A Dissertation

entitled

Strategies for Membrane Protein Studies and Structural Characterization of a Metabolic

Enzyme for Antibiotic Development

Buenafe T. Arachea

Submitted to the Graduate Faculty as partial fulfillment of the

requirements for the Doctor of Philosophy Degree in Chemistry

Dr. Ronald E. Viola, Committee Chair

Dr. Max O. Funk, Committee Member

Dr. Donald Ronning, Committee Member

Dr. Marcia McInerney, Committee Member

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

The University of Toledo August 2011

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

Strategies for Membrane Protein Studies and Structural Characterization of a Metabolic Enzyme for Antibiotic Development

Buenafe T. Arachea

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

The University of Toledo August 2011

Membrane proteins are essential in a variety of cellular functions, making them viable targets for drug development. However, progress in the structural elucidation of membrane proteins has proven to be a difficult task, thus limiting the number of published structures of membrane proteins as compared with the enormous structural information obtained from soluble proteins. The challenge in membrane protein studies lies in the production of the required sample for characterization, as well as in developing methods to effectively solubilize and maintain a functional and stable form of the target protein during the course of crystallization.

To address these issues, two different approaches were explored for membrane protein studies. The first approach utilized different soluble domains as fusion partners with an alpha helical membrane protein (KcsA) to evaluate the effectiveness of this method in forming lattice contacts to produce crystals for high resolution studies. The fusion constructs were successfully cloned and expressed in C41 E. coli cells. The fusion with the maltose-binding protein (MBP) was purified and subjected to crystallization.

Conditions for crystal formation and growth were identified and the MBP-fusion was

iii further characterized by dynamic light scattering measurements and mass spectrometry.

The second approach focuses on the extraction efficiencies of different detergent types to solubilize recombinant and constitutive membrane proteins from bacterial membranes. Using 1D gel electrophoresis for separation and MALDI-TOF spectrometry coupled with database searching for protein identification, a total of 30 unique membrane proteins including the overexpressed model protein (KcsA) were detected from the bacterial membranes. The identified proteins were isolated from the total and outer bacterial membranes, and these proteins are involved in a variety of functions which include cell respiration, ion and molecular transport, as well as in membrane biogenesis and assembly. Our study provides an initial detergent screen set that could be expanded to facilitate the selection of detergents for optimal extraction of different target membrane proteins.

Aside from membrane proteins, other attractive targets for drug development include proteins that are involved in key metabolic pathways that are crucial in the survival of pathogenic microorganisms. The enzyme aspartate semialdehyde dehydrogenase (ASADH) is located at a significant branchpoint in the aspartate biosynthetic pathway, a pathway utilized by bacterial organism to produce four essential amino acids and metabolites that serve as precursor for various cell processes. Blocking this pathway in general, and inhibiting the ASADH enzyme in particular, is fatal to microorganisms, thus raising interest in structure determination of ASADH from various families to aid in the design and development of selective inhibitors of this enzyme.

The first structure of ASADH purified from a fungal species, the yeast Candida albicans, was crystallized in the presence of its nucleotide cofactor. The fungal enzyme is

iv a functional dimer with similar overall fold and domain organization to its bacterial counterparts. More detailed structural comparison between the fungal and bacterial

ASADHs revealed differences in secondary structural elements and in the nucleotide cofactor binding that may explain the lower catalytic activity observed for the fungal enzyme. Moreover, alterations in the dimer interface through the deletion of a helical subdomain and the replacement of amino acids involved in critical hydrogen bonding network results in the disruption of intersubunit communication channels required to support an alternating site catalytic mechanism. Elucidation of the structural details of this fungal enzyme allows an expanded assessment of ASADH as a possible target for antifungal drug development.

Selective inhibitors of this fungal ASADH were identified from kinetic screening of custom made fragment libraries (Gao et al., 2010). These inhibitors were cocrystallized with the enzyme as a binary complex or a ternary complex in the presence of the NADP cofactor or the dinucleotide analog (2’5’, ADP). Co-crystallization yielded good diffraction quality crystals and these inhibitor complexes were structurally characterized by x-ray crystallography. These inhibitor studies were also expanded by evaluating ASADH targets from families of antibiotic resistant strains to help establish selectivity between different ASADH forms.

This dissertation is dedicated to my parents, my greatest source of love, strength and inspiration. Thank you for always being there.

Acknowledgments

I express my appreciation to my supportive adviser and mentor, Dr. Ronald E.

Viola, for all the guidance during my stay in his lab. His commitment in research has motivated me to pursue my experiments despite the challenges attached with it. I appreciate his willingness to share his vast knowledge and expertise in many areas. He always makes sure that each student from his lab gets the best training not only in conducting research but also in other significant areas that would help us in our independent career as researchers. My stay in the Viola lab is one great experience that I would always value.

Thank you to the past and present members of the Viola Research Group, especially to Dr. Alexander Pavlovsky and Dr. Xuying Liu. You were the two persons in the lab that I closely worked with. Thank you for sharing your time and expertise to help me in carrying out my experiments.

I am also grateful to our collaborators Dr. Sami Saribas, Dr. Dragan Isailovic and

Zhen Sun for all of their help in my research. I had a good time working with all of you.

To my family who has supported my decision to pursue further studies abroad, I appreciate all the love and understanding.

Lastly, to Jonathan Johnson, thank you for the love, encouragement and patience.

You are one of my life’s sweetest blessings.

Table of Contents

Abstract……………………………………………………………………………... iii

Acknowledgments…………………………………………………………………... vi

Table of Contents………………………………………………………………….... vii

List of Tables……………………………………………………………………….. xiii

List of Figures………………………………………………………………………. xv

Chapter 1 Introduction…………………………………………………………… 1

1.1 Membrane Proteins: Structure and Function ………………………………... 1

1.2 Current Methods for the Structural Determination of Membrane Proteins….. 5

1.2.1 Detergents as Tools for Membrane Protein Studies………………... 5

1.2.2 Lipidic Phase Crystallization ……………………………………… 7

1.2.3 Antibody-mediated Crystallization………………………………… 10

1.2.4 Fusion Proteins for Hydrophilic Surface Expansion……………….. 12

1.3 Aspartate Biosynthetic Pathway……………………………………………... 14

1.4 Aspartate-β-Semialdehyde Dehydrogenase………………………………….. 16

1.5 Inhibitors of ASADH………………………………………………………… 21

Chapter 2 Membrane Fusion Proteins……………………………………………. 23

2.1 Introduction…………………………………………………………………... 23

2.2 Target Membrane Fusions………………………………………………….... 25

vii 2.3 Generation of Fusion Constructs……………………………………………... 27

2.3.1 Generation of att Sites……………………………………………… 27

2.3.2 Generation of Entry Clone…………………………………………. 29

2.3.3 Generation of Fusion Protein Expression Clones………………….. 29

2.4 Pilot Protein Expression of the Fusion Constructs…………………………… 32

2.5 Cell Growth…………………………………………………………………... 37

2.6 Membrane Fusion Protein Extraction………………………………………... 37

2.7 Purification of H6-MBP-KcsA Fusion………………………………………. 38

2.8 Crystallization of H6-MBP-Fusion…………………………………………... 42

2.9 Data Collection and Structure Determination………………………………... 48

2.10 Characterization of the H6-MBP-KcsA Fusion……………………………… 53

2.10.1 DNA Sequencing of the Expression Clone………………………. 53

2.10.2 Mass Spectrometric Analysis…………………………………….. 55

2.11 Summary and Future Work…………………………………………………... 59

Chapter 3 Detergent Screening for Membrane Protein Extraction……………..... 61

3.1 Introduction…………………………………………………………………... 61

3.2 Materials and Methods……………………………………………………….. 62

3.2.1 Reagents and Chemicals…………………………………………… 62

3.2.2 Plasmid and Cells…………………………………………………... 63

3.2.3 Cell Culture and Growth…………………………………………… 63

3.2.4 Cell Membrane Preparation………………………………………... 63

3.2.5 Extraction and Solubilization………………………………………. 64

3.2.6 Gel Electrophoresis………………………………………………… 65

viii 3.2.7 MS Sample Preparation and Protein Identification………………… 65

3.2.8 MALDI-MS Analysis……………………………………………… 66

3.3 Results ……………………………………………………………………….. 67

3.3.1 Total Protein Extraction from Different Membranes………………. 67

3.3.2 Protein Identification……………………………………………….. 70

3.3.3 Constitutive Inner Membrane Proteins…………………………….. 82

3.3.3.1 Bioenergetic Complexes…………………………………. 82

3.3.3.2 Inner Membrane Proteins with other Functions………….. 85

3.3.4 Constitutive Periplasmic Proteins………………………………….. 87

3.3.5 Constitutive Outer Membrane Proteins…………………………….. 88

3.3.5.1 Porins…………………………………………………….. 88

3.3.5.2 Transport OMPs………………………………………….. 93

3.3.5.3 OMPs Involved in Biogenesis……………………………. 94

3.4 Discussion……………………………………………………………………. 95

3.4.1 Detergent Profiles for Total Membrane Extraction………………... 97

3.4.2 Detergent Profiles for Inner and Periplasmic Membrane Extraction. 98

3.4.3 Detergent Profiles for Outer Membrane Extraction………………... 100

3.4.4 Possible Insights on Protein-Detergent Interactions……………….. 101

3.5 Summary……………………………………………………………………... 103

Chapter 4 Structure of the First Fungal Ortholog of Aspartate β-Semialdehyde

Dehydrogenase……………………………………………………………………… 104

4.1 Introduction…………………………………………………………………... 104

4.2 Cloning, Expression and Purification of caASADH………………………… 106

ix 4.2.1 Cloning and Expression……………………………………………. 106

4.2.2 Purification of Native caASADH...... 106

4.2.3 Purification of the Selenomethionine Form of caASADH………… 109

4.3 Crystallization of caASADH………………………………………………… 111

4.3.1 Crystallization of the Apoenzyme Form of caASADH……………. 111

4.3.2 Crystallization of the Selenomethionine Form of caASADH……... 114

4.3.3 Crystallization of the caASADH-Binary Complex with NADP…... 114

4.4 Structural Determination of caASADH-Cofactor Complex…………………. 115

4.4.1 Data Collection and Processing……………………………………. 115

4.4.2 Structure Determination, Refinement and Validation……………… 118

4.4.3 Structural Refinement……………………………………………… 118

4.5 Structure of the caASADH-Cofactor Complex……………………………… 120

4.5.1 Overall Structure…………………………………………………… 120

4.5.2 Secondary Structure Comparison…………………………………... 122

4.5.3 Active Site Comparison……………………………………………. 125

4.5.4 Differences in Cofactor Binding…………………………………… 128

4.5.5 Intersubunit Communication………………………………………. 131

4.5.6 Structure-Activity relationships……………………………………. 134

4.6 Conclusions…………………………………………………………………... 137

Chapter 5 Crystallographic Screening of ASADH Targets and Inhibitor

Complexes…………………………………………………………………………... 138

5.1 Introduction…………………………………………………………………... 138

5.2 ASADH from Staphylococcus aureus……………………………………….. 140

x 5.2.1 Target Identification………………………………………………... 140

5.2.2 Cloning of ASADH from S. aureus………………………………... 148

5.2.2.1 Gene Amplification………………………………………. 148

5.2.2.2 Generation of the attB1-saASADH-attB2 PCR Products.. 150

5.2.2.3 Generation of the Entry and Expression Clones 151

5.2.3 Protein Expression…………………………………………………. 155

5.2.4 Stability Studies of saASADH……………………………………... 158

5.2.5 Purification of saASADH………………………………………….. 162

5.2.6 Screening of Crystallization Conditions for saASADH…………… 165

5.2.6.1 Sparse-matrix Screening…………………………………. 165

5.2.6.2 Systematic Evaluation of Conditions…………………….. 167

5.2.6.3 Incorporation of Glycerol in the Crystallization Trials…... 168

5.3 Enzyme-Inhibitor Complexes of Fungal ASADH (caASADH)……………... 170

5.3.1 Identified Inhibitors from Fragment Libraries……………………... 170

5.3.2 Crystallization of Enzyme-Inhibitor Complexes…………………... 172

5.3.2.1 Co-crystallization Approach……………………………... 172

5.3.2.2 Soaking Approach………………………………………... 175

5.3.3 Data Collection and Structure Refinement………………………… 176

5.3.4 Assessment of Structural Data and Ligand Binding……………….. 176

5.3.5 Comparison with Inhibitor Binding Studies in other ASADH Forms 179

5.4 Summary and Future Work…………………………………………………... 181

References…………………………………………………………………………... 182

A Mass Spectrometric Analysis and Database Searching………………………. 193

xi A1 Identification of KcsA………………………………………………… 193

A2 Identification of Constitutive Proteins from Total Membrane………... 195

A3 Identification of Constitutive Proteins from Outer Membrane……….. 197

xii

List of Tables

Table 1.1 Detergents commonly used for membrane protein studies…………... 6

Table 1.2 Summary of known membrane protein structures solved using lipidic phase crystallization methods……………………………………………...... 9

Table 1.3 Summary of ASADH structures…………………………………….. 20

Table 1.4 ASADH inhibitors identified from fragment based library screening. 22

Table 2.1 Expected Mr values of the fusion protein…………………………… 36

Table 2.2 Crystallization conditions from H6-MBP-KcsA sparse matrix screening………………………………………………………………………….. 43

Table 2.3 Data collection and refinement statistics for H6-MBP-KcsA fusion monomer…………………………………………………………………... 50

Table 3.1 Membrane protein extraction efficient for different surfactants…...... 69

Table 3.2 Peptide mass fingerprint analysis for KcsA…………………………. 74

Table 3.3 Summary of identified membrane proteins from total membrane fractions…………………………………………………………………………… 77

Table 3.4 Summary of cytoplasmic proteins detected from total and outer membrane fractions……………………………………………………………….. 78

Table 3.5 Summary of identified membrane proteins from outer membrane fractions…………………………………………………………………………… 80

xiii Table 3.6 Summary of constitutive membrane proteins extracted by test detergents from total and outer membranes………………………………………. 81

Table 4.1 Specific activities of the native form of fungal ASADH from

Candida albicans (caASADH)…………………………………………………… 108

Table 4.2 Specific activities of the selenomethionine form (SeMet) of fungal

ASADH from Candida albicans (caASADH)…………………………………… 110

Table 4.3 Data collection statistics for the caASADH-NADP complex………. 117

Table 4.4 Refinement statistics for the caASADH-NADP complex…………... 119

Table 4.5 ASADH structure and activity relationship…………………………. 135

Table 5.1 Mutations in the ASADH sequence of S. aureus variants…………... 145

Table 5.2 Primer designs used for the amplification and cloning of saASADH. 150

Table 5.3 Stability studies of saASADH………………………………………. 160

Table 5.4 Effect of glycerol levels on saASADH enzyme activity……………. 161

Table 5.5 Specific activities of the ASADH from S.aureus (saASADH)……... 163

Table 5.6 Inhibitors of fungal caASADH (Candida albicans) identified from fragment library screening………………………………………………………... 171

Table 5.7 Diffraction data and refinement statistics of inhibitor complexes of caASADH……………………………………………………………………….... 178

xiv

List of Figures

Figure 1.1 Representative structures showing the two types of integral membrane proteins………………………………………………………………... 4

Figure 1.2 Antibody-mediated crystallization of membrane proteins…………. 11

Figure 1.3 Aspartate biosynthetic pathway…………………………………….. 15

Figure 1.4 ASADH-catalyzed reaction………………………………………… 16

Figure 1.5 Proposed catalytic mechanism of ASADH enzymes………………. 17

Figure 1.6 Overall structure of ASADH from E. coli………………………….. 19

Figure 2.1 Generation of fusion constructs…………………………………….. 28

Figure 2.2 Analysis of the entry clone…………………………………………. 30

Figure 2.3 Generation of fusion protein expression clones……………………. 31

Figure 2.4 Protein expression analysis of various fusion constructs…………... 34

Figure 2.5 Purification of H6-MBP-KcsA fusion……………………………… 40

Figure 2.6 Dynamic light scattering measurements of H6-MBP-KcsA fusion... 41

Figure 2.7 Tetrameric H6-MBP-KcsA fusion crystals………………………… 43

Figure 2.8 Expansion of initial crystallization conditions for H6-MBP-KcsA… 45

Figure 2.9 Crystals of the tetrameric and monomeric H6-MBP-KcsA fusion…. 47

Figure 2.10 Structure determination of putative H6-MBP-KcsA monomer fusion……………………………………………………………………………... 51

xv Figure 2.11 Sequence analysis of the fusion expression construct……………… 54

Figure 2.12 Mass spectra of H6-MBP-KcsA tetramer fusion…………………… 56

Figure 2.13 Mass spectra of the putative H6-MBP-KcsA monomer fusion…….. 58

Figure 3.1 Optimization of protein loading concentrations for protein band isolation. ………………………………………………………………………….. 71

Figure 3.2 Constitutive membrane proteins extracted by test detergents from the total cell membrane of E. coli……………………...... 75

Figure 3.3 Constitutive membrane proteins extracted by test detergents from the outer membrane of E. coli……...... 89

Figure 4.1 Characterization of native caASADH……………………………… 108

Figure 4.2 Characterization of the selenomethionine form of caASADH……... 110

Figure 4.3 Crystals of the apo enzyme form of native caASADH…………….. 113

Figure 4.4 Crystals of caASADH in complex with its nucleotide cofactor,

NADP……………………………………………………………………………... 116

Figure 4.5 Overall structure of caASADH in complex with NADP…………… 121

Figure 4.6 Topological map showing the secondary structure of caASADH….. 122

Figure 4.7 Sequence alignment of ASADH enzymes from different organisms. 124

Figure 4.8 Active site comparison of ASADH………………………………… 127

Figure 4.9 NADP binding conformations…………………………………….. 130

Figure 4.10 Dimerization interface of fungal ASADH versus bacterial ASADH. 133

Figure 5.1 Amino acid sequence alignment of various strains of

Staphylococcus aureus……………………………………………………………. 143

xvi Figure 5.2 Agarose gel electrophoresis profiles of the cloning of saASADH…. 149

Figure 5.3 DNA sequencing analysis of the saASADH entry clone…………... 153

Figure 5.4 Confirmation of the expression clones……………………………... 154

Figure 5.5 Protein expression of saASADH…………………………………… 157

Figure 5.6 Purification of saASADH………………………...... 164

Figure 5.7 Sparse matrix screening of saASADH……………...... 166

Figure 5.8 Protein-inhibitor complexes of fungal caASADH from co- crystallization experiments……………………………...... 174

Figure 5.9 Apo crystals of caASADH used for inhibitor soaking studies……... 175

Figure 5.10 Example of bound inhibitor in the active site of spASADH……….. 180

xvii

Chapter 1

Introduction

1.1 Membrane Proteins: Structure and Function

Membrane proteins (MPs) are biomolecules that are associated with the lipid bilayer of biological membranes. They possess a hydrophilic surface that interacts with the solvent in the outside environment as well as with polar head groups of lipid molecules present in the membranes (Iwata, 2003). Membrane proteins also contain a hydrophobic core buried within the lipid bilayer and this core is in contact with the alkyl chain of the lipid moieties. Depending on the degree of association with the membrane, membrane proteins are generally classified either as an integral or peripheral membrane protein. Integral membrane proteins penetrate the hydrophobic regions of the bilayer and form strong interactions. They can extend through the lipid bilayer or they can be attached only on one side of the membrane by an alpha helix or a covalently attached lipid.

