the assay, screening of other chemical libraries with the goal of discovering potential inhibitors of this enzyme will be performed. Identifying potent inhibitors against GlgE may help identify core structural features for inhibition and allow for the synthesis of new molecule libraries based on the compound.

124

Chapter 4

Biochemical characterization and structural studies of the

Mycobacterium tuberculosis treahalose-6-phosphate phosphatase, TPP2

4.1 Introduction

4.1.1 Trehalose and its role in Mycobacteria

Trehalose (α-D-glucopyranosyl-(1→1)-α-D-glucopyranoside) is a non-reducing disaccharide, which serves a variety of functions in organisms (Figure 41). Plants, fungi, insects, and many bacteria utilize trehalose as a stress protectant 89. Under drought conditions, trehalose production is increased to prevent desiccation, while at very low temperatures, high concentrations of trehalose can serve as an anti-freeze, protecting against the formation of ice crystals 89. It also is utilized quite extensively as a source of energy. Many organisms contain trehalase enzymes that can convert one molecule of trehalose, to two molecules of glucose, providing a fast burst of energy compared to traditional glycolysis 90.

125

Figure 41: Cartoon image of trehalose.

In addition to the above-mentioned roles of trehalose, members of the Mycobacteria family utilize a mycolic acid derivatized trehalose as one of the main building blocks of the mycobacterial outer membrane (mycomembrane). The mycomembrane is very hydrophobic; this is the result of an abundance of mycolic acids found in the mycomembrane in the form of free glycolipids and mycolated cell wall components 91. This hydrophobicity makes it an extremely good barrier to passively block many harmful stresses from the organism 5, 8, 9.

These mycolated cell wall components are synthesized by the Antigen85 Complex of enzymes, which attach the mycolic acids to these cell wall components 92. The primary carrier of this mycolic acid comes in the form of trehalose monomycolate which is synthesized by the action of Pks13, which transfers free mycolic acids to 6’-OH group of trehalose 93. This is then exported to the periplasm by the MmpL3 transporter where it can be utilized 94. Because of its essential role in to the formation of the mycomembrane, understanding the initial trehalose biosynthesis is of paramount importance (Figure 42)

126

Figure 42: Trehalose biosynthetic and utilization pathways.

4.1.2 Trehalose biosynthesis and storage in Mycobacterium tuberculosis

Mycobacterium tuberculosis has a number of pathways to synthesize trehalose

(Figure 42). The TreXYZ pathway produces trehalose from an α-1,4 glucan 40. This pathway liberates one maltose unit from the glucan, which is then isomerized to a molecule of trehalose. A second enzyme that is capable of producing trehalose is TreS, which isomerizes maltose to trehalose. This reaction is reversible and favors flux towards maltose 40, 49. The last pathway is the TPS/TPP2 pathway. The first step in this pathway is the formation of trehalose-6-phosphate (T6P) by trehalose phosphate synthase (TPS) 95. TPS is a glucosyltransferase which utilizes UDP-glucose and glucose-6-phosphate to produce T6P.

T6P is then de-phosphorylated by trehalose-6-phosphate phosphatase, TPP2, to form the final product, trehalose 95.

127

Trehalose recycling is also very important in Mycobacterium tuberculosis. The bacterium contains very specific transporters for the uptake of trehalose, namely the LpqY-

SugABC transporter 48. This trehalose can then simply be reused to make trehalose monomycolate again, or in times of carbon surplus, stored in the form of an α-1,4 glucan which is produced in the GlgE pathway (Chapter 2). This glucan can then be broken down by the aforementioned TreXYZ pathway to liberate free trehalose depending on the bacterium’s needs.

4.1.3 TPP2 and its potential as a drug target in the treatment of TB

Of all the trehalose-producing pathways, the TPS/TPP2 is the sole de novo biosynthetic pathway for trehalose biosynthesis 95. In fact, it is the only essential trehalose synthetic pathway for viability of the organisms, as both TreXYZ and TreS are dispensable 96.

More specifically, TPP2 is required for the viability of the organism, while TPS appears to be dispensable. This essentiality makes TPP2 an attractive drug target, as the purposed mechanism of death is similar to that of GlgE; toxic levels of T6P lead to stresses which are ultimately lethal to the organism 97. Therefore, exploiting a similar mechanism of killing could potentially lead to shorter treatment times, and decrease the selection of drug resistant strains. Currently, there are no known TPP2 inhibitors, and structural information concerning the M. tb TPP2 is lacking. This makes characterizing TPP2 of utmost importance for designing inhibitors for this enzyme.

128

4.2 Materials and Methods

4.2.1 Molecular cloning of the otsB2 gene, D147N mutant, and E. coli treA

The otsB2 (Rv3372) gene was amplified by polymerase chain reaction from M. tb

H37Rv genomic DNA and placed between the NdeI and EcoRI cut sites of a modified pET-

28 plasmid. The resulting pDR28-otsB2 plasmid encodes a recombinant TPP2 possessing an

N-terminal 6x histidine tag. A mutant form of TPP2, otsB2-D147N, was made using site- directed mutagenesis using pDR28-otsB2 as a template.

The E. coli treA gene lacking the N-terminal signal sequence was amplified by polymerase chain reaction using E. coli K12 strain genomic DNA taken from whole cells.

The PCR product was placed between the NcoI and XhoI cut sites of pCDF-Duet plasmid to produce pCDF-Duet-treA. This produced a recombinant TreA protein possessing a C- terminal 6x histidine tags. The sequences of all expression plasmids were confirmed by DNA sequencing.

4.2.2 Expression and purification of TPP2, TreA, and mutants

The pDR28-otsB2 plasmid was used to transform E. coli T7 Rosetta cells. Bacterial cells were cultured at 37 °C in Luria Broth containing kanamycin (50 mg/mL) until an OD600 of 0.6 was reached. Protein expression was induced by the addition of 1 mM IPTG at 16 °C.

Bacterial cells were harvested by centrifugation after incubating for 48 hours at 16 °C.

Pelleted cells were resuspended in a buffer containing 5 mM imidazole, 100 mM NaCl, 20 mM Tris pH 8.0, and 5 mM β-mercaptoethanol.

The resuspended cells were incubated with Lysozyme (10 μM) and DNaseI (100 μM) for thirty minutes on ice prior to lysis by sonication. The crude lysate was then pelleted by 129

centrifugation at 15,000 g for 30 minutes. The supernatant was applied to a 5 mL HiTrap

TALON crude column (GE Healthcare) that had been equilibrated with the resuspension buffer. TPP2 bound to the cobalt column was washed with 15 column volumes of resuspension buffer and was eluted using a linear gradient with a concentration of imidazole ranging from 5 mM to 150 mM over 15 column volumes.

The fractions containing TPP2 were pooled and immediately applied to a HiTrap Q

FF anion exchange column (GE Healthcare). Any bound proteins were eluted by a linear gradient (0.1-1.0 M) of NaCl over 20 column volumes. Fractions containing TPP2 were subsequently pooled and precipitated by the addition of ammonium sulfate to a final concentration of 2.6 M. Precipitated protein was pelleted by centrifugation, and resuspended in a buffer containing 20 mM Tris pH 7.5, 1 mM EDTA, and 1 mM DTT. The dissolved protein was dialyzed against the same buffer for 16 hours to remove any residual ammonium sulfate.

TPP2 that would be used for crystallization experiments was also subjected to size- exclusion chromatography. Concentrated protein was applied to a Hi-Load Superdex 200 size exclusion column (GE Healthcare) equilibrated with a buffer containing 20 mM Tris pH 7.5,

150 mM NaCl, 1 mM MgCl2, and 0.3 mM TCEP. Fractions containing TPP2 were subsequently pooled and concentrated. The purity of the protein at all stages of purification was confirmed using SDS-PAGE. The D147N variant form of TPP2 was purified in the same manner as the wild-type enzyme.

The pCDF-Duet-treA expression plasmid was used to transform BL21* E. coli.

Expression of TreA was carried out according to the procedures described above for the recombinant TPP2. The purification of recombinant TreA is the same as TPP2 with the

130

following exception. The pellet obtained from the ammonium sulfate precipitation was resuspended and dialyzed against a buffer containing 20 mM Tris pH 7.5 and 1 mM EDTA.

The concentration of TPP2 and TreA were determined by absorbance spectroscopy

(280 nm). The theoretical extinction coefficients were obtained using the ProtParam tool on the ExPASy proteomics server. The predicted extinction coefficient for TPP2 was 26,470 M-1 cm-1, and the predicted extinction coefficient for the secreted form of TreA was 122,270 M-1 cm-1.

4.2.3 Preparation of the Amplex Red working solution

The Amplex Red working solution was made per the instructions provided with the

Amplex Red Glucose/Glucose Oxidase Assay Kit (Invitrogen). The Amplex Red working solution contained the following: 50 µM 10-acetyl-3,7-dihydroxyphenoxazine (Amplex Red),

1 U Horseradish Peroxidase, and 10 U glucose oxidase. All components were combined in the assay buffer containing 10 mM HEPES pH 7.5 and 1 mM EDTA. The Amplex Red working solution was stored at -20 °C to minimize the conversion of the Amplex Red to resorufin.

4.2.4 Determination of the enzyme concentrations to be used in the assay and control reactions

Production of resorufin from enzymatic activity was measured using absorbance at

530 nm for all assays. Measurements were taken using a synergy H4 Hybrid Reader

(BioTek). The ability of glucose oxidase to utilize T6P as a substrate was tested using different concentrations of T6P (2.5, 20, and 100 µM). These were combined with the

Amplex Red working solution and reaction buffer. Reactions set up in this manner were then 131

used as the blank for all subsequent assays. The optimal concentration of TreA for the assay was determined using a concentrations range of TreA (20-400 nM). These reactions contained 5 nM TPP2, 600 µM T6P, Amplex Red working solution and the reaction buffer to a final volume of 50 µL. Initial rates were plotted against TreA concentration. To confirm the linear dependence of the reaction as a function of TPP concentration, TPP concentration was varied from 0.5 nM to 30 nM while all other components were held constant (600 µM T6P,

40 nM TreA, and the Amplex Red working solution). Both of the above reactions were performed in duplicate.

4.2.5 Michaelis-Menten kinetics determination of TPP2 using the Amplex Red assay

All assays were performed at 37 °C. The 50 µL reactions contained 5 nM TPP2, 100 nM TreA, and 25 µL Amplex Red working solution that were added to an appropriate volume of reaction buffer. Solutions containing T6P at concentrations ranging from 12.5 µM to 4 mM were added to a series of reactions. Assays were performed in 96-well half-area plates (Costar) by measuring release of resorufin at 530 nm. A series of negative control reactions lacking TPP2 were performed in the same manner. Initial velocity values were calculated by determining the slope between the two and five minute time points. The Prism

5 software was used to perform non-linear regression analysis of data.

4.2.6 Assay validation for high-throughput screening applications

The Z’ value was determined in the same manner as the steady-state kinetics assays, except that T6P concentration was held constant at 400 µM. The Z’ value was determined using 48 positive reactions containing TPP2 and 48 negative control reactions without TPP2.

132

4.2.7 NIH Clinical Collection screening

The NIH Clinical Collection was screened against TPP2 activity. Compounds were added at a final concentration of 100 µM to a reaction containing all other components at normal assay conditions. Sets of 40 compounds were tested simultaneously along with four positive and four negative controls.

4.2.8 Crystallization attempts for TPP2

Despite many attempts, diffraction quality crystals of the TPP2 enzyme that lead to data that could be used to solve the structure were never obtained. Most experiments utilized protein concentrations between 5-20 mg/mL. Crystallization screens used were the Index

Screen (Hampton Research) and the Classics Lite Suite (Qiagen). Initial screen hits were optimized by adjusting component concentrations or by the use of Additive or Detergent

Screens (Hampton Research). A variant form of the protein, TPP2-D147N, was created for the purpose of co-crystallization with substrate. Other substrate analogues, metals, detergents, and transition state mimics were also used in attempts to crystallize the protein. Additionally, crystallization using both hanging and sitting drop under oil, as well as varying temperatures were used to improve crystals. Very specific examples of successful crystal production will be discussed in further detail in the results section. Table 9 identifies a broad range of conditions used during crystallization experiments (Table 9).

133

Table 9: Screens and conditions used during TPP2 crystallization experiments

Screens Type Divalent Substrates Transition state Mutants Metals mimics (M2+) Index Hanging Mg2+ trehalose Aluminum D147N (Hampton) Drop Fluoride AlCl3 + NaF AlFx Classics Lite Sitting Zn2+ T6P Orthovanadate (Qiagen) Drop Under NaVO4 Oil VO3x

½ Index Ni2+ n-dodecyl-β- D-maltoside Additive Ca2+ 4-dodecyl- trehalose Detergent

4.3 Results and Discussion

4.3.1 Developing a new assay to measure TPP2 activity

The kinetic properties of TPP2 had previously been determined utilizing the

Malachite Green assay to monitor phosphate release 98. However, as was described in

Chapter 3, colorimetric phosphate detection assays such as the Malachite Green and

Quinaldine Red have some inherent shortcomings that make them less than ideal for HTS applications: they are discontinuous, lack sensitivity, are laborious with the need for many quenching steps, and have a very small dynamic range over which phosphate can be detected.

For this purpose, we set out to design and optimize a real-time assay for monitoring the activity of TPP2 that would not rely on the detection of phosphate. Our aims would be three- fold: improve upon the sensitivity and dynamic range, make the assay require less steps, and for the assay to measure activity in a continuous manner.

134

The products of the TPP2 catalyzed reaction are inorganic phosphate and trehalose.

As stated, phosphate detection is well documented, but we wanted to move away from this.

In contrast to this, detection of trehalose is much more difficult. Trehalose cannot be measured directly, so the need for a coupled assay was apparent. Trehalose can be broken down into glucose via the action of trehalase 99. The advantage of this being that glucose detection is very common and many methods already exist. An additional benefit is the doubling of signal; for every one trehalose that is produced, two molecules of glucose can be utilized, amplifying the overall signal. For the detection of glucose we chose to use the

Amplex Red/Glucose Oxidase Kit (Invitrogen). The final product of this coupled assay is resorufin, a compound, which absorbs at 530 nm. Therefore, enzymatic activity of TPP2 would be measured indirectly by the production of resorufin (Figure 43).

135

Figure 43: Schematic of the TPP2 Amplex Red assay

4.3.1.2 Optimization of enzyme concentrations for the assay

As a proof of concept to see if the coupled assay would work, a series of control reactions were performed. In these, the concentrations of TreA (100 nM) and TPP2 (5 nM) were held constant, while the concentration of T6P was varied. In addition to this, control reactions were performed in the absence of TPP2 to monitor the background hydrolysis of

T6P. These reactions showed very minimal background hydrolysis of T6P and the activity observed during the reaction was concentration dependent on T6P (Figure 44)

136

Figure 44: Initial progress curves from the Amplex Red assay . The concentration of T6P was varied while holding other assay conditions constant. Activity was concentration dependent on T6P and had no background signal in the absence of TPP2.

Because of the requirement of TreA in the coupled assay, another control experiment was performed in order to ensure that the rate-limiting step of the assay was being catalyzed by TPP2 and not TreA. The concentrations of TPP2 and T6P were held constant at 5 nM and

600 μM respectively, while the concentration of TreA was varied from 20 to 400 nM. The initial velocities as a function of TreA concentration were plotted to determine the concentration dependence (Figure 45).

137

Figure 45: Determination of the TreA concentration dependence. TreA concentration was adjusted to ensure it was not the rate-limiting step in the coupled assay. 100 nM of TreA was used as the standard concentration for all future assays.

From the data, the rate of the reaction plateaued near 40 nM of TreA, at which point, all rates were relatively constant. Therefore, we chose 100 nM TreA for all future experiments in order to ensure a large excess of TreA. As a final control, the rate of the reaction as a function of TPP2 was tested. TreA and T6P were held constant at 100 nM and

600 μM. Initial rates were plotted against TPP2 concentration. Inspection of the graph shows a linear relationship of activity over all concentrations of TPP2 tested, indicating that TPP2 under these assay conditions is the rate-limiting step of the assay (Figure 46).

138

Figure 46: TPP2 concentration dependence on the Amplex Red assay. TPP2 concentration was varied, but activity was linear across all concentrations, indicating TPP2 was catalyzing the rate-limiting step of the coupled assay.

4.3.1.3 Michaelis-Menten kinetics for TPP2

The next step in our characterization of TPP2 was to determine the Michaelis-Menten kinetics using the Amplex Red assay. All experimentally determined enzyme concentrations were used, while the concentration of T6P was varied from 2.5-800 μM. Reactions were performed in triplicate and the initial rates of each reaction were determined from the slope between the two and five minute time points measuring the absorbance at 530 nm. Initial rates were determined with the Beer-Lambert law using an extinction coefficient of 30,201

M-1cm-1 for resorufin. However the effective extinction coefficient was doubled since for every one trehalose produced, two glucose molecules were utilized. Fitting of the initial rates of each reaction against substrate concentration to the Michaelis-Menten model was done using the Prism 5 software (Figure 47). 139

Figure 47: Michaelis-Menten curves for T6P. T6P concentration was varied from 12.5 µM to 4 mM, while other components of the coupled assay were held constant. Error bars (standard deviation) were calculated from triplicate reactions.

Km and kcat were calculated to be 353.7 ± 44.6 µM trehalose and 225.8 ± 12.8 µM

2 T6P/min respectively (R = 0.9809). The second-order rate constant of kcat/Km was

-1 -1 determined to be 0.0106 µM s . The determined Km and kcat/Km for TPP2 is comparable to the previously determined in the Malachite Green assay. The table below compares TPP2 to other trehalose phosphate phosphatases from other organisms (Table 10).

140

Table 10: Comparison of kinetic parameters of different TPP enzymes and the assay used to determine them.

Organism Km (µM) kcat (µM/min) kcat/km (s-1) Assay M. tuberculosis 354 225.8 0.0106 Amplex Red M. tuberculosis 600 n/d n/d Malachite Green E. coli 2500 14.3 0.0058 Malachite Green M. smegmatis 1600 n/d n/d Malachite Green T. acidophilum 2700 10.0 0.0037 Malachite Green

The original Km determined using the Malachite Green assay is only 2-fold different from the one determined using the Amplex Red assay. This small difference may be due in part to the problems with utilizing these phosphate detection assays. Most surprising is the difference in Km observed in other species possessing a TPP enzyme (Table 10). The M. tb

TPP2 has an N-terminal domain of approximately 140 residues which is absent in the other

TPP enzymes shown. The function of this domain is currently unknown; however it may be contributing additional interactions with T6P increasing the activity of the enzyme. The potential role this domain is playing will be discussed further in Chapter 4.3.2.3.

4.3.1.4 Z’ determination and NIH screening

Much like the Quinaldine Red and PNP-M5 assays in Chapters 2 and 3, the suitability of the Amplex Red assay for high-throughput screening needed to be established 59.

Therefore, we determined the Z’ value for this particular assay using 48 positive control reactions, and 48 negative control reactions (Equation 1). Reactions contained 5 nM TPP2,

141

100 nM TreA, and 400 µM of T6P in addition to the standard Amplex Red working solution.

Negative control reactions omitted the TPP2. The rate measured between time points of two minutes and five minutes were used for activity measurements and absorbance measured at

530 nm. The Z’ for the assay was determined to be 0.76 (Table 11)

Table 11: TPP2 Amplex Red Z’ determination

Absorbance 530 nm (mAU/min)

Average (-) 9.7

Standard Deviation (-) 0.55

Average (+) 33.3

Standard Deviation (+) 1.4

Z’ score 0.76

The high Z’ score shows that the assay is suitable for HTS applications. As a result, we again utilized the NIH Clinical Collection screen to identify any potential inhibitors against TPP2. Compounds were screened at 100 µM and the activity compared to a control reaction. After screening all 450 compounds, only one demonstrated any measurable change in activity against TPP2. The compound found was piceid.

4.3.1.5 Piceid inhibition against TPP2

Piceid is a β-linked, glycosylated resveratrol derivative that is commonly found in grapes (Figure 48A) 100. We initially hypothesized that the compound may occupy the trehalose-binding site of TPP2 through the glucosyl moiety, and that resveratrol was occupying an additional hydrophobic site. The addition of piceid to the reaction causes an 142

extended lag period at the start, and after a period of time, the enzymatic activity returns to normal levels (Figure 48B). This is reproducible and concentration dependent; the higher the initial concentration of piceid, the longer the lag period. However, we did not know the mechanism through which the activity of the TPP2 returned.