Solubilization of an integral membrane protein requires the use of detergents that disrupt hydrophobic interactions within the phospholipid bilayer. Detergents are effective molecules in replacing the lipids that aid in the solubilization of the protein hydrophobic

1 surfaces. On the other hand, peripheral membrane proteins associate with the membrane through electrostatic contacts and hydrogen bonding with the hydrophilic domains of integral membrane proteins as well as the polar head groups of the lipids. These proteins can be released from the membrane by using relatively mild conditions (Sadowski et al.,

2008; von Heijne, 1996).

Two basic architectures are observed in integral membrane proteins, the alpha helix bundle and the beta barrel. Both structural classes manifest the necessity to form ordered hydrogen-bonding secondary structures in the non-aqueous environment of the lipid bilayer (von Heijne, 2007; Tamm et al., 2001). Beta barrel type of membrane proteins are usually found in the outer membrane of Gram-negative bacteria and also in organelles like mitochondria and chloroplasts. Reported sizes of beta barrel proteins range between 8 to 22 beta strands, forming monomeric as well as higher oligomeric structures (Schulz, 2000). On the other hand, alpha helical proteins are found in all cellular membranes with sizes ranging from a single helix to 20 or more transmembrane helices tightly packed in the membrane (von Heijne, 2007).

A consensus structure for a transmembrane helix made up of 20-30 amino acids was proposed by Wallin et al. (1997). The study showed that transmembrane helices are generally composed of a central helical region rich in aliphatic and phenylalanine residues and an aromatic belt composed of tryptophan and tyrosine amino acids which form favorable interactions with the membrane lipids. On the other hand, the ends of the helices are filled with charged residues such as lysines and arginines that mediate contacts with the polar environment, whereas helix-breaking proline residues are located outside of the helix and in areas of structural irregularities (Wallin et al., 1997).

2 The main structural features of known beta barrel proteins were summarized in a review article by Schulz (2000). Briefly, beta barrel integral membrane proteins have an even number of antiparallel strands, and the neighboring strands are linked together by either short turns or longer loops. The outer surface which interacts with the membrane lipid contains aliphatic residues with an aromatic band near the bilayer surfaces. These aromatic residues present in the strands have been proposed to participate in anchoring the protein to the membrane. Residues present in the inner core were assessed to have intermediate polarities, though there were cases wherein the protein interior is filled with polar residues that form hydrogen bond network such as in the structures seen from small beta barrel proteins like OmpA, OmpX and ompLA (Schulz, 2000; von Heijne, 2007)

The biological functions of membrane proteins are diverse and include respiration, photosynthesis, active ion transport, passive nutrient intake, membrane-bound enzymes, signal tranducers, multidrug efflux, protein export and membrane anchors. This multitude of functions makes membrane proteins attractive targets for therapeutic drugs, thus the need for extensive structural studies of these types of proteins (Schulz, 2000; von

Heijne 2007).

A B

Figure 1.1 Representative structures showing the two types of integral membrane proteins. (A) Vitamin B12 BtuCD with an α-helical transmembrane region ( PDB Code 1L7V) and (B) a β-barrel transmembrane region of ferric citrate uptake receptor protein, FecA (PDB Code 1KMO). Figures adapted from Walian et al., (2004).

4 1.2 Current Methods for the Structural Determination of Membrane

Proteins

The number of solved membrane protein structures is relatively small as compared to all the soluble protein structures available in the PDB database. There are currently 287 unique membrane protein structures which include protein targets of the same type from different species (Membrane Protein of Known 3D Structures, http://blanco.biomol.uci.edu/mpstruc/listAll/list; accessed on May 18, 2011). This slow progress in membrane protein structures reflects the difficulties encountered in performing structural studies of these types of proteins. Despite these challenges, several methodologies have emerged to crystallize these difficult proteins.

1.2.1 Detergents as Tools for Membrane Protein Studies

The use of detergents has been widely applied to membrane proteins. Detergents are amphiphilic molecules with a hydrophilic head group and a hydrophobic tail that closely mimic the natural environment of membrane proteins. These two functional groups enable effective solubilization of membrane proteins in aqueous solutions. For example, the hydrophilic portion which could be ionic, nonionic or zwitterionic provides interactions for solubility in water. On the other hand, the hydrophobic portion allows partitioning of the detergent into the lipid bilayer which eventually leads to bilayer dissolution and protein solubilization. These hydrophobic portions surround the membrane proteins to prevent it from aggregating. With these properties, detergents are quite effective in maintaining the functional integrity and stability of membrane proteins away from the membrane. Thus, these molecules are invaluable tools in membrane

5 protein extraction, solubilization, purification and crystallization experiments.

Selecting the right detergent for membrane protein studies is crucial and sometimes individual detergent screening is necessary in each step from membrane protein extraction to crystallization studies. Crystallization of MPs in the presence of detergents is usually carried out using the vapor-diffusion method. Several references have summarized the primary detergents utilized for high resolution structural studies of

MPs (Iwata, 2003; Prive, 2007; Newstead et al., 2008; le Maire et al., 2000; Raman et al.,

2006). Among all the variety of available detergents, the non-ionic types (alkyl sugar detergents) have been most successfully employed in solubilization and crystallization studies of MPs (Table 1.1) although other detergent types have also been used. Though these non-ionic detergents top the list, it is still necessary to test various kinds of detergents in carrying out membrane protein studies since each membrane protein have specific solubility and functional properties.

Table 1.1 Detergents commonly used for membrane protein studies.

Detergent CMC Membrane Protein (mM) Database Count N-octyl-β-D-glucopyranoside (OG) 18 158 N-dodecyl-β-D-maltopyranoside (DDM) 0.17 149 N,N-dimethyldodecylamine-N-oxide (LDAO) 1-2 126 N-decyl-β-D-maltopyranoside (DM) 1.6 79 Octyltetraoxyethylene (C8E4) 8 66 Dodecyl phosphocholine (DPC) 1.1 44 Dodecyloctaoxyethylene (C12E8) 0.09 34 Sodium dodecyl sulfate (SDS) 10 28 N-nonyl-β-D-glucopyranoside (NG) 6.5 28 Triton X-100 0.23 27 CHAPSO 8 23 Zwittergent 3-12 2.8 21

6 1.2.2 Lipidic Phase Crystallization

Aside from the use of detergents, an attractive and emerging technique for solving structures of membrane proteins is the lipidic phase crystallization. This approach includes the lipidic cubic phase (Landau & Rosenbusch, 1996), lipidic sponge phase

(Wadsten et al., 2006) and bicelle crystallization methods (Faham & Bowie, 2002). All protocols take into account the ability of solubilized MPs to produce well ordered crystals by reconstituting them into their natural lipid environment prior to crystallization, which may contribute to enhanced protein stability resulting in efficient crystal formation.

Generally, lipidic cubic phases (LPC) are prepared by first combining the purified protein solution with lipids such as monoolein (MO) in a 2:3 volume ratio to spontaneously form the cubic phase. Precipitant and additives are then added into the protein-containing LCP and crystals are allowed to grow either in batch crystallization or vapor-diffusion set-ups (Johansson et al., 2009; Caffrey, 2000). This membrane system has been described to form bicontinuous lipid bilayer infused with aqueous channels which facilitates protein diffusion, thus leading to nucleation and crystal growth (Landau

& Rosenbusch, 1996).

A newer version of lipidic phase crystallization is the lipidic sponge phase (LSP), which is a liquid analogue of the cubic phase. Sponge phase formation can be initiated by the addition of precipitants such as polyethylene glycols (PEGs), 2-methyl-2,4- pentanediol (MPD), dimethylsulfoxide (DMSO) and jeffamine to the cubic phase, typically in a volume ratio of 1:4. Compared with the cubic phase, the liquid form of the sponge phase allows easier set-up of vapor diffusion based crystallization screens, as well as the use of automated nanodrop crystallization robots and standard optimization

7 protocols to be directly applied to the crystallization experiments. (Johansson et al., 2009;

Wadsten et al., 2006).

The applicability of this approach in crystallizing membrane proteins was recently explored by developing a 48-condition LSP screen that was tested against 11 different proteins. Each condition in the LSP screen consists of an aqueous buffer, lipid MO, precipitants (PEG 400, PEG 1500, PEG 4000 and jeffamine M600) in the presence of salts (Li2SO4, (NH4)2SO4, MgCl2, NaCl) with a pH range between 5.5 to 9.0. Crystal hits were obtained from 8 membrane proteins from bacterial and eukaryotic origin that could be further optimized by standard crystallization protocols (Wohri et al., 2008; Johansson et al., 2009). This method of lipidic sponge phase can be further exploited as a tool to obtain novel structures of membrane proteins.

Another variation of lipidic phase crystallization is the bicelle crystallization method (BC) conceived by Faham & Bowie (2002). Bicelle crystallization proceeds by combining purified protein with lipid/amphiphile mixture on ice to keep the mixture in liquid form. The protein/bicelle mixture is then mixed with precipitant solution to set-up drops either in a silanized cover slip or on a sitting drop sample well and crystals are allowed to grow at room temperature, or at 37°C if the gel-like phase is desired.

Optimization is carried out in a similar fashion with standard crystallization procedures.

This method offers the advantages of providing a bilayer environment to the protein of interest, use of conventional crystallization set-up, and easier sample manipulation and crystal handling at room temperature where the medium is fluid (Faham & Bowie ,2002;

Faham et al., 2005).

8 The techniques described above are responsible for the structure determination of various membrane proteins such as bacterial rhodopsins, GPCRs and beta-barrel type proteins (Table 1.2). Good diffraction quality Type 1 crystals were obtained from these methods that were reported to exhibit protein surface contacts for crystal lattice stabilization (Johansson et al., 2009).

Table 1.2 Summary of known membrane protein structures solved using lipidic phase crystallization methods.a Protein Resolution Source PDB Method (Å) Organism Entry Used

Bacteriorhodopsin 2.5 H. salinarum 1AP9 LCP Halorhodopsin 1.8 H. salinarum 1E12 LCP Sensory rhodopsin II 2.1 N. pharaonis 1H68 LCP Sensory rhodopsin II- 1.94 N. pharaonis 1H2S LCP transducer complex Reaction center 2.35 R. sphaeroides 1OGV LCPb Anabaena sensory 2.0 Anabaena 1XIO LCP rhodopsin b Engineered human β2 – 2.4 Homo sapiens 2RH1 LCP adrenergic receptor OpcA outer membrane 1.95 N. meningitides 2VDF LCPb adhesin b Human A2A adenosine 2.6 Homo sapiens 3EML LCP receptor Reaction center 2.2 R. sphaeroides 2GNU LSP Light harvesting complex II 2.45 Rps. acidophila 2FKW LSP BtuB 1.95 E.coli 2GUF LSP Reaction center 1.95 Bl. viridis 2WJM LSP Bacteriorhodopsin 2.0 H. salinarum 1KME Bicelle β2-Adrenergic G-protein- 3.4/3.7 Homo sapiens 2R4R Bicelle coupled receptor Voltage-dependent anion 2.3 Mus musculus 3EMN Bicelle channel Xantorhodopsin 1.9 S. ruber 3DDL Bicelle a Adapted from Johansson et al., 2009 b Originally reported as LCP crystallization but likely to have proceeded via LSP

9 1.2.3 Antibody-mediated Crystallization

One celebrated method for membrane protein crystallization is the successful generation of high resolution structures in complex with antibody fragments. Various structures (Figure 1.2) have been solved using this approach. The binding of Fv fragments or Fab domains increases the polar surface of membrane proteins, thereby providing additional hydrophilic area for surface contacts leading to improved crystal quality. This complex also provides adequate space for the incorporation of detergent micelles. Crystal contacts are mainly contributed by the antibody fragments, with limited contacts between solvent accessible surfaces of target proteins. As such, this technique could be particulary be helpful with membrane proteins having transmembrane helices with very short surface-exposed loops (Hunte & Michel, 2002).

Earlier examples of antibody co-crystal structures include cytochrome c oxidase

(COX, PDB Code 1QLE) from Paracoccus denitrificans (Ostermeier et al.,1995;

Ostermeier et al., 1997); yeast cytochrome bc1 complex (QCR, PDB Code 1EZV) (Lange

& Hunte, 2002) and potassium channel protein KcsA (PDB Code 1K4C) from

Streptomyces lividans (Zhou et al., 2001). All of these structures diffracted between 2.0 to

2.5 Å resolutions, with diffraction quality crystals obtained from conventional vapor diffusion methods. The latest addition to this set of structures is the crystal structure of the human β2 adrenergic GPCR co-crystallized with Fab antibody (Rasmussen et al.,

2007). The complex yielded crystals grown in lipid environment by the bicelle crystallization method with the structure determined at a lower resolution of 3.4/3.7 Å.

10 The antibody-mediated crystallization is an elegant approach for solving membrane protein structures; however, its use is limited by the amount of additional work needed to isolate antibodies as well as the cost associated with it (Hunte & Michel,

2002). Thus, developing new alternative methods that could be generally applied to crystallize membrane protein is of prime interest.

B C

Figure 1.2 Antibody-mediated crystallization of membrane proteins. Asymmetric units of (A) cytochrome c oxidase, COX and (B) KcsA with antibody fragments. (C) Main contacts occur between antibody fragments in the biological unit of KcsA. Figures A and B were reproduced using by Pymol. Figure C adapted from Protein Databank.

11 1.2.4 Fusion Proteins for Hydrophilic Surface Expansion

An alternative to the antibody approach for membrane protein crystallization is the addition of fusion proteins to increase available hydrophilic surface. Fusion proteins, sometimes called affinity tags, are utilized to facilitate purification using affinity columns as well as to promote high level expression of target proteins. Popular fusion tags include polyhistidine tag, maltose binding protein, glutathione-S-transferase, green fluorescent protein (GFP) (Terpe, 2003).

These fusion proteins have the potential to facilitate crystallization by acting as a scaffold for the crystal lattice. These constructs could also be utilized to increase the solvent-exposed surface of proteins, thus imitating the role of the antibody fragments.

Once good diffraction quality crystals are obtained from the fusion constructs, structure determination of the target protein can be achieved by using the native structures of the fusion protein partner as search models in molecular replacement calculations (Hunte &

Michel, 2002).

The fusion of different soluble partners (MBP, GST, thioredoxin) has enabled acquisition of structural information for soluble proteins, with some examples described in Section 2.2. Such fusions were also applied for the crystallization of membrane-bound proteins. For example, cytochrome bo3 ubquinol oxidase from E. coli was produced as a

C-terminal fusion with protein Z subunit, however elucidation of the structure was not successful due to low resolution data at 6 Å (Byrne et al., 2000). A similar approach was used for the crystallization of lactose permease as a fusion with soluble cytochrome b562 , where the soluble partner was linked between transmembrane helices instead of the protein terminus. Though the transporter was functional using the fusion system, no

12 diffraction quality crystals have been reported so far (Prive & Kaback, 1996).

A successful example using this strategy of increasing hydrophilic surface through fusion proteins is the high resolution structure of human β2- adrenergic receptor fused with T4 lysozyme solved at 2.4 Å resolution. The soluble protein was inserted at the third intracellular loop of the receptor, which yielded better diffracting crystals as compared to the Fab-complex form of the same protein. Crystals were obtained using lipid cubic phase crystallization instead of the conventional vapor diffusion set-ups. The combination of the fusion strategy with the in cubo method proved to be beneficial in obtaining diffraction quality crystals of the receptor (Cherezov et al., 2007; Rosenbaum et al.,

2007; Bill et al., 2011).

Several methodologies have emerged in an attempt to advance structural studies of membrane proteins. The strategies outlined above are either used exclusively or in combination with other existing techniques to produce membrane protein crystals for high resolution studies. These diverse approaches as well as the development of new methodologies will provide venues to overcome the challenges in the crystallization of membrane proteins.

The first part of this dissertation focuses on developing strategies for membrane protein studies. We have explored the feasibility of the soluble protein fusion method in the crystallization of an alpha helical type membrane protein and our results are presented in Chapter 1. Chapter 2 describes the extraction efficiencies of different detergent types to solubilize recombinant and constitutive membrane proteins from bacterial membranes.

13 1.3 Aspartate Biosynthetic Pathway

The aspartate biosynthesis pathway utilizes L-aspartic acid for the production of four essential amino acids: lysine, methionine, isoleucine and threonine that are required for protein synthesis (Viola, 2001). In addition to these four amino acids, this pathway is also the source of several metabolites that play crucial roles in various functions of the cell (Figure 1.3). Important metabolite products include dihydrodipicolinate, a precursor of dipicolinate used by gram-positive bacteria for sporulation (Ragkousi et al.,2003) and diaminopimelate a required component of peptidoglycan polymers involved in bacterial cell wall synthesis (van Heijenoort, 2001). An additional product of this pathway is S- adenosyl-methionine which plays key roles in methylation reactions and in the production of quorum sensing signaling molecules that are critical for virulence factors in infectious organisms (Lyon & Novick, 2004).