Figure 48: Piceid causes lag time during the Amplex Red assay. (A) Cartoon structure of piceid, a glycosylated resveratrol derivative. (B) Piceid causes an extended lag time to occur during the Amplex Red assay before activity returns to normal levels. This is concentration dependent.

We initially hypothesized that TPP2 could potentially hydrolyze piceid, and that the observed lag period was caused by this. Once all the piceid was consumed, normal activity would resume. However, follow-up experiments where piceid was pre-incubated with TPP2 showed the same lag period regardless of incubation time. Therefore, we used the Malachite

143

Green assay an orthogonal method to confirm inhibition. Surprisingly, the Malachite Green assay showed no inhibition of TPP2 by piceid as reactions with and without the compound showed the same activity. Therefore, the observed lag period was the result of piceid interacting or inhibiting another part of the coupled assay.

Piceid is a resveratrol derivative, which itself is a powerful scavenger of reactive oxygen species 101. Hydrogen peroxide is produced during the assay when glucose is oxidized by the action of Glucose Oxidase. This hydrogen peroxide is then used by the

Horseradish peroxidase to convert Amplex Red to resorufin. As such, the most likely cause of the lag period seen in the assay is the result of piceid scavenging the peroxide produced.

Once all the piceid has been oxidized, the activity of enzyme appears to return to normal levels. As a result, piceid is not inhibiting TPP2 but blocking peroxide from accumulating in the assay. Unfortunately as a result, no compounds were found in the screen which appeared to inhibit the activity of TPP2.

4.3.2 Crystallization of TPP2

To gain greater insight into the biochemical workings of TPP2, we set out to crystallize and solve the structure of the protein. TPP2 is highly specific towards its substrate

T6P, having almost no activity using other phosphorylated sugars 95. Therefore, understanding the interactions between T6P and the enzyme could aid in the development of inhibitors towards TPP2 by exploiting novel interactions observed in the active site. However, despite many attempts, diffraction quality crystals, which would give suitable data for structural elucidation, could not be obtained. As such, efforts to determine the structure are ongoing. Below are several selected examples of conditions which produced protein crystals.

144

4.3.2.1 Initial TPP2 crystallization utilizing the TPP2-trehalose-AlFx complex

TPP2 likely proceeds through a pentacoordinate transition state resulting in nucleophilic attack of the phosphate on T6P 102. In this TPP2 crystallization experiment we utilized aluminum chloride and sodium fluoride to produce a transition state mimic within the protein 103. When these are reacted with the protein, in the presence of magnesium and trehalose, a transition state mimic adopted by T6P and the enzyme is formed. This could then aid in the crystallization of the protein. To do this, TPP2 was reacted with 3 mM trehalose, 1 mM MgCl2, 3 mM AlCl3 and 3 mM NaF for 1 hour. The resulting reaction was the purified using size exclusion chromatography to remove excess substrates that had not reacted with the TPP2. The protein was the concentrated to 10 mg/mL and screened using the Index

Screen. Crystallization was performed using the sitting drop method, under oil, and at 4 °C.

One condition, H2, 0.2 M potassium sodium tartrate tetrahydrate and 20% w/v

PEG3350 produced a single plate crystal that was diffraction quality. The TPP2-trehalose-

AlFx complex diffracted to a modest 3.5 Å resolution. However, there was long-range disorder along one crystal axis, which prevented us from obtaining data necessary to solve the crystal structure (Figure 49).

145

Figure 49: Diffraction of the TPP2-trehalose-AlFx crystals. Both frames were collected from a single crystal. Left) Diffraction shows good order in two dimensions. Right) This orientation exhibits significant disorder along the third axis

Using the initial success from the H2 condition, attempts were made to improve the quality of the TPP2 crystals. These included a 2D screen of the original H2 condition, changing the protein concentration, as well as performing Additive and Detergent screens. A single condition produced a crystal that was useful for X-ray diffraction experiments. TPP2 at a final concentration of 15 mg/mL was modified with the same trehalose-AlFx complex as previously described and crystalized in the H2 condition. Additionally, one compound, glycyl-gylcyl-glycine, from the Additive Screen yielded a crystal (Figure 50).

146

Figure 50: Reproduced TPP2-trehalose-AlFx crystals after optimization. However, this crystal did not diffract to suitable resolution for structural elucidation.

Despite the success in making this crystal, it did not diffract during X-ray diffraction experiments. Several more attempts were made to reproduce the initial crystals and improve upon them using the H2 condition. However reproducing the crystals was difficult, and most additives or detergents did not improve upon crystallization.

4.3.2.3 TPP2 crystallization utilizing a TPP2-n-dodecyl-β-D-maltoside -AlFx complex

The TPP2 from Mycobacterium tuberculosis is unique amongst other TPP enzymes in that it possesses an N-terminal domain of approximately 140 residues (Figure 51).

147

Figure 51: Sequence alignment between the M. tb and T. acidophilum TPP enzymes. The M. tb enzyme contains an N-terminal domain of 140 residues which is absent in many other TPP enzymes.

Secondary structural analysis of TPP2 shows that it contains four predicted alpha helical regions, but is mostly unstructured, particularly the initial 20 residues. A recent structure of the apo form of a TPP enzyme from Brugia malayi, which also contains this N- terminal domain, only showed density for the last two helices of this domain 97. This indicates that a majority of the domain is highly dynamic and could not be resolved. The dynamic nature of this N-terminal domain may have played a role with the issues of reproducibility and crystal disorder observed in previous TPP2 crystals. 148

It has also been shown that trehalose monomycolate production is stimulated by the addition of T6P 104, 105. Therefore, we hypothesized that TPP2 may utilize mycolated trehalose-6-phosphate as a substrate, in addition to its confirmed activity with T6P (Figure

52C). We speculated that the N-terminal domain may in fact play a role in binding to the mycolic acid portion of mycoyl-T6P substrate (Figure 52A).

Figure 52: TPP2 with modeled in N-terminal domain. (A) The T. acidophilum TPP structure with an oval representing the additional N-terminal domain found in the TPP2. The black circle is the location of the predicted active site with the catalytic magnesium ion. (B) Structure of n-dodecyl-β-D-maltoside used as a mimic for trehalose monomycolate. (C) Structure of trehalose-6-phosphate monomycolate.

As a result, improvements of TPP2 crystals may require stabilization of this domain.

To accomplish this we set out to co-crystallize the TPP2 in the presence of an esterified sugar 149

to mimic a mycolated-T6P. Esterified trehalose derivatives were not available at the time of these experiments and using TMM was unfeasible due to solubility issues with the molecule.

Therefore, we selected an n-dodecyl-β-D-maltoside detergent as a substrate analogue (Figure

52B).

In addition to crystallization with the detergent, we also utilized the AlFx transition state analogue that produced the initial TPP2 crystals. To do this, TPP2 was reacted with 1 mM n-dodecyl-β-D-maltoside, 1 mM MgCl2, 3 mM AlCl3 and 3 mM NaF for 2 hours. The reaction was then washed several times in crystallization buffer in order to remove excess reactants from the mixture using ultracentrifugation. This reaction was then used for crystallization attempts with the Index screen. Both hanging drop and sitting drop under oil were utilized.

One condition produced very large plate crystals and contained 0.2 M KCl, 50 mM

HEPES pH 7.5, and 35% v/v pentaerythritol ethoxylate (15/4 EO/OH). The condition was optimized by adjusting the ratio of solution to water. Conditions that had crystals were 30%,

40%, and 70% well solution (Figure 53). Even though the condition produced sizeable crystals, the resolution obtained from the X-ray diffraction experiments was very low (< 20

Å). Despite this, the idea that stabilization of the N-terminal domain could produce better diffracting crystals if a more appropriate trehalose monomycolate mimic could be found, is still sound.

150

Figure 53: Crystals from the TPP2-n-dodecyl-β-D-maltoside -AlFx complex. Percentages represent the dilution of the well solution.

4.3.2.4 TPP2 crystallization utilizing a TPP2 4-dodecyl-trehalose vanadate complex

Stabilization of the N-terminal domain of TPP2 remains a sound hypothesis to promote crystallization of the protein. However, using the n-dodecyl-β-D-maltoside may not have been a suitable substrate mimic to stabilize the N-terminal domain. As such, a better mimic was needed. Therefore, Anatrace (Maumee, OH) synthesized a 4-dodecyl-treahalose.

Despite not being an exact mimic of the natural trehalose monomycolate, it may be a better approximation than the maltoside. Additionally, we utilized orthovanadate as a transition

151

state mimic rather than the aluminum fluoride. Orthovanadate complexes have been used extensively to mimic the transition states of many phosphoryl transferase enzymes 106. Using this, TPP2 was co-crystallized in the presence of the 4-dodecyl-trehalose and orthovanadate.

TPP2 at 9 mg/mL was reacted with 1 mM MgCl2, 1 mM sodium orthovanadate, and 1 mM of the 4-dodecyl-trehalose. This was screened using the Index Screen, utilizing both hanging drop and sitting drop under oil. After several months, one condition, 100 mM HEPES pH 7.5, and 1.4 M sodium citrate, produced crystals (Figure 54).

Figure 54: Crystals obtained from the TPP2 4-dodecyl-treahalose vanadate complex.

These crystals are a recent discovery and further optimization is underway. Cursory attempts to break the crystals indicate that they are protein. Additionally, they are of a different morphology than previous TPP2 which tended to form plates. This change could

152

indicate that the crystals have packed differently due to the presence of the 4-dodecyl- trehalose and therefore give better diffraction quality.

4.3.2.5 Crystallization of an inactive mutant, TPP2-D147, and T6P

TPP2 is a member of the haloacid dehydrogenase (HAD) superfamily, which includes a diverse set of enzyme and includes P-type ATPase, phosphoserine phosphatase and other trehalose phosphatases 97, 107. Based on sequence similarity of a trehalose-6-phosphate phosphatase from Thermoplasma acidophilum, it is likely that D147 of the M. tb is the catalytic nucleophile of the enzyme. Using this information, we constructed the TPP2-

D147N variant, which should be capable of binding substrate but unable to catalyze the de- phosphorylation of T6P.