The commitment step of this pathway is the phosphorylation of aspartic acid to β- aspartyl phosphate catalyzed by multiple aspartokinases (AK). Next in the pathway is the enzyme aspartate-semialdehyde dehydrogenase (ASADH) which is located at a critical branch point. From this junction, the pathway branches out to produce lysine from the precursor diaminopimelate, while the second route continues to synthesize other products of this pathway in the presence of two other key enzymes homoserine dehydrogenase

(HSD) and homoserine kinase (Cohen, 1985). These four enzymes complete the core set of catalytic proteins involved in the amino acid and metabolite production in the aspartate pathway (Figure 1.3).

14 This pathway is present in bacteria, fungi and higher plants, but is absent in mammals. Deletions of the asd gene encoding for ASADH were reported to be lethal to the microorganism and gene deletion in several bacterial strains such as Legionella pneumophila (Harb & Kwaik, 1998), Salmonella typhimurium (Galan et al., 1990) and

Streptococcus mutans (Cardineau & Curtiss, 1987) have caused loss of viability.

Blocking the aspartate pathway is fatal to microorganisms, thus raising interest in the design and development of selective inhibitors for antimicrobial therapy.

Figure 1.3 Aspartate biosynthetic pathway.

15 1.4 Aspartate-β-Semialdehyde Dehydrogenase

Aspartate-β-semialdehyde dehydrogenase (ASADH) catalyzes the reductive dephosphorylation of β-aspartyl phosphate (BAP) to aspartate-β-semialdehyde (ASA) in the presence of NADPH. Identification of the functionally important residues of this enzyme was initially carried out by sequence comparisons among ASADH family against representative sequences of a related enzyme, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Ouyang & Viola, 1995). Subsequent mutational and kinetic studies in combination with structural data acquired from various ASADH organisms have precisely established the roles of each active site residue in substrate binding and recognition, thus leading to the elucidation of the mechanism behind catalysis (Blanco et al., 2003, Blanco et al., 2004; Viola, 2001; Viola et al., 2011).

Figure 1.4 ASADH-catalyzed reaction.

The proposed kinetic mechanism of ASADH catalysis (Blanco et al., 2003), supported by crystal structures of the H. influenza enzyme involves the following steps:

An active site acid/base catalyst (His273) facilitates deprotonation of a nearby cysteine residue (Cys136), which leads to a nucleophilic attack on the carbonyl carbon of the substrate ASA forming a tetrahedral intermediate (Figure 1.5A). Transfer of a hydride ion

16 from the tetrahedral intermediate to the NADP coenzyme then produces an acyl-enzyme intermediate (Figure 1.5B), in which the carbonyl carbon is subsequently attacked by an enzyme-bound phosphate. The collapse of this intermediate (Figure 1.5C) releases the active site thiolate group and generates the product β-aspartyl phosphate (BAP). This fully reversible reaction proceeds in the reverse direction under physiological conditions to generate ASA from BAP.

Figure 1.5 Proposed catalytic mechanism of ASADH enzymes.

17 Aside from the acid/base catalyst (His273) and the catalytic nucleophile

(Cys136), other active site residues include Arg270 and Glu243 which stabilize the ASA substrate carboxyl and amino groups, respectively. The positively charged residues

Lys246 and Arg103 are also present in the active site, and they function in proper positioning of the bound phosphate for catalysis.

The crystal structure of ASADH obtained from the gram-negative bacteria E. coli

(Figure 1.6) provided the first insights into the overall structure of this enzyme (Hadfield et al., 1999; Hadfield et al., 2001). Structure determination of this same enzyme from various organisms (Table 1.3) further gave structural details that support the kinetic mechanism and emphasized salient features of this enzyme between different families.

The overall structure of ASADH is similar among all the solved structures. The functional unit is a dimer, with each subunit composed of an N-terminal coenzyme binding and a C-terminal dimerization and catalytic domain. However, a number of insertions and deletions differentiate the ASADH enzymes. These structural variations include the presence or absence of a helical domain in the dimerization interface, different residues involved in NADP binding and orientation, and the identity of linking residues in needed for intercommunication between subunits.

Structural features of the ASADH enzyme in comparison with three different enzyme families will be discussed in detail in Chapter 4, where we present the first structure of ASADH from a fungal source, the latest addition to the set of ASADH structures. This chapter also highlights detailed functional information on the secondary and overall fold, the constellation of active site residues, differences in both cofactor binding and inter-subunit communication. These key features could likely explain the

18 distinctive catalytic efficiency of each enzyme family despite a highly conserved catalytic site.

Figure 1.6 Overall structure of ASADH from E. coli. The asymmetric unit is shown with each monomer highlighted in different colors. The nucleotide cofactor, NADP is shown as a stick model. Figure generated with Pymol from coordinates deposited in Protein Databank (PDB Code: 1GL3)

Table 1.3 Summary of ASADH structures. Reported crystal structures of ASADH complexes obtained from various organisms. Organism Form / Ligand Resolution PDB Reference (Å) Code

E. coli Apo 2.50 1BRM Hadfield et al., 1999 NADP + SMCS 2.60 1GL3 Hadfield et al., 2001

V. cholerae Apo 2.77 1MC4 Blanco et al., 2003 NADP + SMCS 1.84 1MB4

H. influenza Apo 2.04 1NWC Blanco et al., 2003 Arsenate 2.28 1TA4 Faehnle et al., 2004 Periodate 2.15 1TB4 C136S, ASA + Pi 2.06 1PQP Blanco et al., 2004 H277N, NADP + SMCS 1.92 1PQU R103K 2.15 1PR3 R270K 2.40 1PS8 K246R 2.06 1PU2 E243D 2.05 1Q2X R103L 2.06 1QZA

M. jannaschii Se-Met 2.30 1YS4 Faehnle et al., 2005

S. pneumoniae Apo 2.10 2GYY Faehnle et al., 2006 NADP 1.80 2GZ1 2’,5’ADP 2.10 2GZ2 NADP + ASA 2.10 2GZ3

V. cholerae II apo 2.20 2QZ9 Viola et al., 2008 ASA 2.03 2R00

C. albicans NADP 2.20 3HSK Arachea et al., 2010

20 1.5 Inhibitors of ASADH

The requirement of amino acids in protein biosynthesis, as well as the absence of the aspartate pathway in mammalian cells, makes ASADH a viable candidate for antibiotic development. Several inhibitors which include both synthetic and fragment- based inhibitors have been reported for this enzyme.

Synthetic inhibitors that mimic the natural substrate β-aspartyl phosphate were developed by the group of Cox et al. (2002, 2005). Inhibitors with calculated inhibition constants that were reported include difluoromethylene phosphonates (Ki Pi = 95 μM), methylene phosphonate (Ki Pi = 2.13 mM; Ki ASA = 0.75 mM ), and phosphoramidate (Ki

Pi = 214 μM; Ki ASA = 92 μM).

Selective inhibitors against three representative ASADH families (gram-positive, gram-negative and fungi) were identified from custom-made fragment library screens

(Gao et al., 2010). Different binding selectivities were seen among the tested forms despite the highly conserved active site residues (Table 1.4). Their findings showed that bacterial ASADHs are inhibited by different amino acid analogues, whereas halo acids and substituted aromatic acids are more potent towards the fungal form. The benzophenone analogues on the other hand selectively inhibit the gram-negative form of the enzyme (Gao et al., 2010).

Chapter 5 will present our studies on structural characterization of fungal

ASADH-inhibitor complexes, with inhibitors derived from the fragment library screen.

This chapter will also highlight our efforts on crystallization of the ASADH enzyme isolated from a methicillin-resistant S. aureus that is also currently being analyzed for inhibitors to help establish selectivity among the different ASADH forms.

Table 1.4 ASADH inhibitors identified from fragment based library screening. (from Gao et al., 2010)

a Inhibitors Ki values (mM) Selectivity S. pneumoniae V. cholerae C. albicans sp:vc:ca L-cystine 0.036 ± 0.004 0.00023 ± 0.00006 >20b 160:1:>87000 L-cysteine ethyl ester 0.85 ± 0.22 0.00043 ± 0.00011 >20b 2000:1:>46000 L-cysteine sulfinate 1.3 ± 0.4 0.25 ± 0.02 >20b 5:1:>80 L-homocystine >200b 0.47 ± 0.10 >20b >400:1:>40

S-carbamoyl-L-cysteine >200b 0.13 ± 0.02 >20b >1500:1:>150 b b D-glucosamic acid >200 3.2 ± 0.6 >20 >60:1:>6 b Propionate 0.46 ± 0.14 0.57 ± 0.07 >20 1:1.2:>40 D-2,3-diaminopropionate 0.27 ± 0.05c 0.47 ± 0.08c >10b 1:1.7:>40 3-bromopyruvate >200b >200b 0.34 ± 0.07 >500:>500:1 Maleimide >200b >200b 0.14 ± 0.04 >1400:1400:1 N-iodosuccinimide 0.006 ± 0.002 0.006 ± 0.002 >10b 1:2.8>1600 b b Pyridoxal-5-phosphate 0.14 ± 0.04 >200 >20 1:>1400:>140 4-Hydroxybenzophenone 2.4 ± 0.42 0.33 ± 0.05 10 ± 2 8:1:30 b b 5-Chloro-2-nitrobenzaldehyde >10 0.53 ± 0.17 >10 >20:1:>20 2,6-Dibromoquinonechlorimide >10b 0.99 ± 0.11 >10b >10:1>10 2-Chloro-3’,4’- >10b 1.2 ± 0.2 >10b >8:1:>8 dihydroxyacetophenone b b Bromomethyl cyclohexane >20 >20 0.25 ± 0.08 >80:>80:1 b b 3,4-Dihydroxyphenylacetate >20 >20 0.63 ± 0.14 >30:>30:1

aRatio of Ki values against the three forms of ASADH tested bNo inhibition was observed; the Ki is estimated to be at least 10 times greater than the highest concentration examined. cAdjusted for inhibition by only the D-isomer. Bold font indicates compounds from this set of inhibitors that were selected for further examination and development.

Chapter 2

Membrane Fusion Proteins

2.1 Introduction

Membrane proteins are essential in a variety of cellular functions, making them important potential drug targets. Compared with soluble proteins, membrane-bound proteins are more difficult to deal with because of its hydrophobic nature. The challenge in membrane protein structural studies lies in the production of the required protein sample for characterization, as well as in developing methods to effectively solubilize and maintain a functional and stable form of the target membrane during the course of crystallization experiments. Progress in the structural investigation of membrane proteins has proven to be a difficult task, thus limiting the number of solved structures of membrane proteins as compared to the structural information available for soluble proteins.

To address these issues, different approaches have been introduced to obtain crystal structures of membrane proteins. These include the use of specialty detergents that are commonly used for membrane protein solubilization (Prive, 2007; Walian et al.,

2004; Newstead et al., 2008). Another technique is the lipidic cubic phase and bicelle methods which were introduced to provide a more natural environment for membrane

proteins, both of which were reported to be effective in obtaining crystal structures of various types of rhodopsin (Landau & Rosenbusch, 1996; Faham & Bowie, 2002 ; Faham et al., 2005; Kolbe et al., 2000).

Another alternative approach for membrane protein crystallization is the use of antibody fragments to provide an additional polar surface area that could be utilized for effective lattice contacts. This antibody-mediated crystallization has been shown to improve the crystallization of membrane proteins such as cytochrome C oxidase (COX) from Paracoccus denitrificans (Ostermeier et al., 1995 ; Ostermeier et al., 1997), yeast cytochrome bc1 complex (Hunte et al., 2000) and potassium channel protein KcsA from

Streptomyces lividans (Zhou et al., 2001). The binding of antibody fragments to the membrane proteins were reported to provide additional polar surface contacts resulting in a more rigid, three-dimensional structure. While this method was proven successful in elucidating higher resolution structures of COX, bc1 and KcsA, this method has not proven to be effective in the crystallization of other membrane protein targets. In addition, the use of antibodies is a time consuming and expensive approach, eliminating its possibility to be used in routine crystallization experiments.

Though each of the three described approaches are elegant tools in membrane protein structural studies, there is still a need for alternative techniques to provide a more general method to obtain crystal structures of membrane proteins. The addition of fusion domains to the membrane protein is an attractive substitute to antibody-mediated crystallization. These fusion domains can also increase available hydrophilic surface contacts that may lead to crystal formation, thus serving as an antibody mimic without the need to express an antibody for each membrane protein target. The use of soluble

fusion domains like maltose binding protein (MBP), glutathione-S-transferase (GST) and thioredoxin has been reported to be successful in the crystallization of soluble proteins

(Smyth et al., 2003). However, so far this approach has not been proven to be efficient in obtaining high resolution structures of membrane proteins.

In this study, we aim to investigate the use of different soluble protein domains in fusion with membrane proteins to obtain diffraction quality crystals for high resolution studies. The approach was initially tested against a model protein, KcsA, and later will be applied to structural studies of other membrane protein targets.

2.2 Target Membrane Fusions

For this study, we utilize the potassium ion channel (KcsA) from Streptomyces lividans as a model protein. KcsA is an integral membrane protein that catalyzes selective potassium ion transport. The initial structure of KcsA was determined by Doyle et al.,

(1998) at 3.0 Å resolution, however the mechanism of ion coordination and hydration of this channel has been fully illustrated upon solving the structure of KcsA complexed with a Fab fragment (Zhou et al., 2001). This protein was selected as a test case for an alpha helical type of integral membrane protein for initial experiments. Also, the structure of

KcsA has been solved with an antibody fragment which will allow comparison between the effectiveness of the membrane protein fusion technique against antibody-mediated crystallization for structural characterization of membrane proteins.

Different soluble domains that were commonly used as affinity tags were chosen based on reported crystal structures of fusion proteins. Large affinity tags such as maltose binding protein (MBP), glutathione-S-transferase (GST) and thioredoxin (TRX) have

been previously fused with target proteins and successful crystallization of these fusion proteins have been reported. For example, fusion of MBP into protein targets like the ectodomain of human T cell leukemia virus type 1 protein (gp21), S. aureus accessory regulator R (SarR) and the MATa1 protein fragment from S.cerevisiae enabled structure determination of these proteins (reviewed by Smyth et al., 2008). GST-fusion proteins have also been utilized to obtain crystals for the structure determination of 8 different sets of small peptides and biological regulatory domains with residues ranging from 6 to 319 amino acids in size. The crystallization and structural features of these fused GST segments are summarized in a review article by Zhan et al., (2001). Fusion with thioredoxin has also been demonstrated to drive the crystallization of vancomycin resistance protein (vanH) from Entercoccus faecium with crystals diffracting to 3 Å, however no structure has been reported (Stoll et al., 1998). A more recent study utilized this same carrier-driven crystallization approach led to the structure solution a homology motif domain of splicing factor Puf60 (Corsini et al., 2008)

The utility of these three fusion tags was exploited to create fusions with the potassium channel protein, KcsA. In addition, two other affinity tags, such as NusA protein and ketosteroid isomerase, were included in our initial set of fusion constructs.

These five soluble domains have molecular weights ranging from 13 kDa to 54 kDa, which could also be further evaluated for the effect of fusion size in making crystal contacts. All the fusion constructs carried an additional hexahistidine moiety attached at the N-terminus of each soluble domain to allow purification by affinity chromatography.

Another advantage of using these fusion tags is the availability of crystal structures of these that could serve at templates for molecular replacement methods.

2.3 Generation of Fusion Constructs

To facilitate the generation of different fusion constructs, Gateway Cloning

Technology was employed to rapidly clone the target KcsA sequence into multiple destination vectors to express various tagged versions of the protein. This technology is based on bacteriophage lambda site-specific recombination system, wherein the reactions are driven by strand exchange between specific attachment (att) sites on interacting DNA molecules.

Two reactions are involved in Gateway Cloning (Figure 2.1). The first reaction is the BP recombination which generates an attL-containing entry clone from a reaction between the target gene containing an attB site and a donor vector with an attP site. This reaction is catalyzed by BP Clonase enzyme (Integrase, E.coli Integration Host Factor,

IHF). On the second step, the attL-containing entry clone is then recombined with an attR-containing destination vector to create the desired fusion protein expression clone.

This reaction is catalyzed by LR Clonase (Integrase, E.coli Integration Host Factor, IHF,

Excisionase, Xis).

2.3.1 Generation of att Sites

The target KcsA gene was first amplified by polymerase chain reaction (PCR) using gene-specific primers that incorporate attB sites on the target gene. The sequence of the forward primer is as follows: 5’-GGG ACA AGT TTG TAC AAA AAA GCA

GGC TTC ATG CCA CCC ATG CTG TCC GGT TCT TCT G-3’, whereas the reverse primer has the following sequence, 5’ GAC CAC TTT GTA CAA GAA AGC TGG

GTC TCA CCG GCG GTT GTC GTC GAG CAT TCG-3’. The att sequences are

highlighted, whereas KcsA specific sequences are underlined. After setting up various reaction mixtures, gene amplification was obtained in a 50 μL reaction mixture containing 20 ng DNA template, 0.4 μM each forward and reverse primers, 0.4 mM dNTP mix, 1 U of Vent polymerase, and 5 μL of polymerase buffer. Thermocycler conditions were set up as follows: denaturation step at 367 K for 30 sec; annealing step at

333 K for 30 sec; extension step at 345 K for 1 min for a total of 28 cycles.

Figure 2.1 Generation of fusion constructs. Diagram showing recombination reactions to create entry and expression clones using Gateway Cloning Technology.

2.3.2 Generation of Entry Clone

The PCR product (attB-KcsA) obtained from the previous step was purified using

QIAQuick PCR purification kit (Qiagen) and was cloned into an attP site containing vector (pDONR 221) in a BP recombination reaction to create the entry clone. Briefly,

17 ng of attB-KcsA was incubated with 150 ng of pDONR 221, 2 μL of BP clonase and

TE buffer to make a 10 μL reaction mixture. The BP reaction was allowed to proceed for

12 hours at 298 K, followed by transformation into OmniMax 2T1 cells in LB agar plates containing 50 μg/mL kanamycin. Colonies were selected and entry clones were isolated using commercially available plasmid extraction kit, and was subsequently sent for sequencing.

Agarose gel profiles of the entry clone show an intense band migrating at a region of lower size as compared to the expected size for the entry clone (~3030 bp). However, further confirmation from sequencing analysis revealed that the gene insert in the entry clone showed 100% sequence identity with the target KcsA gene, indicative of successful generation of the entry clone (Figure 2.2).