The activity of the D147N variant was tested utilizing the Amplex Red assay, and the relative activity was compared to the wild-type enzyme (Figure 55). The mutant displayed almost no activity compared to the wild-type, and any signal observed was likely the result of background hydrolysis of T6P to trehalose.

153

Figure 55: Activity comparison of the wild-type TPP2 to the D147N variant. The D147N mutant displays almost no activity due to the absence of the catalytic nucleophile.

Once we verified the inactivity of the mutant, crystallization attempts with the T6P substrate were performed. The protein solution contained 3 mg/mL TPP2-D147N, 2 mM

MgCl2, and 1 mM T6P. The concentration of protein at 3 mg/mL is much lower than most other experiments. One additional modification was to also reduce the concentration of the

Tris in the crystallization buffer from 20 mM to 10 mM. Crystallization experiments used the hanging drop method with the Index Screen. Condition 53 of the Index Screen, 0.2 M ammonium acetate, 0.1 M Tris pH 8.5, and 45% v/v (+/-)-2-methyl-2,4-pentanediol, produced crystals (Figure 56). 154

Figure 56: Initial crystals obtained from the TPP2-D147N and T6P complex.

The identification of these crystals is very recent, and experiments are ongoing to reproduce, and improve upon them. Like those crystals utilizing the 4-dodecyl-trehalose- orthovandate, the morphology of these crystals is different than the plate crystals. Again this may represent a change in packing that may have improved the quality of diffraction that will be obtained.

4.4 Conclusions and Future Works

We were successful in designing a new assay to measure the activity of the TPP2.

The Amplex Red assay does not rely on the release of phosphate like traditional methods of measuring phosphatase activity. This assay is sensitive and reproducible, which is evident in the high Z’ value calculated. Utilizing the assay we screened a number of compounds to identify inhibitors for TPP2. However, none were identified that were legitimate hits.

155

Utilizing different compound libraries could yield better results in the future. The ease of the

Amplex Red assay makes screening large compound libraries easy and timely. We also designed and tested the activity of a TPP2 mutant (TPP2-D147N), which showed almost no activity compared to the wild-type.

Despite our best attempts, the crystal structure of the TPP2 enzyme has eluded us.

Various conditions did provide crystals, though most were not of sufficient quality to give data which could be used to solve the structure. Recent attempts have yielded two new crystal forms with different compounds bound. Further optimization of these crystals is needed, however as they have yet to be tested for diffraction, they may be suitable at this point.

156

Chapter 5

Structural characterization of the Mycobacterium tuberculosis Mycothiol-S-transferase

5.1 Background

5.1.1 Biosynthesis of mycothiol (MSH)

Mycobacterium tuberculosis and many other Actinomycetes utilize mycothiol (MSH) as a low molecular weight thiol that is involved in protecting the bacterium from oxidative stresses, electrophilic toxins, free radicals, and xenobiotics 108. Similar systems exists across other organisms, most notably the glutathione (GSH) that is found in eukaryotes and other bacteria, and bacillithiol (BSH) in Bacillus species (Figure 57) 109, 110. MSH is produced in

Mycobacterium tuberculosis by a series of enzymes within the MshA/A2/B/C/D pathway 108.

157

Figure 57: The three most common reducing thiols. Mycothiol is common in Actinomycetes. Bacilithiol is found in the Bacillus species, while Glutathione can be found in eukaryotes and other bacteria.

MshA, a glycosyltransferase, catalyzes the transfer of UDP-N-acetyl-glucosamine

(UDP-GlcNAc) to inositol-1-phosphate (Ins1P) to yield N-acetyl-glucosamine-inositol-1- phosphate (GlcNAc-Ins-1-P). GlcNAc-Ins1P is then dephosphorylated an unidentified phosphatase to yield N-acetyl-glucosamine-inositol, MshA2 (GlcNAc-Ins). The next step in mycothiol biosynthesis is the deacetylation of the GlcNAc-Ins by the deacetylase MshB; the product of this reaction is glucosamine-inositol (GlcN-Ins). MshC generates cysteine- glucosamine-inositol (Cys-GlcN-Ins) via the transfer of cysteine to GlcN-Ins in an ATP requiring reaction. The final step in this pathway is the acetylation of the Cys-GlcN-Ins to produce mycothiol by the mycothiol synthase MshD (Figure 58) 108.

158

Figure 58: Mycothiol biosynthetic pathway. Enzymes are colored in blue, substrates in green, and products in red.

5.1.2 Utilization of Mycothiol

Mycothiol is readily oxidized both in vitro and in vivo. The result of this is the formation of an oxidized dimer of mycothiol, mycothione. Re-activation of the mycothione to two mycothiol molecules is accomplished by a mycothione reductase. In Mycobacteria 159

and other Actinomycetes this is performed by the NADPH dependent enzyme, Mcr 111

(Figure 59).

Reduced MSH serves as the primary reductant to detoxify a variety of different compounds and oxidants that the bacteria may encounter. Compounds neutralized by MSH are ultimately transported out the cell, via two-step process (Figure 59). First, mycothiol-S- conjugates are cleaved at the amide by the Mycothiol-S-conjugate amidase (Mca). This yields free GlcN-Ins and the mercapturic acid conjugate. The freed Glc-Ins can be recycled back into the MSH biosynthetic pathway to produce additional MSH. The mercapturic acid conjugate is the transported outside of the cell by an as of yet identified mercapturic acid conjugate transporter(s) (Figure 59)108, 112.

Figure 59: Detoxification by the mycothiol. monoChlorobimane is (mBCl) is a common thiol reactive compound for studying thiols and serves as the example “toxin”.

160

5.1.3 The Mycobacterium tuberculosis mycothiol-S-transferase, MtMST

The identities of both MycR and Mca have been known for some time. However, the identity of the enzyme responsible for the transfer of mycothiol to target molecules was unknown. Very recently, Newton et al. identified the enzyme required for the transfer of

MSH 113. The Mycobacterium tuberculosis mycothiol-S-transferase (MtMST) was capable of modifying monoChlorobimane (mBCl), a thiol reactive compound, with MSH, in addition to several other thiol reactive antibiotics 113. They also demonstrated the S-transferase activity of a bacillithiol-S-transferase from Bacillus subtils (BsBST), a glutathione-S-transferase from

Enterococcus faecalis (EfGST), and the mycothiol-S-transferase from Mycobacterium smegmatis (MsMST). Based on phylogenic analysis, all four of these enzymes belong to the

DinB_2 superfamily of enzymes 113. The MtMST specifically belongs to the domain of unknown function 664 family (DUF664). Until recently the only enzyme from the DinB_2 family with a confirmed function was the mycothiol-dependent maleylisomerase from

Corynebacterium callunae 114. Therefore, the discovery of these enzymes defines one potential role for this family of enzymes.

5.1.3 Mycothiol, drug resistance, and a new therapeutic target

Specific mutations or insertions to the genes encoding mshA/B/C/D enzymes in

Mycobacterium smegmatis can increase the susceptibility of the bacterium to oxidative stress, alkylating agents, and some antibiotics, while at the same time, increasing the resistance of the bacterium to some other antibiotics 108, 115, 116, 117. However, similar studies performed in

M. tb, specifically to mshA/C, produced bacteria that were non-viable, while the corresponding M. smegmatis bacteria were still viable 118, 119, 120. The hypothesis to explain this phenomenon was that M. smegmatis grows much faster than M. tb and can likely 161

outgrow many of the harmful stresses placed on it, while the M. tb cannot 118. This suggests that the mycothiol pathway may play a significant role in the evasion of the bacterium by many of the common anti-tubercular drugs whereby MSH neutralizes them.

The apparent essentiality of the mycothiol pathway opens up a new avenue for drug development against TB. Therefore, designing inhibitors to block either the production or utilization of MSH could provide a new therapeutic avenue to fight infections. Both the mshA and mshC genes have been shown to be essential. However, the essentiality of the MtMST has not been proven yet. Despite this, designing inhibitors that target the enzyme could potentially help in a number of ways. Firstly, inhibiting the transfer of MSH to exogenous molecules may be enough to kill the bacterium outright, or cause it to grow at a much slower rate than was observed in M. smegmatis strains 120. Additionally, blocking the activity of

MtMST could potentially make the bacterium more susceptible to currently available anti- tubercular drugs 115. Therefore, we determined that understanding the structure and biochemical aspects of the MtMST was worth investigating.

5.2 Materials and Methods

5.2.1 Molecular cloning of Rv0443c

A synthetic, codon-optimized gene encoding MtMST was purchased from Integrated

DNA Technology and was recombined with a pET32-based plasmid using the Gibson assembly kit (New England BioLabs) to make the pDR32-Rv0443c plasmid. This construct contains an N-terminal, cleavable thioredoxin and poly-histidine tag. The gene sequence was confirmed by DNA sequencing (Eurofins MWG Operon).

162

5.2.2. Expression and purification of MtMST

The pDR32- Rv0443c plasmid was used to transform E. coli BL21* cells. Bacterial cells were cultured at 37 °C in Luria Broth containing ampicillin (100 mg/L) until an OD600 nm of

0.6 was reached. Protein expression was induced by addition of 1 mM isopropyl -D-1- thiogalactopyranoside at 16 °C. The bacterial cells were harvested after a 24-hour induction using centrifugation. The bacterial pellet was resuspended in a buffer containing 20 mM Tris pH 8.0, 5 mM imidazole, 0.2 M sodium chloride, and 5 mM -mercaptoethanol.

The resuspended cells were incubated with Lysozyme (10 μM) and DNaseI (100 μM) for thirty minutes on ice prior to lysis by sonication. The crude lysate was then pelleted by centrifugation at 15,000 g for 30 minutes. The supernatant was applied to a 5 mL HiTrap

TALON crude column (GE Healthcare) that had been equilibrated with buffer A. MtMST bound to the cobalt column was washed with 15 column volumes of resuspension buffer and was eluted using a linear gradient with a concentration of imidazole ranging from 5 mM to

150 mM over 15 column volumes. The fractions containing MtMST were pooled and dialyzed overnight against 20 mM Tris pH 8.0, 0.2 M sodium chloride and 5 mM - mercaptoethanol. Recombinant human rhinovirus 3C protease was added to cleave the poly- histidine and thioredoxin tags. The dialyzed MtMST was again subjected to cobalt affinity chromatography to remove the different tags as well as the protease. The protein purity at all stages of purification was assessed using SDS-PAGE.

Ammonium sulfate (2.8 M) precipitation was used to concentrate MtMST. After centrifugation, the pelleted protein was dissolved in the crystallization buffer (20 mM Tris pH 8.0, 0.3 mM tris(2-carboxyethyl)phosphine) and dialyzed overnight against the same buffer. The protein concentration was determined using absorbance spectroscopy (280 nm) 163

and the theoretical extinction coefficient (36,440 M-1 cm-1) calculated by the ProtParam function from the ExPASY proteomics server.

5.2.3 Crystallization and structural determination

Crystals of MtMST were grown using the hanging-drop vapor diffusion technique.

Initial crystals of the MtMST were obtained using a protein concentration of 9.4 mg/mL and a well solution containing 0.1 M Tris pH 8.5, 0.2 M magnesium chloride and 25 % w/v polyethylene glycol 3350. Improved crystals were obtained by the addition of n-nonyl--D- glucoside at a final concentration of 4.5 mM. Additionally, the MtMST was co-crystallized in the presence of 2-deoxy-2-azido-glucosazide inositol at a final concentration of 3 mM. The crystal was flashed-cooled in liquid nitrogen prior to data collection. Diffraction data were collected at the LS-CAT beamline at the Advance Photon Source Argonne National

Laboratory (APS-ANL, IL). HKL2000 was utilized to index, integrate, and scale the diffraction data 50.

Molecular replacement was performed using the Enterococcus faecalis EF_3021

(PDB accession code: 3CEX) as a search model 51, 52. Rigid body refinement, simulated annealing, positional and B-factor refinements were carried out using PHENIX 53. Manual refinement of the structure was performed using COOT and structural validation was performed using Molprobity 54, 55.

5.2.4 Virtual Screening using Glide

The mycothiol molecule was built in PHENIX using the eLBOW and the REEL ligand restraints editor module 56. The MtMST was prepared for docking studies using the protein preparation wizard in Glide (Schrodinger). Geometric and energy minimization was 164

performed at pH 7.0 using the OPLS_2005 force field. The docking grid was built around the

24 surface near to the dimer interface of the protein using the grid generation wizard in

Glide.

5.2.5 Assay for assessing MtMST activity and antibiotic modification

Assessment of MtMST activity was performed in a similar manner as Newton et. al.

113. Enzymatic reactions were carried out in 100 µL volumes at 37 °C on a synergy H4

Hybrid Reader (BioTek) using 1 μM of enzyme. Buffer conditions for the assay were as follows: 20 mM HEPES, 100 mM NaCl, and 100 μM DTT. MSH (JEMA Biosciences) and monoChlorobimane (Sigma-Aldrich) were added to final concentrations of 50 μM and 50

μM respectively. All reactions were performed in triplicate. Reactions were performed in opaque, black 384-well plates and activity monitored using fluorescence intensity with excitation and emission wavelengths of 394 nm and 490 nm respectively. Progress curves were measured over the course of 40 minutes. Reaction rates were calculated using the Gen5 software (BioTek). Modification of antibiotics was tested using the above assay conditions.

Reactions containing buffer, MtMST, MSH, and 1 mM of an antibiotic were incubated for 4 hours at 37 °C. After this time, mBCl was added to the mixture and the activity measured as detailed previously.

5.3 Results and Discussion

5.3.1 Crystallization strategies for MtMST

To gain greater insight into the biochemical workings of the MtMST, we attempted to crystallize, and determine the structure of the enzyme. In the absence of having the natural

165

substrate, mycothiol, an advanced synthetic intermediate of its organic synthesis, 2-deoxy-2- azido-glucosazide inositol, was used in co-crystallization attempts to potentially identify the enzyme active site, and improve crystallization overall (Figure 60).

Figure 60: Structures of Mycothiol and the advanced synthetic intermediate used for crystallization attempts.

Initial crystallization screening was done using the Index Screen. Several conditions were found that produced crystals of varying morphologies and sizes (Figure 61). The condition that produced the best crystals contained 0.1 M Tris pH 8.5, 0.2 M magnesium chloride and 25 % w/v polyethylene glycol 3350 (Figure 61C). These crystals were large plates were used for further optimization.

166

Figure 61: Initial crystals of MtMST. The plate crystals produced in panel C were chosen or further optimization and contained 0.1 M Tris pH 8.5, 0.2 M magnesium chloride and 25 % w/v polyethylene glycol 3350.

A 2D screen of the hit condition, in addition to both Additive and Detergent screens were performed to optimize the crystals. The final condition that produced the best crystal contained 0.1 M Tris pH 8.5, 0.2 M magnesium chloride and 25 % w/v polyethylene glycol

3350, in addition to 4.5 mM of n-nonyl--D-glucoside detergent (Figure 62). This crystal was ultimately utilized for the structural studies.

167

Figure 62: Diffraction quality crystal of MtMST. This final condition contained 0.1 M Tris pH 8.5, 0.2 M magnesium chloride and 25 % w/v polyethylene glycol 3350, 4.5 mM of n-nonyl--D-glucoside, and 3 mM 2-deoxy-2-azido-glucosazide inositol.

5.3.2 Data collection and structural refinement of the MtMST structure

The crystal of the MtMST was analyzed at Argonne National Labs and diffraction data extended out to 1.36 Å. The protein, which crystallized in the C2221 space group, had the following unit cell parameters: a = 65.4 Å, b = 98.9 Å, c = 50.0 Å; α = β = γ = 90° (Table

12). Based on sequence alignment, the MtMST shared about 40% sequence identity the

EfGSTfrom Enterococcus faecalis. The EfGST had demonstrated GSH transferase activity, and the structure was known 113. Therefore, molecular replacement was used to phase the structure utilizing this Enterococcus faecalis EF_3021 (PDB accession code: 3CEX) as a search model. The final structure contains a single molecule in the unit cell. The protein is completely resolved from both termini. The MtMST contains the highly conserved 4-helical bundle domain found in the DinB superfamily as it was hypothesized from the phylogenic analysis (Figure 63) 113, 121.

168

Table 12: Data collection and refinement statistics for the MtMST. Values for the highest shells are in parentheses

Data Collection Apo-MtMST PDB Code n/a

Space Group C2221 Unit Cell Dimensions a, b, c (Å) 65.4, 98.9, 50.0 α, β, γ (°) 90, 90, 90 Resolution (Å) 50.0 – 1.36

Rmerge 9.0 (61.4) I/σI 32.0 (3.3) Completeness (%) 89.4 (72.9) Redundancy 12.3 (5.6) Refinement Resolution (Å) 29.4-1.36 Unique Reflections 31,567

Rwork/Rfree 0.1912/0.2229 No. Atoms Protein 1368 Water 228 Ligand n/a B-factors (Å2) Protein 17.3 Water 27.7 Ligand n/a R.m.s deviations Bond lengths (Å) 0.009 Bond angles (°) 1.11 Ramachandran Favored (%) 100 Outliers 0

169

Figure 63: Overall MtMST structure. MtMST contains the highly conserved 4-helical bundle domain of the DinB superfamily.

5.3.3 MtMST monomeric structure contains the highly conserved, 4-helical bundle domain of the DinB superfamily

The key feature shared by MtMST and the DinB superfamily is the highly conserved,

4-helical bundle domain (Figure 63) 121. This 4-helical bundle along with the intervening loop regions defines the overall structure of the protein. The first helix of the 4-helical bundle at the N-terminus (1) is 25 residues long (Figure 64A) and leads into a long, but ordered, 18 residue loop (12L) which itself contains a small 5 residue alpha helix (Figure 64C). This loop dives over the top of the bundle leading into the second -helix (2) that contains 20

170

residues (Figure 648B). 2 is connected to the next helix (3) by a very long loop of 39 residues that crosses over the bundle (23L) (Figure 5-8C). This loop region contains several helices, giving it a loop-helix-loop-helix-turn-helix structure. This loop transitions into 3, an 18-residue helix, which runs parallel to 2 (Figure 2A). A final, 21-residue loop

(34L) connects 3 and 4 (Figure 64D). This loop contains a single, 5-residue alpha helix, followed by a beta-hairpin. In this structure, the sidechains of the loop connecting 3 and 4 are largely unresolved. The final helix (4) is 25 residues and terminates at the C- terminus. (Figure 64B) This gives the MtMST structure the characteristic up-down-down-up

4-helical bundle, with 1 and 4 abutting each other, with 2 across from 1 and 3 across from 4.

171

Figure 64: Structural elements of MtMST. (A) 13 helices, (B) 24 helices, (C) 12L and 23L, (D) 34L.

Many enzymes of the DinB superfamily exist as dimers and MtMST is no exception

121 . MtMST crystalized in the C2221 space group with one molecule in the asymmetric unit.

However, a crystallographic dimer was present and this is commonly observed in the

172

structures for this family of enzymes. Additionally, the presence of an MtMST dimer was confirmed using native-PAGE and size exclusion chromatography (Figure 65).

Figure 65: Determination of the MtMST dimer. Chromatograms from size-exclusion indicate (left) that MtMST elutes sooner than the similarly sized Lysozyme indicating a potential dimer. Native-PAGE also suggests the presence of a dimer (right).

The MtMST dimer, like others of the DinB superfamily, is formed by an extensive interface between the 14 surface of each monomer 121. This dimerization interface, in addition to the 24 surface and connecting loop regions, form a pocket at the 24 surface (Figure

66A). This pocket has been hypothesized to be the location of the active site for this enzyme and as a result of this the active site of each monomer are on opposing faces (Figure 66B) 113.

173

Figure 66: MtMST dimer and binding pocket. (A) Stereo view of the MtMST dimer. (B) Top view showing the 14-dimerization interface. (C) Surface view of the pocket (black circle) formed at the 24 as a result of the dimer and 12 and 23 loop regions.

To truly ascertain that this pocket is the active site, 2-deoxy-2-azido-glucosazide inositol, an advanced synthetic intermediate of MSH, was co-crystalized with the MtMST.

However, no density was observed for the molecule during refinement of the X-ray crystal structure, suggesting that the compound was unable to bind to the protein. The insufficient

174

binding affinity may be due to the absence of the N-acetyl cysteine moiety of the natural substrate, MSH.