2.3.3 Generation of Fusion Protein Expression Clones

The five fusion protein expression vectors were constructed thru an LR recombination reaction. Different vectors encoding different soluble domains were kindly provided by Dr. Michael Wiener and Dr. Peter Horanyi (University of Virginia). The destination vectors encode H6-Nus, H6-KSI, H6-GST, H6-Trx and H6-MBP and will introduce an N-terminal fusion to KcsA to yield the fusion protein. Each reaction mixture was composed of 100 ng of entry clone, 150 ng of destination vector, 2 μL of LR Clonase

and TE buffer to make 10 μL reaction mixture, followed by incubation at 298 K for 12 hours. After the incubation period, 1 μL of each reaction mixture was transformed into

OmniMax T1 cells with resistance to ampicillin for selection. Expression clones were linearized with restriction enzymes and were visualized in agarose gels. Comparison of the DNA gel profiles of the empty vectors against the expression clones indicated that the fusion constructs migrate at the expected size (Figure 2.3).

A M (kbp) Entry Clone

10 8 6 5 4 3

0.5

Figure 2.2 Analysis of the entry clone. (A) Agarose gel electrophoresis of the isolated entry clone (KcsA-pDONR221). (B) Sequencing analysis of the entry clone showing exact match of the nucleotide sequence between the gene insert (labeled as Entry) and the target KcsA sequence.

B Protein Domain Vector Size (bp) Expression Clone Plasmid Size (bp) (1) H6-NUS 8724 (2) H6-NUS-KcsA 7503 (3) H6-KSI 7485 (4) H6-KSI-KcsA 6264 (5) H6-TRX 7728 (6) H6-TRX-KcsA 6507 (7) H6-GST 8932 (8) H6-GST-KcsA 7711 (9) H6-MBP 9367 (10) H6-MBP-KcsA 8146

Figure 2.3 Generation of fusion protein expression clones. (A) Agarose gel profiles of the expression clones resulting from the LR recombination reaction between the entry clone and various destination vectors. (B). Summary of the destination vector size and the expected size of the fusion constructs.

2.4 Pilot Protein Expression of the Fusion Constructs

Pilot protein expression of the fusion constructs was carried out following the method of Mohanty and Wiener (2004) with some modifications. Briefly, each of the fusion constructs was transformed into four different E. coli cell lines (C41, C43,

BL21(DE3), XL-1 Blue) and 25 mL cultures were grown at 310 K with 250 rpm shaking until OD600 reached 0.8-1.0. The production of the protein was induced with 0.5 mM

IPTG and was allowed to grow for four more hours. One mL aliquots were taken every hour during post-induction to monitor cell growth and control samples were taken prior to IPTG induction. Cell samples were centrifuged at 12,000 rpm and the resulting pellet was resuspended in SDS sample buffer. After 10 minute incubation, the sample was spun again at the same speed and 15 μL of the supernatant from this step was loaded onto an

SDS-PAGE gel. The gel was then transferred to a nitrocellulose membrane for Western blot analysis using a polyhistidine antibody for detection. Protein expression was qualitatively determined by comparing the control and sample lanes for the presence of protein bands that represent the target fusion.

All of the fusion proteins were well expressed in C41 and BL21 (DE3) cells only, with little or no expression in either C43 or XL-1 Blue cells (Figure 2.4). At least two protein bands were present in each of the fusion proteins except for the H6-KSI-KcsA fusion wherein a single distinct band was observed in the Western blot profiles. These multiple bands may represent both monomeric and higher oligomeric forms of the fusion protein, though these bands migrate at higher positions relative to the expected size based on the molecular weight markers. For example, the expected size for H6-KSI-KcsA fusion would be 34 kDa (Table 2.1), but a higher band with an apparent Mr of ~43 kDa

was observed. This gel shifting phenomenon appears to be common with various helical membrane proteins where the apparent molecular weight can deviate widely with the formula weight of membrane proteins. Rath et al., (2009) compared the migration distance of several helical membrane proteins on SDS-PAGE gels and described the shifts in migration in terms of dMW values, where dMW = (apparent MW – formula

MW)/formula MW x 100%. For 17 types of helical membrane proteins analyzed which include the tetrameric form of KcsA , dMW values between -46% to +48% were determined. Negative values indicate faster migration whereas positive dMW values indicate slower movement of the protein in the gel.

In our case, we have seen higher apparent Mr for either the monomeric or higher oligomeric forms of all five fusion proteins. This aberrant migration could occur for various reasons. Detergent binding could affect the migration where lower amounts of

SDS molecules bound to the fusion protein will result in higher gel shifts. This detergent binding is in turn affected by the degree of protein folding present in the fusion constructs. If the fusion protein is well folded and has a compact structure because of more favorable protein-protein versus protein-detergent interaction, there is a tendency to bind low amounts of detergent which will consequently lead to slower migration in the gel. Moreover, the anomalous migration of possible oligomeric forms of the fusion proteins in our samples could be attributed to the stability of the tertiary structure which may be resistant to detergent treatments.

A. H6-NUS-KcsA C41 C43 XL1 Blue BL21 (DE3)

0 1 2 3 4 M 0 1 2 3 4 M 0 1 2 3 4 0 1 2 3 4

182 182 116 116 82 82 64 64 49 49 37 37

B. H6-KSI-KcsA C41 C43 XL1 Blue BL21 (DE3)

M 0 1 2 3 4 0 1 2 3 4 M 0 1 2 3 4 0 1 2 3 4

182 182 116 116

82 82

64 64 49 49 37 37

26 26 19 19 15 15

Figure 2.4 Protein expression analysis of various fusion constructs. Western blot profiles of (A) H6-NUS-KcsA and (B) H6-KSI-KcsA. Four different E. coli cell lines were tested. Numbers (0-4) indicate the post-IPTG induction time in hours. The molecular weight marker (M) ladder is labeled with the corresponding Mr values (15-182 kDa).

C. H6-Trx-KcsA C41 C43 XL1 Blue BL21 (DE3)

M 0 1 2 3 4 M 0 1 2 3 4 M 0 1 2 3 4 0 1 2 3 4

182 182 116 116

82 82

64 64

49 49

37 37

26 >181.8 26 19 19 15 15

D. H6-GST-KcsA C41 C43 XL1 Blue BL21 (DE3) 0 1 2 3 4 M 0 1 2 3 4 M 0 1 2 3 4 0 1 2 3 4

182 182 116 116 82 82 64 64 49 49 37 37 26 26 19 19 15 15

Figure 2.4 Protein expression analysis of various fusion constructs. Western blot profiles of (C) H6-Trx-KcsA and (D) H6-GST-KcsA. Four different E. coli cell lines were tested. Numbers (0-4) indicate the post-IPTG induction time in hours. The molecular weight marker (M) ladder is labeled with the corresponding Mr values (15-182 kDa).

E. H6-MBP-KcsA C41 C43 XL1 Blue BL21 (DE3) M 0 1 2 3 4 0 1 2 3 4 M 0 1 2 3 4 0 1 2 3 4

182

116 182

82 116

64 82

49 64

37 49

37 26

19 26 15 19

Figure 2.4 Protein expression analysis of various fusion constructs. Western blot profiles of (E) H6-MBP-KcsA. Four different E. coli cell lines were tested. Numbers (0- 4) indicate the post-IPTG induction time in hours. The molecular weight marker (M) ladder is labeled with the corresponding Mr values (15-182 kDa).

Table 2.1 Expected Mr values of the fusion protein. Protein Cell Paste Yield Mr Tag (kDa) Mr Fusion (kDa) (g/L) KcsA 1.5 17 H6-NUS-KcsA 4.5 54 72 H6-KSI-KcsA 4.5 16 34 H6-TRX-KcsA 4.5 13 31 H6-GST-KcsA 5.0 25 43 H6-MBP-KcsA 4.5 42 60

2.5 Cell Growth

Large scale production of the fusion proteins was done with C41 cells using the same conditions described from the pilot expression analysis. C41 cells were chosen over

BL21 cells as the expression cell line, because comparison of cell growth curves of C41 and BL21 cells showed a growth lag in BL21, which may indicate that initial production of the fusion proteins caused toxicity to the cells. Moreover, C41 cells have been reported to be a viable E. coli strain for the production of recombinant membrane proteins

(Miroux & Walker, 1996).

The average cell paste yields of the fusion proteins ranges from 4 to 5 g/L, which were three times higher than the yield obtained from the growth of native KcsA (1.5 g/L in XL1 Blue cells). Cells were harvested by centrifugation and stored at -80°C until use for membrane protein extraction.

2.6 Membrane Fusion Protein Extraction

All protein extraction steps were done at 277 K unless otherwise specified. Cell paste were resuspended in Buffer K (20 mM Tris-HCl, pH 7.5, 150 mM KCl, and 1 mM dodecyl maltoside (DDM), protease inhibitor cocktail), homogenized by hand and subsequently subjected to cell lysis by sonication (50% power level, 30 sec exposure/cycle, total of 5 cycles). The resulting lysate was spun at 35,000 rpm for 1 hr to pellet out the membrane fraction which was resuspended in Buffer K + 8 mM DDM. The higher concentration of the detergents allows effective solubilization of the membrane fusion protein. This membrane fraction was homogenized by hand, followed by slow shaking on a platform shaker for 1 hr to extract the membrane fusion protein. After

shaking, the sample was centrifuged at 40,000 rpm for 1 hr to pellet the extracted membrane and retrieve the supernatant containing the detergent-solubilized membrane fusion protein.

2.7 Purification of H6-MBP-KcsA Fusion

Small scale purification of the H6-MBP-KcsA fusion was initially done using gravity flow columns for optimization purposes. Crude H6-MBP-KcsA fusion was passed thru a 5 mL bed volume Co-IMAC column pre-equilibrated with Buffer K + 5 mM imidazole and step elution was performed using the same buffer with 100, 200 and 500 mM imidazole. Relevant fractions were seen eluting at 100 and 200 mM imidazole concentrations with prominent bands at ~64 kDa marker and a higher molecular weight band above the 182 kDa region which may represent the monomer and a higher oligomeric form of the fusion (Figure 2.5A). Due to lesser amounts of impurities the samples eluting at 200 mM imidazole were collected, concentrated and loaded onto a small scale gel filtration column (Superdex 200 TM, colume volume of 15 mL, separation range of 10-600 kDa globular proteins) for further separation. After 1.5 CV elution with

Buffer K, minimal separation between the two oligomeric forms was seen after the gel filtration step, which could probably due to the inefficient separation due to the smaller column volume used.

To facilitate production of pure protein for crystallization trials, the H6-MBP-

KcsA fusion was purified following the same procedure as previously described but with the use of automated purification systems. A linear gradient between 5 mM -500 mM imidazole was utilized for the Co-IMAC purification (Figure 2.5C) while a higher

resolution gel filtration matrix (Sephacryl HR200, separation range 5-250 kDa globular proteins) with a longer bed volume was used for the second step (Figure 2.5D). Also, high molecular weight markers (Invitrogen, LC5699) was run as standard for gel electrophoresis.

Similar protein profiles were observed for both batches of purifications. Elution chromatogram from the gel filtration step showed two protein peaks (Figure 2.5D) and these protein peaks correspond to two different forms of the H6-MBP-KcsA fusion protein. The SDS-PAGE profile showed separation of the monomer fusion with that of the tetramer fusion with very faint bands co-eluting with the target samples. The tetrameric form travels above the 247 kDa marker, while the monomeric fusion migrates at ~50 kDa region (Figure 2.5C). There is a discrepancy observed between the migrations of the monomer fusion when using two different protein markers. In the small scale purification, monomeric form travels as a ~64 kDa protein whereas same protein was seen at the ~50 kDa region when a high Mr marker was used. It is unknown why the monomer fusion has distinctive apparent Mr values when run in gels with different markers. The identities of these protein fusions could be further confirmed by mass spectrometry analysis.

Both the protein peaks were subjected to dynamic light scattering measurements

(DLS) to determine the amount of polydispersity as well as to get additional information about the size of the macromolecules. Three trials were performed for each fusion protein with 0.5 mg/mL concentration dissolved in Buffer K. The Mr of the samples in solution was in the range of the expected molecular weights of the monomer (60 kDa) and tetramer H6-MBP-KcsA fusion (240 kDa) (Figure 2.6). As expected based on the purity

of the sample on the gel profile, a moderate amount of % polydispersity (less than 30%) was observed for both forms of fusion protein. However, these samples could still be used for preliminary crystallization trials, thus we proceeded with screening for crystal conditions for both the tetramer and monomer fusion.

M crude unb wash 5mM 5 mM 100 mM 200 mM 500 mM M F3 F4 F5 F6 F7 A B

182 182

116 116

82 82

64 64

49 49 37 37

26 26 19 19 15 BTABk2p45001:10_UV2_280nm BTABk2p45001:10_UV3_260nm BTABk2p45001:10_Logbook 15 mAU 9 9

C Markers Crude Co-IMAC400 Peak I Peak II D Gel filtration Chromatography 420

247 214 LEGEND: 160 Abs280 nm 107 Abs 260 nm

300

200

100

I II 0 0 50 100 150 ml Figure 2.5 Purification of H6-MBP-KcsA fusion. Small scale purification (A, B) and large scale automated purification using Co-IMAC and gel filtration chromatography (C, D)

A. Peak I : Tetramer Fusion Trial 1 Trial 2 Trial 3

B. Peak II: Monomer Fusion Trial 1 Trial 2 Trial 3

C. Summary of DLS Results

Peak 1 R (nm) % Polydispersity Mr (kDa) Tetramer Fusion Trial 1 6.4 9.7 262 Trial 2 6.4 24.8 262 Trial 3 6.4 27.2 263

Peak 2 R (nm) % Polydispersity Mr (kDa) Monomer Fusion Trial 1 3.5 17.8 61 Trial 2 3.4 19.5 60 Trial 3 3.5 18.1 63

Figure 2.6 Dynamic light scattering measurements of H6-MBP-KcsA fusion. Measurements were performed in triplicates for both (A) Tetramer fusion and (B) Monomer fusion. (C) Summary of parameters gathered from the DLS analysis.

2.8 Crystallization of H6-MBP-Fusion

The tetramer and monomer H6-MBP-KcsA fusion were both screened at a protein concentration of 10 mg/mL with a total of 576 sparse matrix conditions from various commercial and custom made crystallization kits. Crystallization experiments were done by mixing equal volumes of each fusion protein with reservoir buffer, followed by incubation in a 96-well sitting drop vapor diffusion tray containing 70 μL of reservoir buffer in each well. The trays were then incubated at 293 K and were monitored for crystal growth.

After several weeks of incubation, crystals with different morphologies (Figure

2.7) grew from drops set up for the tetrameric form, whereas no crystals hits were obtained from the monomer fusion in all of the conditions tested. Four different conditions were found to be favorable in obtaining initial hits (Table 2.2). Clusters of plate crystals as well as thin needles were isolated in the drops. The needles were too small to be tested for diffraction quality; however the plate crystals from the other three conditions were confirmed to be protein crystals. The best resolution was obtained from the plate crystals grown in the presence of 30% PEG 4000, 0.1 M Na-cacodylate pH 6.5,

0.08 M MgCl2, which diffracted to 15 Å at the synchrotron. Though these crystals diffract at very low resolution, these preliminary diffraction data had encouraged us to pursue and optimize the identified initial conditions to get better diffraction quality crystals.

The initial conditions were expanded using 24-well hanging drop vapor diffusion set up with different pH values in one dimension against precipitant concentration on the other. Different drop sizes were prepared relating to different ratio between the fusion protein and the reservoir solution.

Table 2.2 Crystallization conditions from H6-MBP-KcsA sparse matrix screening. Kit Precipitant Buffer Additive Morphology

(A) Natrix 30% PEG 4000 0.1 M Na- 0.08 M Cluster of plate cacodylate MgCl2 crystals pH 6.5 (B) JB Basic 30% PEG 8000 0.1 M MES 0.20 M Small plate pH 6.5 CH3COONa crystals (C) JB Basic 20% PEG 4000 0.1 M Hepes 10% v/v Needle clusters pH 7.5 2-propanol (D) Nextal 0.5 M 0.1 M Na- 1.0 M Rectangular plate Classics (NH4)2SO4 citrate pH Li2(SO4) crystals 5.6

A B

C D

Figure 2.7 Tetrameric H6-MBP-KcsA fusion crystals.

The expansion yielded “oil-like” crystals with no distinct three-dimensional shape. Conditions A, B and C gave similar oil-like crystals which appeared after three weeks of incubation at 293 K. Further incubation at the same temperature did not lead to the formation of crystals. Crystal screening for diffraction quality was attempted for these crystal forms, however these crystals are very fragile and displayed a “buttery “ consistency when touched by the loop, making it hard to scoop for diffraction quality screening as well as in transferring them as crystal seeds (Figure 2.8). These “oil-like” crystals were seen for the tetramer fusion only, while mostly clear drops to light precipitation was observed for the monomeric form.

These “oil-like” crystals may either represent detergent or protein crystals that could phase out of the solution. Whether these crystals are fusion protein crystals or not are unknown and have not been explored. The identity of these samples could be confirmed by x-ray diffraction screening, but the crystals are too fragile to be mounted on the loop. One possible alternative is to take enough samples, followed by washing with buffer to minimize the presence of PEG precipitants and then by running an SDS-PAGE gel. Upon staining with either Coomassie blue or silver stain, protein crystals will show bands whereas detergent crystals would not be visualized in the gel. Alternatively, measuring the absorbance at 280 nm would also confirm whether these ar protein or detergent crystals.

Since we are getting mostly “oil-like” crystals and clear drops from most of the expansion conditions, we have decided to increase the protein concentration to twice the original value (20 mg/mL) and redo the expansion for both the tetrameric and monomeric forms of the protein using the four initial conditions. The excess protein samples were

further used to identify new conditions from other sparse matrix screens that were not previously tested.

A1 A2

B C

10 um 20 um

Figure 2.8 Expansion of initial crystallization conditions for H6-MBP-KcsA. Oil-like crystals were obtained from expansion of the initial sparse-matrix screening.