5.3.4 Docking Studies reveal a potential Mycothiol binding site in MtMST

Because crystallization attempts with the 2-deoxy-2-azido-glucosazide inositol were not successful, identifying the active site of the MtMST is difficult. In lieu of obtaining the structure of MtMST in complex with the 2-deoxy-2-azido-glucosazide inositol, Glide docking studies were performed in an attempt to identify the mycothiol-binding site. We hypothesized that the potential active site is likely the pocket produced by 24, and the 1 of the other monomer in the MtMST dimer.

Firstly, many of the DinB family of enzymes are metal binding, with structures submitted to the Protein Databank having bound nickel or zinc ions. These metals are coordinated by a triad of histidines on the 24 surface (Figure 67) 121. It was suggested that the metal in this site is used to coordinate the sulfhydryl group of the reducing thiol during catalysis. However, not all DinB structures show metal binding to this particular site including the MtMST structure, which showed no density for a divalent metal ion at this site and did not require a divalent metal ion for enzymatic activity in vitro 113.

Additionally, there is a large hydrophobic patch adjacent to the 24 surface which could potentially interact with target molecules (Figure 67). The area of the patch leading into the 24 surface is lined with several aromatic residues which have been hypothesized to act as a funnel towards the binding pocket 113. This funnel of the MtMST is formed by four residues: Y91, W48, and W139 (Figure 67). Y91 and W48 are well resolved in the structure,

175

but W139 is on the dynamic and poorly resolved 34 loop which may become less dynamic upon the binding of a substrate.

Figure 67: The MtMST active site. (Left) a large hydrophobic patch (black circle) next to the hypothesized active site that may bind to targeted molecules. The black diamond indicates where divalent metals are found in many other DinB proteins, however no metal was observed in the MtMST. (Right) The three aromatic residues that are thought to funnel hydrophobic molecules towards the active site. Y91 may be playing the role of a general base.

To probe the presumed 24 surface as the binding site, we used Glide to perform docking studies. A docking grid was generated around the pocket at the 24 surface, and mycothiol was docked within it. Glide found a low energy site within the pocket, which resulted in a modest Glide docking score of –5.00. Examination of the site reveals several chemical features that further suggest this site is potentially MSH binding site.

The inositol moiety of the mycothiol is binding into a highly negative potential surface of the 24 surface (Figure 67A). This is characteristic of carbohydrate binding enzymes, as the acidic and other polar residues are often required to coordinate the hydroxyl groups of the sugars and inositol-containing compounds 122. A second interesting feature is a

176

small hydrophobic dimple in the surface in which the N-acetyl moiety of the mycothiol appears to bind (Figure 67A). The sulfhydryl moiety is sitting in a slightly positively charged surface that may help coordinate the group while it is in a deprotonated state (Figure 67A). In addition to forming the hydrophobic funnel towards the MSH, Y91 of the 23L may function as a general base for the deprotonation of the cysteine as is seen in other glutathione

S-transferases (Figure 67B) 123. However, at almost 7 Å in distance, this interaction is much too far away to occur. It is possible that structural rearrangement occurs upon binding of both substrate and target molecule, which would bring Y91 into closer proximity of the MSH.

Figure 68: Glide docking of MSH. (A) Docking shows that the MSH binds to a relatively negative potential surface with varying moieties to coordinate the MSH. (B) Y91 may be playing the role of general base to deprotonate the MSH, however the distance of this interaction is very far.

5.3.5 Assessing the activity of MtMST using monoChlorobimane

As previously stated, the knockout of the genes which produce mycothiol in

Mycobacterium smegmatis alters the way antibiotics affect the organism, increasing the susceptibility of the bacterium to some antibiotics, while making it more resistant to others

108, 115, 116, 117. Therefore, we speculated that the mycothiol pathway may play a significant role in the evasion of the Mycobacterium tuberculosis by many of the common anti- tubercular drugs whereby MSH neutralizes them. 177

Before we could test the modification of antibiotics by MSH, an assay needed to be developed to test for MtMST activity in vitro. Previously, Newton et. al. utilized halogenated bimane derivatives to determine the activity of the MtMST 113. Specifically, they used monoChlorobimane (mBCl), which upon thiolation by MtMST, produces a fluorescent bimane product. This allows for easy determination of activity (Figure 69). Using this assay, we showed it to have good signal to noise ratio, and was suitable for doing further experiments (Figure 69). Using this assay we probe the modification of antibiotics by

MtMST.

Figure 69: monoChlorobimane assay for MtMST activity. (Top) general schematic of the cholorobimane assay. (Bottom) Progress curves of the reactions with MtMST, MSH, and mBCl. Positive control contains enzyme, while the negative one does not.

178

5.3.6 Modification of antibiotics by MSH

To determine modification of these antibiotics, a competitive kinetic assay using mBCl is used. If MSH is being used to modify the antibiotics, a decrease in signal will be observed as less MSH will be available to modify the mBCl. Four antibiotics were chosen initially: isoniazid, erythromycin, spectinomycin, and pyrazinamide, all known anti- tubercular drugs. Initial assays showed very little change in activity in the presence of the antibiotics. We speculated that this may be due to the reactivities of the targets. It was likely that the mBCl was a much better thiol acceptor than the antibiotics. Therefore, we pre- incubated the antibiotics with MtMST and MSH for four hours prior to the addition of the mBCl. These reactions showed a marked difference in activity (Figure 70). Isoniazid and erythromycin showed approximately 80% and 85% the activity of the control reaction. The spectinomycin and pyrazinamide performed better as substrates for the MSH modification, with 60% and 70% activity respectively.

Figure 70: Modification of anti-tubercular drugs by MSH. After a 4 hour incubation reactions with the different antibiotics showed a decrease in activity compared to the control reaction.

This initial data suggest that MtMST is able to modify antibiotics. However, characterization of the products needs to be performed, and these experiments are ongoing. 179

5.4 Conclusions and future works

MtMST is a member of the DUF664 family of the DinB superfamily of enzymes.

Here we have solved the first structure of the apo-form of the wild-type MtMST. The

MtMST has the highly conserved, 4-helical bundle domain that is common to this family of proteins. However, we were unable to ascertain the location of the active site using an advance synthetic intermediate of MSH. Therefore, we performed Glide docking studies to find a potential docking site for MSH. Glide identified a small pocket at the 24 surface which gave a reasonable Glide score. Despite this, the actual active site needs to be found.

The crystal structure of MtMST in a complex with MSH and an MSH-bimane complex are ongoing, and initial crystals have been produced. Using these we can identify the binding site of MSH and compounds that it modifies.

We have shown preliminary evidence that MtMST is able to modify common antibiotics used to treat TB infections. Many more antibiotics need to be screened using the assay. Additional confirmation needs to be done to show modification of the antibiotics, potentially using mass spectrometry. Once this is confirmed, identifying where the MSH has been added into the molecule needs to be determined.

180

References

1. WHO. Global tuberculosis report 2013. World Health Organization, (2013).

2. Donoghue HD, et al. Tuberculosis: from prehistory to Robert Koch, as revealed by ancient DNA. The Lancet infectious diseases 4, 584-592 (2004).

3. Brosch R, et al. A new evolutionary scenario for the Mycobacterium tuberculosis complex. Proceedings of the national academy of Sciences 99, 3684-3689 (2002).

4. OFFICIAL T. Diagnosis and treatment of disease caused by nontuberculous mycobacteria. Am Rev Respir Dis 142, 940-953 (1990).

5. Dover LG, Alderwick LJ, Brown AK, Futterer K, Besra GS. Regulation of cell wall synthesis and growth. Current molecular medicine 7, 247-276 (2007).

6. Sakamoto K. The pathology of Mycobacterium tuberculosis infection. Veterinary Pathology Online 49, 423-439 (2012).

7. Van Pittius NG, Gamieldien J, Hide W, Brown GD, Siezen RJ, Beyers AD. The ESAT-6 gene cluster of Mycobacterium tuberculosis and other high G+ C Gram- positive bacteria. Genome Biol 2, 44.41-44.18 (2001).

8. Murray PR, Rosenthal KS, Pfaller MA. Medical Microbiology, with STUDENT CONSULT Online Access, 7: Medical Microbiology. Elsevier Health Sciences (2013).

9. Draper P, Daffé M. The cell envelope of Mycobacterium tuberculosis with special reference to the capsule and the outer permeability barrier. Tuberculosis, 261-273 (2005).

181

10. Glickman MS, Jacobs WR. Microbial pathogenesis of Mycobacterium tuberculosis: dawn of a discipline. Cell 104, 477-485 (2001).

11. Smith I. Mycobacterium tuberculosis pathogenesis and molecular determinants of virulence. Clinical microbiology reviews 16, 463-496 (2003).

12. Aderem A, Underhill DM. Mechanisms of phagocytosis in macrophages. Annual review of immunology 17, 593-623 (1999).

13. Keane J, et al. Infection by Mycobacterium tuberculosis promotes human alveolar macrophage apoptosis. Infection and immunity 65, 298-304 (1997).

14. Kaufmann SH. How can immunology contribute to the control of tuberculosis? Nature reviews Immunology 1, 20-30 (2001).

15. Flynn JL, Chan J. Immune evasion by Mycobacterium tuberculosis: living with the enemy. Current opinion in immunology 15, 450-455 (2003).

16. Mann FM, Xu M, Chen X, Fulton DB, Russell DG, Peters RJ. Edaxadiene: a new bioactive diterpene from Mycobacterium tuberculosis. J Am Chem Soc 131, 17526- 17527 (2009).

17. Ehrt S, Schnappinger D. Mycobacterial survival strategies in the phagosome: defence against host stresses. Cellular microbiology 11, 1170-1178 (2009).

18. Russell DG. Who puts the tubercle in tuberculosis? Nature Reviews Microbiology 5, 39-47 (2007).

19. Pieters J. Mycobacterium tuberculosis and the macrophage: maintaining a balance. Cell host & microbe 3, 399-407 (2008).

20. Herzog H. History of tuberculosis. Respiration 65, 5-15 (1998).

21. Bell E. A souped-up version of BCG. Nature Reviews Immunology 5, 746-746 (2005).

182

22. Fine PE, Carneiro IA, Milstien JB, Clements CJ. Issues relating to the use of BCG in immunization programmes. A discussion document Geneva: World Health Organization 45, (1999).