An increase in the protein concentration for the tetramer fusion still showed the growth of “oil-like” crystals in most drops after a week of incubation. The trays were checked every week for a period of two months and every month thereafter. After six months, small trigonal crystals measuring 10-20 um (Figure 2.9A3) grew on the edge of the drops. When tested for diffraction quality, these crystals diffract to poor resolution

(~20 A). No new crystal hit was identified for the tetramer fusion based from the two screen kits that were tested using the same batch of samples (Figure 2.9A).

The monomer fusion was also screened in parallel using the conditions that were identified for the tetramer fusion (Table 2.2). Expansion of Condition A by varying different parameters (pH, precipitant concentration, additive concentration, drop sizes) gave elongated plate crystals (Figure 2.9B1) while a new crystallization condition was discovered from sparse matrix kits. Clusters of plate crystals emerged from drops incubated in the presence of 20% PEG 2000 MME, 0.1 M MES pH 6.5, 0.1 M NaCl, 0.1

MgCl2 at 293K (Figure 2.9B2). Single crystals were isolated from the drops and were used as seeds for optimization of crystal quality. Slight variation of the initial conditions gave bigger crystals of the same morphology (Figure 2.9 B3) that are now suitable to be tested for diffraction quality.

A. H6-MBP-KcsA Tetramer Fusion Crystals 1 2 3

B. H6-MBP-KcsA Monomer Fusion Crystals 1 2 3

Figure 2.9 Crystals of the tetrameric and monomeric H6-MBP-KcsA fusion. Crystals of the tetramer fusion (A1) was obtained in the presence of 30% PEG 4000, 0.1 M Na-cacodylate pH 6.5, 0.08 M MgCl2. Expansion of this initial condition resulted to an initial growth of oil-like crystals (A2), which was followed by the growth of trigonal crystals after six months incubation (A3). Clusters of plate crystals (B1) of the monomer fusion grew from expansion trays of 30% PEG 4000, 0.1 M Na-cacodylate pH 6.5, 0.08 M MgCl2. Another condition from sparse matrix screening of the monomer fusion gave plate crystals (B2) which grew from 20% PEG2000MME, 0.1 M MES pH 6.5, 0.1 M NaCl, 0.1 M MgCl2. Expansion of this B2 condition led to crystals with the same morphology (B3).

2.9 Data Collection and Structure Determination

Fusion crystals suitable for x-ray data collection were cryoprotected with either

25% ethylene glycol or 25% glycerol, dipped in liquid nitrogen and maintained at this condition until data collection. Assessment of diffraction quality was done in-house with a Rigaku FR-E rotating generator equipped with an R-axis IV image plate detector. Data collection under cryo condition was subsequently performed at station GM-CAT Sector

23B at the Advanced Photon Source operating at a wavelength of 1.00 Å. A data set with

84.5% completeness was collected overall and the acquired dataset was processed using

HKL-2000 (Otwinowski & Minor, 1997). Scaling was done with the SCALEPACK program within the CCP4 program suite. The crystal diffracted to 2.5 Å resolution and belongs to the P212121 space group with unit cell parameters a = 50.05 Å, b = 58.42 Å, c

= 124.03 Å and all angles at 90° (Table 2.3). The asymmetric unit is composed of a single molecule giving a Matthews coefficient of 2.24 Å3/Dalton and a solvent content of ~45%.

Both structures of MBP (PDB Code: 1LLS) and KcsA (PDB Code: 1BL8) have been previously solved and were used as search models to facilitate the structure determination of the H6-MBP-KcsA fusion. In addition to these two search models, we also explored the Protein Databank for MBP-fusion structures with a flexible linker region that is analogous to the design of our target H6-MBP-KcsA fusion.

Different search models were used to solve the structure of the fusion by molecular replacements using both the Molrep and Phaser programs within the CCP4 suite. Search models that were utilized include the tetramer and monomer of KcsA

(1BL8), a combination of KcsA (1BL8) and MBP (1LLS) with either full or truncated

KcsA structure, full structure of MBP (1LLS) alone, and full or truncated versions of

MBP-helical protein fusion (1Y4C).

After all the MR calculations performed from both programs, Molrep solutions were only found when MBP was used as a search model. Out of the four model combinations tested, best R factor of 0.41 and a score of 0.58 were obtained for the MR search utilizing the full structure of MBP-helical protein (1Y4C). The other search models have R factors which range between 0.54 to 0.58 and scores of 0.11 to 0.29. No solution was found for calculations using KcsA as search model.

The model was refined by restrained refinement with Refmac achieving a final structure having refinement values of 0.25 and 0.31 for the Rwork and Rfree, respectively.

Generation of electron density maps modeled the MBP moiety in the crystal structure with high confidence, whereas no obvious electron density could be accounted for the fusion partner (KcsA) from maps calculated with the phases from MBP. The final electron density map allowed complete identification of MBP residues 4-366 in the experimental model (Figure 2.10A). No electron density was seen for the linker between the fused proteins. In the search model used, MPB was connected at the N-terminal of an alpha helical protein (DHP) that was solved at 1.9 Å. In this previous study, MBP was also utilized for the structure determination which allowed the complete building of interpretable maps that correspond to 77% MBP and 23% DHP (Figure 2.10B).

Examination of the structure downloaded from protein databank however did not show the glycine rich region that links the MBP with that of DHP. This suggests that the region may be disordered showing no electron density even at high resolution (Laporte et al.,

2005).

Table 2.3 Data collection and refinement statistics for H6-MBP-KcsA fusion monomer. Data Collection Statistics

Parameters Values

Temperature (K) 100

Space Group P212121 Unit Cell dimensions a, b, c (Å) 50.05, 58.42, 124.03 α, β, γ (°) 90, 90, 90 Wavelength (Å) 1.033

Resolution (Å) 2.50 (2.54-2.50) a

Total reflections 372924

Unique reflections 11127

Mosaicity (°) 0.4

a Rsymm 0.092 (0.22)

Output 14.5 (2.7) a

Completeness (%) 84.5 (41.1) a

Redundancy 5.5 (2.2) a

Refinement Statistics

Resolution range (A) 62.02– 2.50

Number of reflections 10552

Rwork/Rfree 0.25/0.31

R.m.s.d bonds 0.009

R.m.s.d angles 1.149

Number of molecules per ASU 1 aHighest resolution shell.

A. Experimental Model B. MR Search Model (1Y4C)

Figure 2.10 Structure determination of putative H6-MBP-KcsA monomer fusion. (A) Overall structure obtained from the refinement of the experimental model showing the presence of MBP moiety and absence of the fusion partner KcsA. (B). The structure of the search model (1YC4) used for molecular replacement showing MBP and its fusion partner (DHP, in red). The linker region connecting these two proteins was missing in the structure. The blue region denotes the N- and C- termini of MBP.

In our case, the structure of the putative H6-MBP-KcsA fusion only accounted for the presence of the MBP, and no additional electron density could be correlated to the presence of the passenger protein KcsA. Based from our gel electrophoresis and DLS data, we are isolating the monomer and tetramer forms of the fusion constructs, though x- ray crystallographic studies suggest otherwise. The missing KcsA in the structure could possibly be due to following reasons:

First, the full fusion protein may have been expressed as a whole or as a truncated version of the fusion during cell growth and protein production, which could result from inefficient generation of the fusion construct. This could be verified by sending the expression clone for sequencing to establish whether the linkage between MBP and KcsA has been formed. In addition, freshly purified protein samples from the last purification step could be sent for mass spectrometric analysis (MALDI) to obtain information on the molecular weight which could be correlated to the oligomeric forms of the fusion protein.

Second, assuming that the fusion protein has been expressed and isolated in full length, it is also possible that the absence of KcsA in the structure could be due to proteolytic cleavage during the crystallization trials. Initial crystals from sparse matrix screens usually grow between 3 weeks to 3 months. The putative full length monomer fusion may have undergone proteolytic cleavage during this time, resulting in two different protein entities in the crystallization drop, one of which is crystallizable while the other is not. In this particular case, the H6-MBP-KcsA fusion crystals that were used as seeds were isolated after 3 months of incubation at 293K. These seed crystals were transferred to drops containing freshly prepared purified protein and well reservoir. From this point, crystals grew after three days and were screened for diffraction quality. The

seed crystals may already be the truncated form of fusion.

Alternatively, the full length form of the protein may be present in the crystallization drop; however the passenger protein (KcsA) may have a flexible conformation and is not stabilized by its fusion with MBP, preventing it to form favorable contacts to crystallize. The structure of KcsA has been solved as an apoprotein as well as a complex with a Fab fragment. In both structures, a carboxy-terminal loop was reportedly cleaved to promote crystallization and such cleavage allowed higher resolution structures of this channel (Zhou et al., 2001). In this study, full length KcsA was linked with MBP, and the presence of this C-terminal loop could have a detrimental effect for crystal growth within our experimental conditions.

2.10 Characterization of the H6-MBP-KcsA Fusion

2.10.1 DNA Sequencing of the Expression Clone

To verify whether we have formed the correct expression clone of the H6-MBP-

KcsA, the clone was sent for DNA sequencing (Integrated DNA Technologies). The DNA sequence was converted to amino acid sequence. Sequencing results were compared with the theoretical sequence of H6-MBP-KcsA in the absence of the linker region for easy identification of the location of the linker region. Amino acid comparison indicated the presence of the MBP protein (highlighted in red), and the KcsA protein (shown in blue) and a 14-amino acid linker region (boxed) that is linking the two proteins (Figure 2.11).

Though the sequencing results failed to give the whole amino acid sequence, the analysis has confirmed that the desired fusion construct has been successfully obtained.

Theoretical Expression Clone

Figure 2.11 Sequence analysis of the fusion expression construct. Amino acid sequence comparison between the expression construct and theoretical sequence of the expected expression clone. The MBP sequence is highlighted in red, KcsA in blue and the 14-amino acid linker region is boxed.

2.10.2 Mass Spectrometric Analysis

Freshly purified samples of the tetramer (Peak I) and monomer (Peak II) fusion obtained from gel filtration chromatography (Figure 2.5D) were sent for mass spectrometric analysis to the Mass Spectrometry and Proteomics Facility at Ohio State

University for determination of experimental theoretical weight of the fusion. A molecular weight of 60335 Da for the monomer fusion and a Mr value of 241,336 Da for the tetrameric form are expected based from mass calculations derived from the amino acid sequence.

MALDI analysis of the sample from the tetramer fusion (Peak I, Figure 2.12) indicated the presence of a major peak at 60,515.9 m/z which suggest the presence of a dissociated monomer in the samples. This dissociated monomer fusion was observed to co-exist with higher oligomeric forms of the fusion protein with minor peaks II to IV seen at 121,043.2, 181,184.2 and 243,464.9 which could be correlated to the dimer, trimer and tetrameric forms of H6-MBP-KcsA, respectively (Figure 2.12). This sample was isolated from the tetramer peak (Peak i) gel filtration), thus we expect the tetramer fusion to be of higher abundance as compared with the other forms of the fusion.

However, it seems like separation of the units of the tetrameric fusion occurs in time which led to the presence of these various Mr peaks. It is possible that dissociation occurs spontaneously and there is equilibrium between the different forms of the fusion.

Alternatively, the separation of the monomer units could be attributed to the change of optimum conditions when the detergent containing storage buffer was exchanged with ammonium acetate buffer that is used for MALDI analysis. Two other minor m/z mass peaks at 30,238.8 and 90,825.7 are also detected, which could either be fragments of the

fusion or protein contaminants that are present in the samples since the Mr values could not be directly associated with the theoretical Mr values for the fusion. Despite the presence of these contaminant peaks, the mass spectrometric data have established the presence of the desired H6-MBP-KcsA fusion which could initially exist as a tetramer that may further dissociate into lower oligomeric forms upon storage.

II III IV

Figure 2.12 Mass spectra of H6-MBP-KcsA tetramer fusion.

The mass spectrum of Peak II (putative H6-MBP-KcsA monomer fusion, Figure) was also obtained to verify our results from x-ray crystallography studies. Peak II was initially assigned as the H6-MBP-KcsA monomer fusion based on its migration in SDS-

PAGE (Figure). However, structure determination of diffraction quality crystals grown from the monomer fusion failed to show the target protein (KcsA) in the structure.

Therefore, mass spectrometric analysis will confirm whether a full length or truncated monomer fusion represents Peak II in the gel filtration profile.

The mass spectrum of Peak II showed an intense peak at 42,104.5 Da (Figure

2.13) and three other minor peaks (21,018.5, 63,417.7, 85,048.7 and 127,082.3). The major peak at 42,104.5 clearly indicates a truncation of the putative monomer fusion having Mr of 60,334 Da. Instead of the expected monomer fusion, the major peak represents H6-MBP which could result from truncation of the original construct. This result then correlates with the structural information obtained from diffraction data of the monomer fusion, whereas only the MBP part was fully accounted in the electron density.

MPB has been reported to assemble in higher oligomeric forms (Wallis & Drickamer,

1999) and the peaks at 85,048.7 Da and 121,082.3 Da could represent the dimer and tetramer forms of MBP. The other two peaks (21,018.5 and 63,417.7) could be protein impurities present in the samples.

Figure 2.13 Mass spectra of the putative H6-MBP-KcsA monomer fusion.

2.11 Summary and Future Work

Five expression fusion constructs with potassium channel protein, KcsA were successfully cloned and expressed in E.coli C41 cells. Among the designed fusion constructs, the fusion between the maltose-binding protein and KcsA was purified in two chromatographic steps. Two oligomeric forms of the fusion were isolated and were subjected to crystallization trials. Characterization by DNA sequencing established the generation of a fusion construct between the two proteins (MBP and KcsA), attached by a

14 amino acid linker. Mass spectrometric analysis of fusion samples indicate the presence of the dissociated monomers that co-exist with higher oligomeric forms of the fusion.

Sparse-matrix screening yielded crystals of the tetramer fusion which diffracted to only

15 Å resolution. Better diffraction quality crystals were obtained for the putative monomer fusion; however structure determination did not show the presence of the target protein in the membrane fusion structure. With these results, extensive optimization is necessary to obtain membrane protein fusion crystals that diffract to good resolution.

Future work could focus on optimizing the linker length as well as the identity of the residues in the linker region. This linker region seems to be crucial in obtaining good diffraction fusion crystals as evidenced by various examples of structures obtained by fusion with tags such as MBP, GST and thioredoxin. The success behind the structure determination of targets protein relied on determining the optimal linker length between the fused proteins. For example, crystals of protein targets from MBP fusion were only obtained when the linker length was reduced from 25 amino acids to 3-5 amino acids.

Linker residues were also changed into polyalanine to provide more rigid support (Smyth et al., 2003). A similar linker length (5 amino acid) proved to be effective in getting

diffracting crystals of protein fused with thioredoxin and the presence of this linker in the crystal structure are well defined in the electron density (Corsine et al., 2008). On the other hand, target protein crystallized as fusion with GST have longer linker length (9 amino acids) and contain a consensus linker sequence of SDLVPRGSM, with the last residue (methionine) being substituted with either serine and arginine in a few cases

(Zhan et al, 2001). Optimization of the length and structure of the linker peptide is important to minimize conformational flexibility as well as to allow proper folding of individual proteins in order to promote crystallization.

Chapter 3

Detergent Screening for Membrane Protein Extraction

3.1 Introduction

The challenge in dealing with membrane-anchored and integral proteins is in the isolation and solubilization of these proteins to yield a water-soluble, functional and homogeneous form in solution. In contrast to soluble proteins, membrane-bound proteins require an environment that mimics the lipid bilayer in order for them to be soluble and properly folded. The use of various detergents with different properties is usually employed to effectively solubilize membrane proteins outside of their physiological environment. However, to date there are no reported data that specifically guide the selection of particular class of detergents to extract a targeted membrane protein. Thus it will be helpful to find new strategies in identifying detergents that could be used for the efficient extraction of membrane proteins in general, as well as of specific classes of membrane proteins, while retaining their functional properties.

In this study, we have employed different detergent types to determine their extraction efficiencies against constitutive membrane proteins isolated from the total and outer membrane fractions of Escherichia coli cells. In addition, the potassium ion channel

protein, KcsA, was recombinantly expressed in E.coli cells and served as a model protein for the study. Eight detergent families were initially tested and the identities of the extracted constitutive membrane proteins were confirmed by MALDI-TOF mass spectrometry, coupled with database searching using the Mascot database (Matrix

Science).

3.2 Materials and Methods

3.2.1 Reagents and Chemicals

All chemicals and reagents were analytical grade and were used as received.

Dithiothreitol, iodoacetamide, ammonium bicarbonate and trifluoroacetic were purchased from Sigma. The solvent acetonitrile was from Fisher, whereas trifluoroacetic acid was obtained from Sigma-Aldrich. Tris base was from USB Corporation. Sodium chloride and Triton X-100 were purchased from Fisher. Precasted acrylamide gels, protein molecular weight markers and sodium dodecyl sulfate were procured from Invitrogen. All of the specialty detergents were obtained from Anatrace. Sequencing grade modified trypsin was from Promega. Bruker DALTONICS provided the α-cyano-4- hydroxycinnamic acid matrix as well as the peptide calibration standard kit. The peptide calibration kit includes the following standards with their respective monoisotopic mass values: angiotensin II: 1046.5418; angiotensin I: 1296.6848; substance P: 1347.7354; bombesin: 1619.8223; ACTH clip 1-17: 2093.0862; ACTH clip 18-39: 2465.1983; somatostatin 28: 3147.4710.

3.2.2 Plasmid and Cells

The pQE32 expression plasmid containing the KcsA gene which serves as our membrane protein control was kindly provided by Dr. Michael Wiener at the University of Virginia. Escherichia coli cell lines (C41 and XL-1 Blue) competent cells were purchased from Novagen.

3.2.3 Cell Culture and Growth

E. coli (XL-1 blue) cells were transformed with expression plasmids containing the KcsA gene grown on plates supplemented with ampicillin. Wild type cells were also transformed with empty expression vectors to confer antibiotic resistance. Colonies were picked and used to grow overnight cultures which were subsequently used as an inoculum for bulk cultures (4 x 1L LB Medium). Cells were grown in an incubator shaker with 250 rpm shaking speed, induced with isopropyl-β-D-1-thiogalactopyranoside

(IPTG) for protein production and then grown to late log phase. Cells were harvested by centrifugation at 10,000 rpm at 4 °C for 10 min and were stored at -80 °C until use. The bacterial cell pellets obtained from the different E. coli cell lines were used as the protein source for the target and constitutive membrane proteins.

3.2.4 Cell Membrane Preparation

Bacterial cells used for the production of the model protein KcsA and constitutive membrane proteins were resuspended in buffered solution and sonicated with a 50% duty cycle for ten cycles with a total exposure time of 5 min to disrupt the cell membrane, followed by centrifugation for 30 min at 10,000 rpm at 4 °C. The resulting membrane-

containing pellets were resuspended in buffer, centrifuged at 10,000 rpm for 20 min, and then repeated with a high spin wash at 35,000 rpm to separate any soluble or loosely associated proteins. The membrane proteins from this washed cell membrane were extracted as described below.

The supernatant obtained from the first centrifugation step (10,000 rpm) served as the source for outer cell membranes (OM). To isolate the outer cell membranes, the supernatant was spun at 34,000 rpm for 1 hr and the pellet was subsequently resuspended with a buffered solution containing 2 mM lauryl sarcosine. This mixture was incubated at room temperature for 1 hr with 20 rpm speed on a platform shaker, and centrifuged at

34,000 rpm to pellet the outer cell membrane.

Yeast cells were resuspended in buffered solution and lysed by using a Bead

Beater cell disruptor, followed by centrifugation and extraction.

3.2.5 Extraction and Solubilization

A variety of different membrane preparations were examined to measure the extraction efficiencies for specific detergents. Stock solutions of each detergent were prepared at 10 times their critical micelle concentration (CMC) values and diluted by a factor of five for the initial working concentrations. Cell membranes that were previously obtained were resuspended in buffer, homogenized with a hand-held cell homogenizer and divided into aliquots. A fixed volume of resuspended pellet was transferred into a buffered detergent solution and incubated with rocking at 4 °C for 3 hours. The samples were then centrifuged at 40,000 rpm for 1 hour to separate and remove the unextracted membranes. A set of six representative specialty detergents were initially chosen to

determine which classes of detergents were most effective for total protein extraction from the different cell types. Control reactions were run with buffer solutions and the extraction efficiencies for these specialty detergents were compared to a commonly used denaturing ionic surfactant (SDS) and a mild, non-ionic surfactant (Tween 20). Triton X-

100 was originally selected as the representative non-ionic surfactant, but its strong absorbance made protein concentration determination more difficult.

3.2.6 Gel Electrophoresis

SDS-PAGE was used to analyze the extraction efficiency of each detergent for the different constitutive and recombinant membrane proteins. Gels were run with equal protein loading and were visualized with Coomassie blue dye. The intensity of each band was quantified by using the ImageJ software package program (rs.info.nih.gov/ij).

Protein concentrations were determined by measuring the absorbance at 280 nm with the

Nanodrop Spectrophotometer. For analysis of total membrane extraction, protein concentrations were normalized per gram of cells for each membrane-detergent preparation.

3.2.7 MS Sample Preparation and Protein Identification

The bands corresponding to constitutive membrane proteins were examined by

MALDI mass spectrometry to identify the individual proteins. Protein bands were excised from the gels and were processed for trypsin digestion following the method of

Shevchenko, et al., (2006). Gel bands were washed and treated with acetonitrile, followed by solvent evaporation in a speedvac. Samples were then reduced by the

addition of freshly prepared DTT (10 mM DTT in ammonium bicarbonate) and incubated for 30 min at 56 °C. Free cysteines were then alkylated with 55 mM iodoacetamide for 30 minutes in the dark, followed by another washing step with acetonitrile. Solvents were completely removed by vacuum centrifugation.

In-gel digestion was performed by incubating the gel pieces with 50 μL of sequence grade trypsin solution (13 ng/μL stock solution in 100 mM ammonium bicarbonate/10% v/v acetonitrile). Enzyme digestion was allowed to proceed for 12 hr at

37 °C. The resulting supernatant which contains the digested peptides were transferred into tubes and spun in a vacuum concentrator for 2 hr to concentrate the peptide digest.

For the mass spectrometric analysis, the samples were prepared using the dried droplet method. Briefly, 1 μL of each sample was mixed with an equal volume of saturated α-cyano-4-hydroxycinnamic acid dissolved in 1:1 (v/v) acetonitrile:water with

1% TFA. The samples were spotted onto an MTP 384 ground steel target plate provided by Bruker Daltonics. Peptide standard mixture was prepared using the same method described above.

3.2.8 MALDI-MS Analysis

Mass spectra were acquired by an UltrafleXtreme MALDI TOF/TOF instrument

(Bruker Daltonics) in positive ion mode using a Smartbeam II laser operating at repetition rate of 1 kHz. External calibration was performed using a peptide standard mixture (Bruker Daltonics peptide calibration kit). The mass spectra were recorded in the m/z range between 800 and 3500 using reflectron mode and flexControl software. The signal obtained from a total of 1000 shots was summed for each peptide mass fingerprint

(PMF) search. For the MS/MS analysis, the LIFT TOF-TOF method was used to analyze the fragments formed by laser induced dissociation (LID). The MS/MS spectra were acquired using 1000-2000 laser shots for each sample and laser power slightly higher than the power used in the PMF experiments. Data analysis and PMF and MS/MS searches were carried out using the flexAnalysis (Bruker Daltonics) and Mascot (Matrix

Sciences), respectively. Search parameters allowed for a mass accuracy of ±0.2 Da, one allowed missed cleavage of trypsin with methionine oxidation and carbamido- methylation of cysteine as modification.

3.3 Results

3.3.1 Total Protein Extraction from Different Membranes

Different species of microorganisms have membranes with different structural organizations and will contain different classes of proteins to carry out the specialized functions of that organism. Extraction studies have been carried out on representative gram-negative enterobacteria (C41, BL21 and XL-1 Blue strains of Escherichia coli) and a methylotrophic yeast (Pichia pastoris) to determine how efficiently different types of surfactants will extract the integral proteins from the membranes of these organisms.

Selected members from different structural classes of non-ionic detergents (DM,

OG, Cymal5, Mega9, Tween20) as well as zwitterionic detergents (SDS, Fos-choline,

LDAO) were initially tested for their capacity to extract integral membrane proteins from these different membranes. Extractions from E. coli membranes were carried out from two different strains as well as from two different membrane preparations (total membrane and outer membrane). On the other hand, membrane samples from P. pastoris

primarily came from the total membrane preparations. For both organisms, extraction was performed by incubating equal aliquots of each membrane fraction for a certain period of time with buffered solutions of detergents at twice their CMC levels. The total protein obtained with each surfactant was compared to measure the overall extraction efficiency.

Extraction of the membrane fractions with buffer alone leads to the release of low levels of protein (Table 3.1). The proteins extracted with the buffer control in the absence of detergents are likely to be weakly associated peripheral membrane proteins. The identity of these proteins extracted with the buffer control was verified by our mass spectrometric analysis and we have confirmed the presence of proteins that are peripherally bound to the membrane. The commonly used surfactant SDS, which is widely reported to effectively solubilize protein at the expense of protein structural integrity, displayed the highest total protein extraction efficiency. Protein extraction with

SDS resulted in at least a 6-fold increase in extracted protein relative to the buffer control in OM protein extraction. SDS has also been observed to be 15 times more efficient in solubilizing proteins present in the inner membrane of E. coli. A similar trend in extraction efficiency for this surfactant was seen for the total membrane preparations of

E. coli and P. pastoris resulting to 18-fold and 12-fold increase in extracted protein, respectively (Table 3.1).

Table 3.1. Membrane protein extraction efficient for different surfactants.

Total protein extracted from different membranes Detergent 2X CMC E. coli E. coli E. coli P. pastoris (mM) (XL-1) (C41 IM) (C41 OM) none N/A 5.2 ± 0.4 7.4 ± 1.5 15.0 ± 2.8 3.1 ± 0.6 SDS 20 94.3 ± 3.5 115.3 ± 15.2 89.3 ± 12.4 37.7 ± 4.5 DM 3.2 26.2 ± 2.0 21.5 ± 3.0 29.8 ± 1.3 7.6 ± 1.0 OG 50 78.0 ± 6.8 85.9 ± 6.5 55.1 ± 9.9 18.0 ± 1.0 CYMAL-5 10 29.6 ± 2.1 46.5 ± 9.1 42.8 ± 2.5 10.0 ± 1.1 MEGA-9 50 33.7 ± 3.5 43.1 ± 1.9 60.9 ± 1.6 8.1 ± 0.7 LDAO 4 9.7 ± 0.2 12.2 ± 1.5 26.3 ± 3.7 4.0 ± 0.2 Fos-choline10 22 89.9 ± 4.0 32.2 ± 4.1 29.2 ± 5.7 16.9 ± 2.3 Tween 20 0.12 9.3 ± 0.7 7.8 ± 1.0 28.3 ± 3.8 3.9 ± 0.2

The non-ionic detergents were found to be the most effective of the specialty detergents in extracting proteins from the outer membrane of E. coli. Among the non- ionic detergents tested, Mega 9 showed the highest extraction efficiency (60.9 mg/g), followed by the alkyl sugar detergents OG and Cymal5 (Table 3.1). All of the other detergents exhibited similar but lower capacity in protein extraction. A similar trend was observed in inner membrane extraction where non-ionic detergents gave better extraction efficiencies. However, OG was found to be particularly effective in extracting proteins from the inner membrane as compared with its alkyl sugar counterparts. OG extracted 2-

fold higher levels of protein than Cymal 5 and 4-fold higher levels than DM. Mega 9 which was observed to be the most effective in extracting OM proteins was found to be equally efficient as Cymal 5 in IM protein extraction. The zwitterionic detergent Fos- choline was also found to extract moderate amounts of inner membrane proteins. The extraction efficiencies of these detergents led to between 3-fold to 11-fold higher protein levels as compared with the buffer control. The mild, non-ionic surfactant Tween 20 was the least efficient extraction detergent, with protein levels comparable to, or only slightly higher, than the buffer control (Table 3.1).

3.3.2 Protein Identification

We are particularly interested in determining the extraction efficiencies of representative detergent families in solubilizing constitutive membrane proteins to identify which types of detergents are efficient in extracting either a general class or a specific type of membrane protein. Our starting hypothesis was that protein bands migrating at the same position on each horizontal row would represent the same single protein or possibly the same mixture of proteins. If this was the case, then we could establish a comparison of detergent extraction efficiency for each protein from the various test detergents based on the band intensity at the same position which could be quantified using densitometric analysis.

The bacterial and yeast membrane preparations (total membrane and outer membrane) solubilized in various test detergents were loaded onto an SDS-PAGE gel to separate the constitutive proteins. Initial trials were performed using the total membranes isolated from XL-1 Blue cells to determine the optimal concentration needed for protein

identification in MS analysis. Our results indicate that ~50 μg of total protein loaded on each lane is an optimal concentration that showed intense and well-resolved bands that could be separately excised from the gel for MS sample preparation (Figure 3.1). To further increase the amount of protein needed to improve the signal to noise ratio during protein identification, triplicate gels were run and, if necessary protein bands migrating at similar positions in each of the gel were pooled and considered as one protein entity.

Samples were then processed as previously described (Section 3.2.7) and protein identification was carried out using MALDI-TOF MS analysis coupled with database searching.

Total microgram protein loaded

M 112 56 28 11 5.6 2.8 1.4 0.71 0.35 0.178 M

180 kDa

115 kDa

82 kDa 64 kDa

49 kDa 37 kDa

26 kDa 19 kDa

15 kDa

Figure 3.1 Optimization of protein loading concentrations for protein band isolation.

The acquired experimental peptide masses were compared with the theoretical peptide masses from protein database using the Mascot search engine. This search generates a list of candidate proteins with corresponding peptide matches. Each candidate protein from the PMF search was then scored and this value was compared with the threshold score for the taxonomy indicated in the search. This threshold score is the minimum score required to obtain positive protein identification within a high confidence level (i.e. greater than 95% certainty) (Perkins et al., 1999). If the PMF score did not exceed the allowed threshold, verification of the peptide matches was carried out by MS-

MS fragmentation. The MS-MS spectrum generates fragment ion masses to give information on the complete or partial amino acid sequence of the peptide. The experimental fragment ion masses are then compared with those present in the database.

Identification of a positive peptide match from the MS-MS fragmentation follows similar scoring function as described in the PMF search.

To optimize the search parameters for the database searching, we utilized the band corresponding to the KcsA protein at ~ 64 kDa region (Figure 3.2). This protein migrates as a stable tetramer (Heginbotham et al., 1997) in standard SDS-sample buffer solution used for electrophoresis. We also have previously worked on this protein and we have confirmed this stable tetramer conformation by subjecting purified KcsA samples to

Western blot analysis, thus allowing us to have a good estimate of the position of the

KcsA in the SDS-PAGE gels.

Protein bands that correspond to the tetrameric form of KcsA from each detergent lane and buffer control were analyzed. Using an initial setting of maximum of 5 missed cleavages, significant PMF scores for KcsA were obtained from samples solubilized in

OG, Cymal5 and Fos-choline (Table 3.2). However, varying the settings of this parameter between 0-5 showed different rankings and PMF scores for KcsA in the list of candidate proteins for all of the detergent samples (Appendix A1). In addition to KcsA, we have also observed other candidate proteins (tnaA, atpB) in this band with higher PMF scores than KcsA in the detergent samples. Since we have not purified the protein, and we are using 1D gels with limited resolving power as compared to 2D gels, it is likely that these two proteins (tnaA, atpB) comigrate in a similar position with KcsA in the gel.

To systematically evaluate our preliminary results, we have performed MS/MS analysis of KcsA as well as the other high ranking proteins in the PMF list. From all the different search parameters, we have focused on varying the number of allowed missed cleavages. The use of sequence specific enzymes like trypsin could result in incomplete cleavage which could add another variable to our analysis. Therefore, in our PMF and

MS-MS search we have varied this parameter from 0-5 and evaluated the resulting scores.

Based from our analysis, an allowed missed cleavage of 1 appears to give reliable result as evidenced by significant scores to positively identify the constitutive proteins.

Using the same cleavage setting in the MS/MS analysis further confirmed the identity of the constitutive proteins. All protein identifications using MALDI-TOF MS and MS/MS were carried out following the set parameters and methods described in Section 3.2.7.

Table 3.2 Peptide mass fingerprint analysis for KcsA. Shown in red is the identification of overexpressed KcsA.

Sample Accession Number Mass (Da) Score Expect Matches Protein Name Code

4-142-1 UGPC_ROSDO (1) 39751 55 1.5 5 Sn-glycerol-3-phosphate import ATB-binding protein R.denitrificans Buffer RS19_MYCS5 (2) 10280 49 6.4 4 30S ribosomal protein S19 M.synoviae COXAM_YEAS1 (3) 13316 43 25 4 COX assembly mitochondrial protein S.cerevisiae 4-142-2 ATPA_DESMR (1) 54645 51 2.6 6 ATP synthase subunit alapha D.magneticus SDS RL9_NEIMA (2) 15752 42 22 4 50S Ribosomal protein L9 N.meningitidis serogroup A MUTH_HAEIA (3) 24914 41 24 4 DNA mismatch repair protein mutH H.influenza 4-142-3 RIMM_ALISL (1) 20085 69 0.045 5 Ribosome maturation factor rimM A.salmonicida Triton KCSA_STRCO (6) 17683 52 1.9 5 Voltage-gated potassium channel S.coelicolor KCSA_STRLI (7) 17683 52 1.9 5 Voltage-gated potassium channel S.lividans 4-142-5 KCSA_STRCO (1) 17683 84 0.0013 8 Voltage-gated potassium channel S.coelicolor OG KCSA_STRLI (1) 17683 84 0.0013 8 Voltage-gated potassium channel S.lividans RUVB_PROM (2) 38276 62 0.2 10 Holliday junction ATP-dependent helicase rubV P.marinus 4-142-6 KCSA_STRCO (1) 17683 90 0.00036 8 Voltage-gated potassium channel S.coelicolor Cymal5 KCSA_STRLI (1) 17683 90 0.00036 8 Voltage-gated potassium channel S.lividans DNAA_PROMS (2) 52344 55 1.1 7 Chromosomal replication initiator dnaA P.marinus 4-142-7 QUEC_RICCK (1) 25643 34 120 3 7-cyano-7-deazaguanine synthase R.canadensis Mega9 DDL_RICTY (2) 36089 30 300 3 D-alanine ligase R.typhi YCIN_ECO57 (3) 9380 28 490 2 Protein yciN E.coli )157:H7 4-142-8 RS12_NOSP7 (1) 14482 56 0.83 8 30S ribosomal protein S12 N.punctiforme LDAO RS12_ANASP (2) 14469 43 17 6 30S ribosomal protein S12 A.sp 4-142-9 RPOB_CLOP1 (1) 139044 64 0.12 12 DNA directed RNA polymerase subunit beta C.perfringens Fos KCSA_STRCO (4) 17683 64 0.12 6 Voltage-gated potassium channel S.coelicolor KCSA_STRLI (5) 17683 64 0.12 6 Voltage-gated potassium channel S.lividans

MW Buffer SDS DM OG Cy5 Mega9 LDAO Fos Tween

fdoG fdoG fdoG

180 rne,eftu1,bgal eftu1,bgal odp1 odp1,nuoG nuoG efg,nuoG efg efg,dhsA efg 115 dnaK dnaK dnaK nuoCD nuoCD,dhsA dhsA,ppiD dhsA dhsA dhsA yaeT,dhsA 82 oppA,pckA atpA,glpD atpA atpA,glpD atpA,glpD atpA atpB,tnaA atpB,tnaA kcsA,tnaA,atpB atpB atpB,tnaA atpB,tnaA atpB 64 eftu1 eftu1 rbsB hflK,dacC ahpC cysK kdgK 49 nlpB kdgK,hslO eftS 37 rbsB,atpG kduD,blaT atpG atpG ompP,dhsB dhsB,sodM sodM 26

19 kcsA kcsA,atpF kcsA 15 ibpA

Figure 3.2 Constitutive membrane proteins extracted by test detergents from the total cell membrane of E. coli. Each column represents protein extraction from each detergent type. Constitutive proteins present in each gel band were identified by mass spectrometry.