23. Wood AJ, Iseman MD. Treatment of multidrug-resistant tuberculosis. New England Journal of Medicine 329, 784-791 (1993).

24. Weiner M, et al. Association between acquired rifamycin resistance and the pharmacokinetics of rifabutin and isoniazid among patients with HIV and tuberculosis. Clinical infectious diseases 40, 1481-1491 (2005).

25. Mukherjee JS, et al. Programmes and principles in treatment of multidrug-resistant tuberculosis. The Lancet 363, 474-481 (2004).

26. De Logu A, et al. In vitro antimycobacterial activity of newly synthesised S- alkylisothiosemicarbazone derivatives and synergistic interactions in combination with rifamycins against Mycobacterium avium. International journal of antimicrobial agents 26, 28-32 (2005).

27. Falzari K, Zhu Z, Pan D, Liu H, Hongmanee P, Franzblau SG. In vitro and in vivo activities of macrolide derivatives against Mycobacterium tuberculosis. Antimicrobial agents and chemotherapy 49, 1447-1454 (2005).

28. Jagannath C, Reddy MV, Kailasam S, O'Sullivan J, Gangadharam P. Chemotherapeutic activity of clofazimine and its analogues against Mycobacterium tuberculosis. vitro, intracellular, and in vivo studies Am J Respir Crit Care Med 151, 1083-1086 (1995).

29. Zumla A, Nahid P, Cole ST. Advances in the development of new tuberculosis drugs and treatment regimens. Nature reviews Drug discovery 12, 388-404 (2013).

30. Moonan PK, et al. Does directly observed therapy (DOT) reduce drug resistant tuberculosis? Bmc Public Health 11, (2011).

31. Ford CB, et al. Use of whole genome sequencing to estimate the mutation rate of Mycobacterium tuberculosis during latent infection. Nature genetics 43, 482-486 (2011).

183

32. Udwadia ZF, Amale RA, Ajbani KK, Rodrigues C. Totally drug-resistant tuberculosis in India. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America 54, 579-581 (2012).

33. Shankar EM, et al. HIV–Mycobacterium tuberculosis co‐infection: a ‘danger‐couple model’of disease pathogenesis. Pathogens and disease 70, 110-118 (2014).

34. Boulle A, et al. Outcomes of nevirapine-and efavirenz-based antiretroviral therapy when coadministered with rifampicin-based antitubercular therapy. Jama 300, 530- 539 (2008).

35. Preiss J, Romeo T. Physiology, biochemistry and genetics of bacterial glycogen synthesis. Advances in microbial physiology 30, 183-238 (1990).

36. Schaechter M. Encyclopedia of microbiology. Academic Press (2009).

37. Ballicora MA, Iglesias AA, Preiss J. ADP-glucose pyrophosphorylase, a regulatory enzyme for bacterial glycogen synthesis. Microbiology and Molecular Biology Reviews 67, 213-225 (2003).

38. Wilson WA, et al. Regulation of glycogen metabolism in yeast and bacteria. FEMS microbiology reviews 34, 952-985 (2010).

39. Chandra G, Chater KF, Bornemann S. Unexpected and widespread connections between bacterial glycogen and trehalose metabolism. Microbiology 157, 1565-1572 (2011).

40. De Smet KA, Weston A, Brown IN, Young DB, Robertson BD. Three pathways for trehalose biosynthesis in mycobacteria. Microbiology 146, 199-208 (2000).

41. Sani M, et al. Direct visualization by cryo-EM of the mycobacterial capsular layer: a labile structure containing ESX-1-secreted proteins. PLoS pathogens 6, e1000794 (2010).

42. Geurtsen J, et al. Identification of mycobacterial α-glucan as a novel ligand for DC- SIGN: involvement of mycobacterial capsular polysaccharides in host immune modulation. The Journal of immunology 183, 5221-5231 (2009).

184

43. Mendes V, Maranha A, Alarico S, Empadinhas N. Biosynthesis of mycobacterial methylglucose lipopolysaccharides. Natural product reports 29, 834-844 (2012).

44. Jackson M, Brennan PJ. Polymethylated polysaccharides from Mycobacterium species revisited. Journal of Biological Chemistry 284, 1949-1953 (2009).

45. Kaur D, et al. Initiation of methylglucose lipopolysaccharide biosynthesis in mycobacteria. PloS one 4, e5447 (2009).

46. Stadthagen G, et al. Genetic basis for the biosynthesis of methylglucose lipopolysaccharides in Mycobacterium tuberculosis. Journal of Biological Chemistry 282, 27270-27276 (2007).

47. Kalscheuer R, et al. Self-poisoning of Mycobacterium tuberculosis by targeting GlgE in an alpha-glucan pathway. Nature chemical biology 6, 376-384 (2010).

48. Kalscheuer R, Weinrick B, Veeraraghavan U, Besra GS, Jacobs WR. Trehalose- recycling ABC transporter LpqY-SugA-SugB-SugC is essential for virulence of Mycobacterium tuberculosis. Proceedings of the National Academy of Sciences 107, 21761-21766 (2010).

49. Ruhal R, Kataria R, Choudhury B. Trends in bacterial trehalose metabolism and significant nodes of metabolic pathway in the direction of trehalose accumulation. Microbial biotechnology 6, 493-502 (2013).

50. Otwinowski Z, Minor W. Processing of X-ray diffraction data collected in oscillation mode. Method Enzymol 276, 307-326 (1997).

51. Adams PD, et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta crystallographica Section D, Biological crystallography 66, 213-221 (2010).

52. McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Storoni LC, Read RJ. Phaser crystallographic software. Journal of applied crystallography 40, 658-674 (2007).

185

53. Afonine PV, et al. Towards automated crystallographic structure refinement with phenix.refine. Acta crystallographica Section D, Biological crystallography 68, 352- 367 (2012).

54. Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and development of Coot. Acta crystallographica Section D, Biological crystallography 66, 486-501 (2010).

55. Chen VB, et al. MolProbity: all-atom structure validation for macromolecular crystallography. Acta crystallographica Section D, Biological crystallography 66, 12- 21 (2010).

56. Moriarty NW, Grosse-Kunstleve RW, Adams PD. electronic Ligand Builder and Optimization Workbench (eLBOW): a tool for ligand coordinate and restraint generation. Acta crystallographica Section D, Biological crystallography 65, 1074- 1080 (2009).

57. Syson K, et al. Structure of Streptomyces maltosyltransferase GlgE, a homologue of a genetically validated anti-tuberculosis target. The Journal of biological chemistry 286, 38298-38310 (2011).

58. Miyata Y, et al. High-Throughput Screen for Escherichia coli Heat Shock Protein 70 (Hsp70/DnaK) ATPase Assay in Low Volume by Exploiting Energy Transfer. Journal of biomolecular screening 15, 1211-1219 (2010).

59. Zhang J-H, Chung TD, Oldenburg KR. A simple statistical parameter for use in evaluation and validation of high throughput screening assays. Journal of biomolecular screening 4, 67-73 (1999).

60. MacGregor EA, Janecek S, Svensson B. Relationship of sequence and structure to specificity in the alpha-amylase family of enzymes. Biochimica et biophysica acta 1546, 1-20 (2001).

61. Janecek S, Svensson B, Henrissat B. Domain evolution in the alpha-amylase family. Journal of molecular evolution 45, 322-331 (1997).

62. Juge N, et al. Comparison of barley malt alpha-amylase isozymes 1 and 2: construction of cDNA hybrids by in vivo recombination and their expression in yeast. Gene 130, 159-166 (1993).

186

63. Leiba J, et al. Mycobacterium tuberculosis maltosyltransferase GlgE, a genetically validated antituberculosis target, is negatively regulated by Ser/Thr phosphorylation. Journal of Biological Chemistry 288, 16546-16556 (2013).

64. Davies GJ, Wilson KS, Henrissat B. Nomenclature for sugar-binding subsites in glycosyl hydrolases. Biochem J 321, 557-559 (1997).

65. Veleti SK, Lindenberger JJ, Ronning DR, Sucheck SJ. Synthesis of a C-phosphonate mimic of maltose-1-phosphate and inhibition studies on Mycobacterium tuberculosis GlgE. Bioorg Med Chem 22, 1404-1411 (2014).

66. Veleti SK, Lindenberger JJ, Thanna S, Ronning DR, Sucheck SJ. Synthesis of a Poly- hydroxypyrolidine-Based inhibitor of Mycobacterium tuberculosis GlgE. J Org Chem, 20 (2014).

67. Schirmer T, Keller TA, Wang YF, Rosenbusch JP. Structural Basis for Sugar Translocation through Maltoporin Channels at 3.1-Angstrom Resolution. Science 267, 512-514 (1995).

68. Wimmerova M, Kozmon S, Necasova I, Mishra SK, Komarek J, Koca J. Stacking Interactions between Carbohydrate and Protein Quantified by Combination of Theoretical and Experimental Methods. PloS one 7, (2012).

69. Robert X, Gouet P. Deciphering key features in protein structures with the new ENDscript server. Nucleic acids research 42, W320-324 (2014).

70. Vasella A, Davies GJ, Bohm M. Glycosidase mechanisms. Curr Opin Chem Biol 6, 619-629 (2002).

71. Svensson B. Protein engineering in the α-amylase family: catalytic mechanism, substrate specificity, and stability. Plant molecular biology 25, 141-157 (1994).

72. Syson K, et al. Structural insight into how Streptomyces coelicolor maltosyl transferase GlgE binds alpha-maltose 1-phosphate and forms a maltosyl-enzyme intermediate. Biochemistry 53, 2494-2504 (2014).

187

73. Patil R, Das S, Stanley A, Yadav L, Sudhakar A, Varma AK. Optimized hydrophobic interactions and hydrogen bonding at the target-ligand interface leads the pathways of drug-designing. PloS one 5, e12029 (2010).

74. Sarkar A, Kellogg GE. Hydrophobicity--shake flasks, protein folding and drug discovery. Current topics in medicinal chemistry 10, 67-83 (2010).

75. Wells JA, McClendon CL. Reaching for high-hanging fruit in drug discovery at protein-protein interfaces. Nature 450, 1001-1009 (2007).