A total of 35 constitutive were identified from the total membrane fraction isolated from bacterial membranes (XL-1 Blue) using peptide mass fingerprint and MS-

MS analysis, in addition to the recombinantly expressed model alpha-helical membrane protein, KcsA (Appendix A2). Eleven proteins were classified as belonging to the inner membrane (IM), whereas three were found in the periplasmic space, and another three were designated as outer membrane (OM) proteins (Table 3.3). Unexpectedly, fifteen cytoplasmic proteins and three proteins of unknown subcellular localization were also detected in the total membrane samples (Table 3.4). These cytoplasmic proteins were identified to be molecular chaperones and proteins that function in a variety of cellular processes which include biosynthesis of proteins, carbohydrates and small molecules as well as in respiratory and redox reactions (Table 3.4). It is likely that these proteins are associated with the membrane or with integral membrane proteins and they were not fully eliminated during the washing step done for the membrane pellet prior to detergent solubilization.

The presence of cytoplasmic proteins identified from cell membrane preparations is not completely unexpected. A proteomic analysis of the cell envelope fraction of E. coli carried out by Fountoulakis & Gasser (2003) reported the presence of 394 different gene products, 25% of which were annotated as membrane proteins, 20% as cytosolic and no annotated subcellular localization were reported for the remaining 55% of the proteins during the time of their study.

Table 3.3 Summary of identified membrane proteins from total membrane fractions.

Protein Protein Name Mr Localization Function ID (kDa) atpA ATP synthase subunit alpha 55420 Inner Membrane ATP synthesis coupled proton transport atpB ATP synthase subunit beta 50352 Inner Membrane ATP synthesis coupled proton transport atpF ATP synthase F(0) sector subunit b 17264 Inner Membrane ATP synthesis coupled proton transport atpG ATP synthase gamma chain 31673 Inner Membrane ATP synthesis coupled proton transport dacC D-alanyl-D-alanine-carboxypeptidase 43639 Inner Membrane Cell wall biogenesis dhsA Succinate dehydrogenase flavoprotein subunit 65019 Inner Membrane Electron transport chain dhsB Succinate dehydrogenase iron-sulfur subunit 27390 Inner Membrane Electron transport chain hflK Modulator of FtsH protease HflK 45517 Inner Membrane Inner membrane biogenesis kcsA Voltage-gated potassium channel 17694 Inner Membrane Ion transport nuoG NADH-quinone oxidoreductase subunit G 101226 Inner Membrane Electron transport nuoCD NADH-quinone oxidoreductase subunit C/D 68425 Inner Membrane Electron transport ppiD Peptidyl-prolyl cis-trans isomerase D 68108 Inner Membrane Folding of OM proteins fdoG Formate dehydrogenase-O major subunit 113291 Periplasm Anaerobic respiration oppA Periplasmic oligopeptide-binding protein 60977 Periplasm Peptide/Protein transport rbsB D-ribose-binding periplasmic protein 30931 Periplasm Sugar transport nlpB Lipoprotein 34 36842 Outer Membrane Outer membrane assembly ompP Outer membrane protease ompP 35477 Outer Membrane Protease yaeT Outer membrane protein assembly factor 90612 Outer Membrane Outer membrane assembly

Table 3.4 Summary of cytoplasmic proteins detected from total and outer membrane* fractions

Protein ID Protein Name Mr (kDa) Localization Function ahpC Alkyl hydroperoxide reductase subunit C 20864 Cytoplasm Redox reactions bgal Beta-galactosidase 117366 Cytoplasm Lactose catabolic process ch601* 60 kDa chaperonin 57467 Cytoplasm Protein folding cysK Cysteine synthase 34526 Cytoplasm Cysteine biosythesis dnaK Chaperone protein 69130 Cytoplasm Protein folding dps DNA protection during starvation protein 18,695 Cytoplasm Redox efg Elongation factor G 77581 Cytoplasm Protein biosynthesis efts Elongation factor Ts 30520 Cytoplasm Protein biosynthesis eftu1 Elongation factor Tu 1 43430 Cytoplasm Protein biosynthesis glpD Aerobic glycerol-3-phosphate dehydrogenase 56889 Cytoplasm Glycerol metabolism hslo 33 kDa chaperonin 32862 Cytoplasm Chaperone ibpA Small heat shock protein 15744 Cytoplasm Chaperone odp1 Pyruvate dehydrogenase E1 component 99954 Cytoplasm Glycolysis pckA Phosphoenolpyruvate carboxylase 59868 Cytoplasm Carbohydrate biosynthesis rne Ribonuclease E 118357 Cytoplasm RNA binding rs1* 30S ribosomal protein S1 61,237 Ribosome Ribosomal protein rl5* 50S ribosomal protein L5 20347 Ribosome Ribosomal protein rl9* 50S ribosomal protein L5 15769 Ribosome Ribosomal protein sodm Superoxide dismutase 23065 Cytoplasm Redox tnaA Tryptophanase 53146 Cytoplasm Trytophan catabolism blaT Beta-lactamase TEM 31669 Unknown Antibiotic resistance kdgK 2-dehydro-3-deoxygluconokinase 34231 Unknown Redox reaction kduD 2-dehydro-3-deoxy-D-gluconate 5-dehydrogenase 27169 Unknown Redox reactions ydiu* UPF0061 protein ydiU 54,382 Unknown Uncharacterized Yiba* Protein yibA 31,874 Unknown Uncharacterized

From the outer membrane fraction, we identified 11 constitutive OMPs which function either in ion or lipid transport, or in assembly of outer membranes in the cell such as in maintaining the bacterial envelope integrity (Table 3.5, Appendix A3). An additional six IMPs are also found in the outer membrane samples, some of which (ATP synthase subunits, NADH-quinone oxidoreductase) have been previously found in the total membrane fraction. Ribosomal and chaperone proteins are also present in the buffer control (Table 3.4).

Our overall goal is to selectively isolate different classes of membrane proteins by efficient detergent extractions. For both cell membrane preparations, a total of 14

(including KcsA) unique inner membrane proteins were extracted in varying combinations by the test detergents (Table 3.6) and positively identified by mass spectrometric experiments (Appendix A2 to A3). Thirteen outer membrane proteins and three periplasmic proteins were also isolated. The different classes of membrane proteins, their extraction efficiencies and the behavior of the protein-detergent complexes will be discussed in the following sections. Our results from the constitutive protein identification from the total and outer cell membrane were combined to give a concise and clear summary of the different protein types extracted by various detergents (Table

3.6).

Table 3.5 Summary of identified membrane proteins from outer membrane fractions.

Protein Protein Name Mr (kDa) Localization Function ID acrA Acriflavine resistance protein A 42229 Inner Membrane Drug efflux Lipid-anchor atpA* ATP synthase subunit alpha 55420 Inner Membrane ATP synthesis coupled proton transport atpB* ATP synthase subunit beta 50352 Inner Membrane ATP synthesis coupled proton transport atpG* ATP synthase gamma chain 31673 Inner Membrane ATP synthesis coupled proton transport ftsH ATP-dependent zinc metalloprotease 70763 Inner Membrane Protein catabolic process nuoG* NADH-quinone oxidoreductase subunit G 101226 Inner Membrane Electron transport fadL Long-chain fatty acid transport protein 48512 Outer Membrane Lipid transport lptD LPS-assembly protein 89836 Outer Membrane Outer membrane assembly nlpB* Lipoprotein 34 36842 Outer Membrane Outer membrane assembly ompA Outer membrane protein A 37294 Outer Membrane Porin, Ion transport ompF Outer membrane protein F 39309 Outer Membrane Porin, Ion transport ompX Outer membrane protein X 18649 Outer Membrane Porin, Ion transport pal Peptidoglycan-associated lipoprotein 18879 Outer Membrane Bacterial envelope integrity tolC Outer membrane protein 53708 Outer Membrane Transport tsx Nucleoside-specific channel-forming protein 33568 Outer Membrane Transport yfiO UPF0169 lipoprotein 27870 Outer Membrane Outer membrane assembly Lipid-anchor ybjP Uncharacterized lipoprotein 19211 Outer Membrane Not mentioned Lipid-anchor * Also identified in the total membrane extractions

Table 3.6 Summary of constitutive membrane proteins extracted by test detergents from total and outer membranes.

Protein Types Buffer SDS DM OG Cymal5 Mega9 LDAO Fos Tween A. Inner Membrane Proteins Model Protein (KcsA) kcsA kcsA kcsA Bioenergetic Complexes ATP synthase atpA atpA atpA atpA atpA atpA atpA atpA atpB atpB atpB atpB atpB atpB atpB atpB atpB atpG atpG atpG atpF

Succinate dehydrogenase dhsA dhsA dhsA dhsA dhsA dhsB dhsB NADH-quinone reductase nuoG nuoG nuoG nuoCD nuoCD Other Functions Cell wall biogenesis dacc Inner membrane biogenesis hflk Folding of OM proteins ppiD Drug efflux acrA acrA acrA Protein Catabolism ftsH B. Periplasmic Proteins Respiratory Complex fdoG fdoG fdoG Transport rbsB rbsB oppA C. Outer Membrane Proteins Porins ompA ompA ompA ompA ompA ompF ompF ompF ompF ompF ompX ompX ompX ompX Transport tolC tolC tsx fadL Assembly and integrity yfiO nlpB nlpB lptD pal yaet ybjp Protease ompP

3.3.3 Constitutive Inner Membrane Proteins

3.3.3.1 Bioenergetic Complexes

The most common type of constitutive inner membrane proteins isolated from our extracted samples belongs to protein complexes that play roles in bioenergetic processes.

Protein subunits belonging to the family of ATP synthase, succinate dehydrogenase and

NADH-quinone reductase were identified. Other inner membrane proteins were also isolated that perform functions in various cellular processes (Table 3.3).

The ATP synthase (also called as F1F0-type ATPase, Complex V) is a large enzyme complex that is responsible for ATP production. The complex is made up of eight subunits in prokaryotes and 16-18 subunits in mammals, with a total molecular weight ranging from 550 kDa to 650 kDa. The ATP synthase is divided into two subcomplexes, namely the F1 catalytic core containing five (α3, β3, γ, δ, ε), and the F0 membrane proton channel composed of three subunits (a1, b2, c9-12), (Capaldi & Aggeler, 2002).

A total of four subunits of this complex were detected from the detergent- solubilized samples. We were able to identify the α-subunit (55 kDa, atpA), the β-subunit

(50 kDa, atpB), the γ-subunit (31 kDa, atpG) and the subunit b (17 kDa, atpF) from membrane preparations solubilized in Fos-choline. Three subunits (atpA, atpB, atpG) were detected from Cymal 5 and LDAO, whereas two subunits (atpA, atpB) were positively confirmed from the six remaining sample preparations including the buffer control (Table 3.6).

All four of these subunits migrate at a relatively higher position compared with their expected Mr position based on the protein markers (Figure 3.1 and 3.2). However, the migrations of each subunit relative to each other are within a reasonable range. For

example, atpA (55 kDa) and atpB (50 kDa) detected from the total membrane preparation migrate closely to each other (Figure 3.1, lanes Buffer, OG, Cymal5, LDAO, Fos-choline) but above the 64 kDa marker. The other subunits atpG (31 kDa) and atpF (17 kDa) were also seen at higher apparent Mr (Figure 3.1, lanes Cymal5, LDAO, Fos-choline), but both of them travel at a lower Mr range as compared to the larger α- and β- subunits.

In the outer membrane detergent preparation, atpA was found in a position twice its expected monomeric weight of 55 kDa from membranes solubilized in SDS, DM and

Cymal 5 (Figure 3.2). The ATP synthase F1 complex contains three alpha subunits, and the identification of atpA in this position may likely indicate the presence of an undissociated dimeric form of the α-subunit. In these three detergent preparations (SDS,

DM, Cymal 5), the α-subunit co-migrates with the trimeric form of the outer membrane protein, ompF, which was also positively identified by mass spectrometry from the same band. It seems like the maltoside based detergents (DM and Cymal5) can extract the monomeric and dimeric forms of atpA, as well as an intact trimeric form of ompF. The behavior of the ompF protein in the different test detergents will be discussed below.

One interesting observation about the behavior of atpA was seen from a comparison of the total membrane and outer membrane preparations from Cymal-5.

Using the same Cymal-5 detergent for solubilization, atpA was completely dissociated from the total membrane preparation (Figure 3.1, Cymal-5 lane), whereas a dimeric form was detected from the outer membrane preparation. It also seems that the presence of dimeric form of atpA is influenced by the presence of ompF. In all of the detergent extractions where dimeric atpA was detected, these samples also contained various forms of ompF. Whether the presence of undissociated atpA dimer is because of the milder

extraction capabilities of the maltoside detergents or is due to possible interaction between atpA and ompF needs further studies.

Another multisubunit protein complex identified is the succinate dehydrogenase

(complex II or succinate:ubiquinone oxidoreductase) which is a functional part of the

Krebs cycle and aerobic respiratory chain. Its crystal structure reveals a heterotetramer complex (dhsABCD, also annotated as SdhABCD) composed of the hydrophilic subunits flavoprotein (dhsA, 65 kDa) and iron-sulfur protein (dhsB, 27 kDa). The other subunits

(dhsC, 14 kDa and dhsD, 15 kDa) are heme-containing proteins that are anchored to the membrane and serve as binding site for ubiquinone (Cecchini, et al., 2003). The crystal structure reveals an overall structure of succinate dehydrogenase as trimer with a total molecular weight of 360 kDa (Yankovskaya et al., 2003).

Only the hydrophilic subunits dhsA and dhsB were found to be extracted in the total membrane fractions. Both subunits were found from samples extracted with OG and

Mega 9, whereas only the flavoprotein subunit was identified from membranes extracted with SDS, Cymal5 and Fos-choline (Table 3.6, Figure 3.1). Each of these subunits is extracted in detergents having at least 10 mM concentrations.

The NADH-quinone reductase (Complex 1) is involved in proton translocation across the mitochondrial inner membrane and in the plasma membrane of prokaryotes.

This complex is made up of 14 subunits that carry redox centers, a flavin mononucleotide and nine iron-sulfur clusters in bacterial organisms. A more intricate form of this respiratory chain complex is present in mitochondria, where an additional 32 subunits are found leading to a very large molecular mass extending to 1000 kDa (Brandt, 2006). Out of the 14 subunits of Complex 1 in bacteria, we identified subunit G (nuoG, 101 kDa) and

subunit CD (nuOCD, 68 kDa) from three detergent preparations (SDS, LDAO, Tween), as well as in the buffer control. It seems that these intact subunits are efficiently extracted by detergents with low CMC values (Table 3.1, Table 3.6).

3.3.2.2 Inner Membrane Proteins with other Functions

A number of inner membrane proteins participating in various cellular fractions have also been determined. Three additional inner membranes were isolated from OG- solubilized fractions. The inner membrane anchored parvulin-like peptidyl-prolyl isomerase D (ppiD, 68 kDa) was resolved in the gel occupying the same position with dhsA (65 kDa) in the OG lane (Figure 3.1), indicative of the presence of the monomeric form of ppiD. Our result is in contrast with those obtained by the group of Stenberg, where ppiD dimer and trimers were detected in 2D gels. However, no other indications of the oligomeric forms of ppiD have been reported (Stenberg et al., 2005).

PpiD was previously described to be similar to surA protein, where the latter protein is known to exhibit both peptidyl-prolyl-isomerase and chaperone activities that are essential in OMP maturation. SurA reportedly facilitates protein folding thru substrate binding and recognition to sequences that are characteristic of OMPs (Hennecke et al.,

2005). Recent studies on ppiD showed the presence of a well-folded parvulin domain, albeit deficient in catalytic activity for prolyl-isomerase (Weininger et al., 2010). On the other hand, ppiD displayed chaperone activity and was found to participate in a system of chaperone activities in the cell to aid periplasmic folding of newly translocated proteins.

(Matern, et al., 2010).

In addition to ppiD, octylglucoside was also effective in solubilizing two other

IMPs, dacC and hflk, both of which were detected from the same gel band. DacC (D- alanyl-D-alanine carboxypeptidase, also known as penicillin binding protein) is a membrane-bound 44 kDa protease that cleaves terminal alanine residues from sugar- peptide cell wall precursors (Waxman, et al., 1982). On the other hand, hflK (46 kDa) is a protein that interacts with another protein hflC to form the complex hflKC. This hflKC complex regulates the proteolytic activity of a membrane-bound ATPase (called ftsH) against uncomplexed forms of the SecY translocase subunit (Kihara, et al., 1996). The hetero-multimeric complex hflKC interacts with the hexameric form of ftsH to form a large FtsH holoenzyme (FtsH)6(HflKC)6 of about 1 MDa size (Akiyama, 2009). We have also isolated the ftsH protein in samples solubilized in milder detergent, Tween 20

(Figure 3.2).

We have also identified acriflavine resistance protein A (acrA, 42 kDa) from outer membrane fractions isolated using DM, Cymal5 and Fos-choline detergents. AcrA is a component of the major multidrug efflux AcrAB-TolC which is proposed to traverse the inner and outer membranes. This stable complex is responsible for the extrusion of antibiotics, detergents, dyes and organic solvents (Ma et al., 1993; Tikhonova et al.,

2004).

3.3.4 Constitutive Periplasmic Proteins

Three periplasmic proteins were detected in our extraction samples. Formate dehydrogenase-O major subunit (fdoG) is a component of formate dehydrogenase, FDH-

O, which functions in anaerobic respiratory pathway in E. coli cells. This large subunit fdoG (113 kDa) was successfully extracted from OG, Mega9 and Fos-choline from total membrane preparations, and protein bands were resolved in the high molecular weight region of the gel. Subunit fdoG together with fdoH (33 kDa) and fdoI (25 kDa) forms the catalytic complex of FDH-O isoenzyme. FdoG contains the catalytic site, whereas the fdoH and fdoI have been determined to be the electron transfer unit and the cytochrome b apoprotein, respectively. (Benoit et al., 1998).

LDAO was found to be effective in solubilizing the periplasmic oligopeptide binding protein (oppA, 61 kDa), a constituent of the protein transport system oligopeptide permease (opP). Oligopeptide permease is involved in peptide uptake that is to be used as carbon and nitrogen sources for the bacteria (Sleigh et al., 1997). Another transport protein that we have identified is the periplasmic D-ribose-binding protein (rbsB, 31 kDa). It primarily functions in high affinity transport of ribose as well as a bacterial receptor for bacterial chemotaxis (Kim, 1992). This protein was detected from Cymal5 and Mega9 solubilized membranes.