76. Lindenberger JJ, Veleti SK, Sucheck SJ, Ronning DR. Crystal structures of Mycobacterium tuberculosis GlgE and complexes with non-covalent inhibitors. Scientific reports, (2015).

77. Lillelund VH, Jensen HH, Liang XF, Bols M. Recent developments of transition-state analogue glycosidase inhibitors of non-natural product origin. Chem Rev 102, 515- 553 (2002).

78. Schramm VL. Transition States, analogues, and drug development. ACS chemical biology 8, 71-81 (2013).

79. Gloster TM, Davies GJ. Glycosidase inhibition: assessing mimicry of the transition state. Organic & biomolecular chemistry 8, 305-320 (2010).

80. Unrau PJ, Bartel DP. An oxocarbenium-ion intermediate of a ribozyme reaction indicated by kinetic isotope effects. Proceedings of the National Academy of Sciences of the United States of America 100, 15393-15397 (2003).

81. Provencher L, Steensma DH, Wong CH. Five-membered ring azasugars as potent inhibitors of alpha-L-rhamnosidase (naringinase) from Penicillium decumbens. Bioorg Med Chem 2, 1179-1188 (1994).

82. Borges de Melo E, da Silveira G, Carvalho I. a- and b-Glucosidase inhibitors: chemical structure and biological activity. Tetrahedron 62, 10277-10302 (2006).

83. Thanna S, Lindenberger JJ, Gaitonde VV, Ronning DR, Sucheck SJ. Synthesis of a 2- deoxy-2,2-difluoro-α-maltosyl fluoride and its X-ray structure in complex with Streptomyces coelicolor GlgEI-V279S. Organic & biomolecular chemistry, (2015).

188

84. Withers SG, Street IP, Bird P, Dolphin DH. 2-Deoxy-2-Fluoroglucosides - a Novel Class of Mechanism-Based Glucosidase Inhibitors. J Am Chem Soc 109, 7530-7531 (1987).

85. Withers SG, Rupitz K, Street IP. 2-Deoxy-2-Fluoro-D-Glycosyl Fluorides - a New Class of Specific Mechanism-Based Glycosidase Inhibitors. Journal of Biological Chemistry 263, 7929-7932 (1988).

86. Yamashita H, Nakatani H, Tonomura Bi. Effect of a p-nitro group of phenyl- maltooligosaccharide substrate on the change of action specificity of lysine-modified porcine pancreatic alpha-amylase. Biochemistry and molecular biology international 35, 79-85 (1995).

87. Kalscheuer R, Jacobs WR, Jr. The significance of GlgE as a new target for tuberculosis. Drug news & perspectives 23, 619-624 (2010).

88. Miura T, et al. Effect of Acarbose (. ALPHA.-Glucosidase Inhibitor) on Disaccharase Activity in Small Intestine in KK-Ay and ddY Mice. Journal of nutritional science and vitaminology 44, 371-379 (1998).

89. Elbein AD, Pan Y, Pastuszak I, Carroll D. New insights on trehalose: a multifunctional molecule. Glycobiology 13, 17R-27R (2003).

90. Clegg JS, Evans DR. The physiology of blood trehalose and its function during flight in the blowfly. (1961).

91. Kaur D, Guerin ME, Škovierová H, Brennan PJ, Jackson M. Biogenesis of the cell wall and other glycoconjugates of Mycobacterium tuberculosis. Advances in applied microbiology 69, 23-78 (2009).

92. Favrot L, Ronning DR. Targeting the mycobacterial envelope for tuberculosis drug development. Expert Rev Anti Infect Ther 10, 1023-1036 (2012).

93. Gavalda S, et al. The Polyketide Synthase Pks13 Catalyzes a Novel Mechanism of Lipid Transfer in Mycobacteria. Chemistry & biology 21, 1660-1669 (2014).

189

94. Tahlan K, et al. SQ109 targets MmpL3, a membrane transporter of trehalose monomycolate involved in mycolic acid donation to the cell wall core of Mycobacterium tuberculosis. Antimicrobial agents and chemotherapy 56, 1797-1809 (2012).

95. Murphy HN, et al. The OtsAB pathway is essential for trehalose biosynthesis in Mycobacterium tuberculosis. Journal of Biological Chemistry 280, 14524-14529 (2005).

96. Sassetti CM, Boyd DH, Rubin EJ. Genes required for mycobacterial growth defined by high density mutagenesis. Molecular microbiology 48, 77-84 (2003).

97. Farelli JD, et al. Structure of the trehalose-6-phosphate phosphatase from Brugia malayi reveals key design principles for anthelmintic drugs. PLoS pathogens 10, e1004245 (2014).

98. Klutts S, et al. Purification, cloning, expression, and properties of mycobacterial trehalose-phosphate phosphatase. Journal of Biological Chemistry 278, 2093-2100 (2003).

99. Jorge JA, Maria de Lourdes T, Thevelein JM, Terenzi HF. Trehalases and trehalose hydrolysis in fungi. FEMS microbiology letters 154, 165-171 (1997).

100. Romero-Pérez AI, Ibern-Gómez M, Lamuela-Raventós RM, de la Torre-Boronat MC. Piceid, the major resveratrol derivative in grape juices. Journal of Agricultural and Food Chemistry 47, 1533-1536 (1999).

101. Bosutti A, Degens H. The impact of resveratrol and hydrogen peroxide on muscle cell plasticity shows a dose-dependent interaction. Scientific reports 5, (2015).

102. Holtz KM, Stec B, Kantrowitz ER. A model of the transition state in the alkaline phosphatase reaction. Journal of Biological Chemistry 274, 8351-8354 (1999).

103. Xu Y-W, Moréra S, Janin J, Cherfils J. AlF3 mimics the transition state of protein phosphorylation in the crystal structure of nucleoside diphosphate kinase and MgADP. Proceedings of the National Academy of Sciences 94, 3579-3583 (1997).

190

104. Shimakata T, Minatogawa Y. Essential role of trehalose in the synthesis and subsequent metabolism of corynomycolic acid in Corynebacterium matruchotii. Archives of biochemistry and biophysics 380, 331-338 (2000).

105. Takayama K, Wang C, Besra GS. Pathway to synthesis and processing of mycolic acids in Mycobacterium tuberculosis. Clinical microbiology reviews 18, 81-101 (2005).

106. Davies DR, Hol WG. The power of vanadate in crystallographic investigations of phosphoryl transfer enzymes. FEBS letters 577, 315-321 (2004).

107. Kuznetsova E, et al. Functional Diversity of Haloacid Dehalogenase Superfamily Phosphatases from Saccharomyces cerevisiae: Biochemical, Structural, and Evolutionary Insights. Journal of Biological Chemistry, M115. 657916 (2015).

108. Newton GL, Buchmeier N, Fahey RC. Biosynthesis and functions of mycothiol, the unique protective thiol of Actinobacteria. Microbiology and Molecular Biology Reviews 72, 471-494 (2008).

109. Salinas AE, Wong MG. Glutathione S-transferases-a review. Current medicinal chemistry 6, 279-310 (1999).

110. Gaballa A, et al. Biosynthesis and functions of bacillithiol, a major low-molecular- weight thiol in Bacilli. Proceedings of the National Academy of Sciences 107, 6482- 6486 (2010).

111. Patel MP, Blanchard JS. Expression, purification, and characterization of Mycobacterium tuberculosis mycothione reductase. Biochemistry 38, 11827-11833 (1999).

112. Steffek M, Newton GL, Av-Gay Y, Fahey RC. Characterization of Mycobacterium tuberculosis mycothiol S-conjugate amidase. Biochemistry 42, 12067-12076 (2003).

113. Newton GL, Leung SS, Wakabayashi JI, Rawat M, Fahey RC. The DinB superfamily includes novel mycothiol, bacillithiol, and glutathione S-transferases. Biochemistry 50, 10751-10760 (2011).

191

114. Feng J, et al. The gene ncgl2918 encodes a novel maleylpyruvate isomerase that needs mycothiol as cofactor and links mycothiol biosynthesis and gentisate assimilation in Corynebacterium glutamicum. Journal of Biological Chemistry 281, 10778-10785 (2006).

115. Rawat M, Newton GL, Ko M, Martinez GJ, Fahey RC, Av-Gay Y. Mycothiol- deficient Mycobacterium smegmatis mutants are hypersensitive to alkylating agents, free radicals, and antibiotics. Antimicrobial agents and chemotherapy 46, 3348-3355 (2002).

116. Rawat M, Av-Gay Y. Mycothiol-dependent proteins in actinomycetes. FEMS microbiology reviews 31, 278-292 (2007).

117. Newton GL, Av-Gay Y, Fahey RC. A novel mycothiol-dependent detoxification pathway in mycobacteria involving mycothiol S-conjugate amidase. Biochemistry 39, 10739-10746 (2000).

118. Sareen D, Newton GL, Fahey RC, Buchmeier NA. Mycothiol is essential for growth of Mycobacterium tuberculosis Erdman. Journal of bacteriology 185, 6736-6740 (2003).

119. Newton GL, Buchmeier N, La Clair JJ, Fahey RC. Evaluation of NTF1836 as an inhibitor of the mycothiol biosynthetic enzyme MshC in growing and non-replicating Mycobacterium tuberculosis. Bioorgan Med Chem 19, 3956-3964 (2011).

120. Newton GL, et al. Characterization ofMycobacterium smegmatisMutants Defective in 1-d-myo-Inosityl-2-amino-2-deoxy-α-d-glucopyranoside and Mycothiol Biosynthesis. Biochemical and biophysical research communications 255, 239-244 (1999).

121. Cooper DR, Grelewska K, Kim C-Y, Joachimiak A, Derewenda ZS. The structure of DinB from Geobacillus stearothermophilus: a representative of a unique four-helix- bundle superfamily. Acta Crystallographica Section F: Structural Biology and Crystallization Communications 66, 219-224 (2010).

122. Bryce R, Hillier I, Naismith J. Carbohydrate-protein recognition: molecular dynamics simulations and free energy analysis of oligosaccharide binding to concanavalin A. Biophysical journal 81, 1373-1388 (2001).

192

123. Liu S, Zhang P, Ji X, Johnson W, Gilliland G, Armstrong R. Contribution of tyrosine 6 to the catalytic mechanism of isoenzyme 3-3 of glutathione S-transferase. Journal of Biological Chemistry 267, 4296-4299 (1992).

193