3.3.5 Constitutive Outer Membrane Proteins

3.3.5.1 Porins

Eleven outer membranes proteins were obtained and confirmed by mass spectrometric data from the various detergent preparations. The major outer membrane protein ompF (outer membrane F, 39 kDa) was detected in addition to ompA (outer membrane A, 37 kDa) from five membrane-detergent preparations. ompF is a beta-barrel type trimeric membrane protein with 16 β-strands comprising each monomer (Cowan, et al., 1992), whereas ompA is a monomer with 8 amphiphatic β-strands (Vogel & Jahnig,

1986). Both proteins form water-filled channels that allow passive diffusion of solutes across the other membrane (Schirmer, 1998).

OmpF was effectively solubilized by the ionic surfactant SDS, in which we have detected the ompF trimer (Figure 3.2, SDS lane, ~115 kDa region). PMF analysis of the clusters of intense gel bands between 64 kDa to 115 kDa also identified ompF either as single protein or as a mixture with other membrane proteins. A faint gel band at the 49 kDa region was also found to contain ompF. The presence of ompF in multiple gel bands were also seen from outer membrane samples extracted from non-ionic maltoside-based detergents (DM and Cymal5). OmpF was detected from five gel bands between 82 to

180 kDa marker from DM-solubilized membranes, while this same protein was confirmed from three bands from the Cymal5 lane (82 to 115 kDa). On the other hand, two other non-ionic detergents were efficient in extracting ompF as evidenced by the very intense bands at 64 kDa in the OG and Mega 9 lanes. A faint band in the 49 kDa region of the OG lane was also attributed to ompF.

MW Buffer SDS DM OG Cy5 Mega9 LDAO Fos Tween

trag, pet

tolC pcoa

180 ompF

ompF hcad ispe nuog ispe , ftsh atpA, ompF atpA, ompF atpA, ompF 115 Ispe, ompF ompF rs1 ompF ompF ftsh ompF ch601 ompF, lptd ompF, atpB, tolC yaet 82 atpA atpB atpB atpB atpA, atpB ompF atpA, atpB 64 ompF ompF acra acra acra cvaa efts nuoF cvaa ompF nlpB ompF, nlpB rho yiba, nlpB 49 Prim, Eftu ompA ompA ompA ompA atpG ompA 37 ompA atpG yhgf, mscm, atpg fadL arnd fadL fadL Nohb , fadl mepa metn, pal yfio 26 ybjp, pal dhsb dps rl5 tsx dnaT 19 trbg rl9 ompX ydiu ompX ompX 15 ompx, ybjp cvaa mbea citd yeea

Figure 3.3 Constitutive membrane proteins extracted by test detergents from the outer membrane of E. coli. Highlighted in red are proteins that need further confirmation.

The presence of the ompF trimers in the gel profile is quite expected because it is the functional form of this protein. However, we were intrigued by the detection of ompF in multiple bands in the gels, thus we tried to compare our data with the literature. The group of Stenberg (2005) identified the ompF trimer from the E. coli cell envelope using

2D gels. Aside from the trimer, other ompF forms were also determined and these were attributed to be the result of smearing and partial denaturation with SDS (Stenberg et al.,

2005). A similar scenario could be applied to our system due to the smearing because of lower resolving power of 1D gels.

Earlier studies on the different conformations of ompF with immunological probes of samples solubilized at different temperatures have also been done. The ompF protein has been found to exist in five major products with apparent Mr values of 95, 105,

112, 120 and 190 kDa (Pages & Bolla, 1988), and the relative distributions of these forms were found to be dependent on the temperature of solubilization. Treatment at 95°C dissociated the oligomers into monomeric ompF units, whereas heating between 20-56°C showed this range of various forms with different relative amounts (Pages & Bolla,

1988). In our case, membrane protein solubilization was carried out at 4°C while samples for electrophoresis were prepared at room temperature. The presence of ompF in multiple gel bands could be attributed to the presence of different oligomeric forms that are partially dissociated from the functional trimer, and that they could exist in equilibrium with each other. The aberrant migration and multiple banding patterns of ompF trimers in

SDS-PAGE gels have also been reported to be associated with the interaction of lipopolysacharides, (LPS) (Diedrich et al., 1990). It has also been found that monomer ompF could still retain bound LPS (Rocque et al., 1987). The ompF protein that we have

isolated may have co-purified with the LPS present in the bacterial membranes.

The above reasons could explain the migration behavior and multiple banding patterns of ompF in all the five-detergent fractions where ompF was detected. However, we also found some interesting differences between the profiles of the group of SDS, DM and Cymal5 with that of OG and Mega lanes. Multiple banding patterns have been observed for the first set of detergents, while the last two only showed a major band at 64 kDa. These results could imply that OG and Mega 9 may be efficient in extracting ompF, however it may cause complete dissociation of the functional trimer to its monomeric units. On the other hand, the presence of ompF trimers and other multiple bands associated with ompF suggests that DM, Cymal 5 and even SDS are efficient towards ompF extraction. The clusters of bands may be indicative of the folding process of ompF monomers into dimers, and then to trimeric ompF.

In a study done by Surrey et al., (1996), urea dissociated ompF was subjected to refolding experiments in the presence of lipid vesicles and detergents. Refolding of ompF was analyzed by Western blot, conductance measurements, as well as fluorescence and circular dichroism spectroscopy. In the presence of lipid vesicles alone, refolding of ompF to trimers was established to be slow with different distributions of oligomers in the gel. The addition of dodecylmaltoside detergent (DDM) significantly improved the yield of ompF trimers, and this was proposed to be due to perturbation of the structure

(Surrey et al., 1996). DDM detergent belongs to the same family of maltoside-based detergents DM and Cymal5 that were used in this study. It is likely that DM and Cymal5 behaves in a similar manner to DDM to promote refolding of ompF, however additional characterization of ompF in the presence of DM and Cymal5 could be carried out to

further confirm such behavior.

The five detergent families that effectively solubilized ompF, were also useful for the extraction of another major outer membrane protein, ompA. The positive control SDS was the most efficient in ompA extraction as evidenced by the very intense band migrating at 37 kDa region. The non-ionic detergents OG, Mega9, Cymal5 and DM are also capable of solubilizing ompA, albeit in varying degrees based on the band intensity.

In contrast with the ompF protein, ompA was isolated as a single band consistent with its reported functional monomeric form (Pautsch & Schulz, 1998), though there was a reported example when ompA was detected as dimers (Stenberg et al., 2005). No other oligomeric form of ompA was resolved in our samples.

The presence of a smaller outer membrane protein ompX (19 kDa) was also verified from SDS, OG, Cymal5 and Mega 9 detergent-OM preparations. This protein migrates at the expected Mr region but the bands are skewed as it reached the end of the gel, probably due to the voltage settings for the electrophoretic runs. This 8-stranded beta barrel ompX protein has been determined to have functions associated with bacterial adhesion and entry to mammalian cells (Vogt & Schulz, 1999).

3.3.5.2 Transport OMPs

The acrA protein, one of the subunits in the acrAB-TolC drug efflux complex has been identified from the outer membrane fractions. Though we were not able to detect acrB in our extraction samples, the other interaction partner TolC was isolated from DM and SDS membrane extractions. A high Mr band (~ 180 kDa) was resolved from the DM lane and we confirmed this as the functional TolC trimer (Koronakis, 2003). In contrast, tolC was identified with a gel band migrating at 82 kDa region in the SDS lane, co- migrating with ompF and atpB. Our results suggest that while decyl maltoside (DM) was able to extract the two components of the acrAB-TolC system, only acrA was detected from Cymal 5 and Fos-choline preparations, while SDS is effective in extracting TolC but no acrA was detected from SDS-soluble membranes (Table 3.6)

The nucleoside-specific channel protein, TsX (33.5 kDa) migrated with an apparent Mr of ~21 kDa in the Cymal5 lane. The long chain fatty acid transporter, fadL

(48.5 kDa) was thus resolved in a lower apparent Mr compared to its expected size. The presence of this fatty acid transported was confirmed from both SDS and DM membrane preparations. Protein bands from other detergents lanes (OG, Cymal5, Mega9) migrating at similar position with respect to fadL from SDS was also subjected to PMF identification, however low scores were obtained that did not allow us to fully designate these bands as fadL proteins. Further MS-MS fragmentation also showed very low intensity ion peaks, insufficient for direct identification. It may be necessary to perform reanalysis of these bands from these three detergent preparations for confirmation.

3.3.5.3 OMPs Involved in Biogenesis

Members of a protein complex machinery participating in outer membrane biogenesis were also identified in our samples. This complex consists of a beta-barrel protein (yaeT, 90 kDa), and three outer membrane lipoproteins ygfL, yfiO and nlpB (Wu et al., 2005). We identified three of the components, except for yfgL. YaeT was obtained from OG-soluble membranes, nlpB (37 kDa) from OG and DM extracted membranes, while SDS solubilized yfiO (28 kDa).

Two other lipoproteins pal (19 kDa) and ybjP (19 kDa) co-exist in a gel band isolated from Mega9-solubilized membranes. In addition, the LPS-assembly protein, lptD

(90 kDa) was detected from Cymal5 samples, co-migrating with ompF. However, the interaction partner of lptD to form the lptD/lptE complex was not identified. The lptD/lptE complex together with periplasmic lptA protein, and IM-associate ATP binding cassette proteins (lptB and lptC) are involved in the LPS assembly for transport of LPS between the inner and outer membranes (Sperandeo et al., 2008).

The outer membrane protease, ompP, was identified from a gel band isolated from the OG lane. This protein was resolved with a lower apparent Mr in comparison with its expected size of 35 kDa. The gel band containing this protease was also found to contain subunit B of succinate dehydrogenase, dhsB (27 kDa) which is close to its apparent Mr value. OmpP is analogous to the ompT protease, with both proteins having 71% sequence identity, though they slightly differ with the preferred specificity of ompP towards lysine at the P1 or P1’ site (Hwang et al., 2007).

3.4 Discussion

Our study is focused on determining the efficiencies of different detergents in the solubilization of constitutive membrane proteins from bacterial and yeast membranes.

Using mass spectrometry, we were able to identify constitutive membrane proteins from bacterial membranes. However, we were not successful in our initial attempts to identify membrane proteins from yeast membranes. The yeast membrane may require other methods of membrane isolation (i.e. successive sucrose gradients) that were previously used to obtain membrane proteins from fungal species (Zahedi et al., 2006; Schafer et al.,

2001).

In addition to the overexpressed potassium channel protein KcsA, we were able to identify 29 membrane proteins, 13 of which are associated to the inner membrane, another 13 from the outer membrane and 3 are located in the periplasmic space (Table

3.6). The majority of the identified inner membrane proteins are protein subunits of multisubunit complexes that play roles in respiratory processes in the cells. On the other hand, the most common protein types extracted from the outer membrane are the porins and transporters involved in OM biogenesis.

The protein components of the E. coli cell envelope (outer membrane, periplasm/peptidoglycan layer, inner membrane) have been extensively analyzed in various studies (reviewed by Weiner & Li, 2008). The set of identified membrane proteins from our study has also been detected from analysis of bacterial membranes using various methodologies (Molloy et al., 2000; Molloy et al., 2001; Fountoulakis &

Gassler, 2003; Gevaert et al., 2002; Lai et al., 2004; Stenberg et al., 2005). For example,

40 unique proteins (which include 21 OMPs, 5 lipoprotein, 3 IMPs) were identified from

the work of Molloy (2000), while a total of 394 different gene products (24 OMPs) were determined by Fountoulakis & Gassler (2003).

Based on our analysis, we were able to confirm the presence of a reasonable sampling of membrane proteins, given the limited resolving power of the 1D gel for separation of the components of the detergent solubilized membranes. The number of detected proteins could be further improved by running a 2D gel for better separation, however this may introduce complications between the ampholytes needed for this method and the detergents tested for this study. Separation on the 1D gel could also be enhanced by the use of bigger gel slabs and a wider range of Mr markers. Alternatively, excised gel bands that may contain protein mixtures could be further separated by LC coupled with MS/MS for protein identification. Peptide separation by LC will decrease ion suppression effect, thus leading to identification of a larger set of peptides from the mixture (Wu & Yates, 2003).

The potassium channel KcsA was overexpressed in E. coli cells and initially served as our test protein for the optimization of our experimental protocols. We expected to detect the presence of KcsA in all the detergent fractions, however KcsA was only positively identified as a tetramer from OG preparations and as a dissociated monomer from SDS and Fos-choline solubilized membrane fractions. KcsA was not conclusively detected from other detergent fractions when using a narrow range of allowed missed cleavages for trypsin. The lack of KcsA identification from the other fractions could be due to the hydrophobic nature of KcsA, and of membrane proteins in general, resulting in limited accessibility of trypsin to cleave its target basic residues.

The limited number of identified proteins from our samples could be caused by various reasons. First, proteins co-migrate in the gel bands leading to protein mixtures and may result in detection of only the most abundant protein. Second, the presence of detergent micelles surrounding the protein of interest can hinder access of the enzyme for proteolysis. The protease to sample ratio and the incubation time should also be evaluated to see if varying the time and ratio will improve detection.

Though we were not able to identify the overexpressed KcsA bands in every detergent extraction, our results still identified a significant number of isolated membrane proteins. Our work provides a platform for testing detergents that could be used for the extraction of target membrane proteins. This initial set of surfactants could be expanded to select the optimal detergent for use with a specific protein type. From our results, we were able to establish detergents that are effective in extracting different sets of proteins and these different detergent profiles for membrane protein extraction are discussed below.

3.4.1 Detergent Profiles for Total Membrane Extraction

Among the tested detergents, the non-ionic glucoside-based surfactant (OG) was the most efficient in extracting the largest number of constitutive and over-expressed membrane proteins, with a total of 15 various protein types that perform functions in different subcellular locations (Table 3.6). OG was effective in the extraction of protein subunits (atpA, atpB, dhsA, dhsB, fdoG) from three respiratory complexes, and was also specifically successful in the extraction of three other inner membrane proteins with specialized functional roles (hflK, dacC, ppiD). This non-ionic detergent is also equally efficient in the solubilization of proteins located in the outer membrane. Proteins

belonging to the porin family as well as OMPs with roles in outer membrane assembly and proteolysis were identified from OG-solubilized membranes. In contrast, OG was not as effective in the extraction of other types of membrane transport proteins, aside from the over-expressed potassium channel, KcsA.

Two other non-ionic detergents also exhibited good extraction efficiencies towards various classes of membrane proteins. A total of eleven membrane proteins were detected from Cymal5 membrane preparation, whereas Mega9 extracted ten different protein types. These numbers are comparable with the extraction efficiency of the commonly used ionic surfactant SDS, which served as our positive control for membrane protein extraction. The zwitterionic detergent Fos-choline and the alkyl-maltoside detergent DM showed similar total membrane extraction efficiencies. On the other hand,

LDAO and Tween 20 were the least efficient in total membrane extractions, only capable of solubilizing a few inner and periplasmic membrane proteins.

3.4.2 Detergent Profiles for Inner and Periplasmic Membrane Protein Extraction

For the extraction of inner membrane proteins, OG and Fos-choline were found to be the most effective and have comparable extraction efficiencies. However, Fos-choline is particularly useful in the extraction of the multisubunit ATP synthase respiratory complex, where four protein subunits were detected from Fos-choline membrane samples. Three subunits of the ATP synthase complex were detected from LDAO and

Cymal5, while all the other detergents exhibited lower extraction profiles, with either one or two subunits confirmed from membrane fractions (Table 3.6).

Other proteins belonging to major bacterial respiratory complexes were also detected in the detergent-solubilized membranes. Two protein subunits (dhsA and dhsB) from succinate dehydrogenase complex are best extracted using the detergents OG and

Mega9. Aside from these two detergents, three other surfactants (SDS, Cymal5 and Fos- choline) could be used to specifically extract the larger dhsA subunit. We have also confirmed the presence of the subunits of the NADH-quinone reductase complex, and our results suggest that detergents with lower CMC values (LDAO, Tween20) appears to be sufficient to remove the high Mr subunit nuoG subunit from the inner membrane. On the other hand, SDS is effective in the extraction of the 68 kDa subunit nuoCD. Interestingly, these two subunits were also detected from the buffer control. These findings may indicate that the use of milder detergents and low detergent concentrations could already be adequate to solubilize this type of inner membrane protein, though we have to verify this with other low CMC detergents.

Inner membrane proteins that are involved in membrane protein and cell wall biogenesis are solely extracted by OG, whereas the acrA protein from the drug efflux acrAB-TolC complex are equally extracted by maltoside-based detergents (DM and

Cymal5) as well as the zwitterionic Fos-choline. On the other hand, ftSH protein involved in protein catabolism was only detected with Tween 20 extraction.

Three proteins residing in the periplasmic space between the inner and outer membranes of E. coli were also identified from extraction samples. The large alpha subunit of formate dehydrogenase (fdoG) was similarly solubilized by OG, Fos-choline and Mega9. In addition, Mega9 extracted the transport protein, rbsB, which was not detected from OG and Fos-choline membrane preparations. RbsB was also efficiently

extracted by Cymal5, while LDAO was the only detergent that appears to solubilize the periplasmic oligopeptide binding protein. Other detergents such as SDS, DM and Tween were unsuccessful in extracting periplasmic proteins.

3.4.3 Detergent Profiles for Outer Membrane Protein Extraction

Proteins from the outer membrane are best solubilized in the non-ionic detergent,

OG, as well as the ionic surfactant SDS, with each detergent extracting a total of six different protein types. These two detergents are capable of extracting proteins from the porin family (ompA, ompF, ompX), as well as other OM proteins which are exclusively obtained from each detergent extraction. For example, yaeT and ompP were selectively extracted from membranes by OG, whereas the long chain fatty acid protein (fadL) and the lipoprotein yfiO were detected only from SDS-soluble membranes. In addition, SDS was capable of extracting the drug-efflux protein tolC, but this same protein was not detected from OG-soluble membranes. Instead, tolC was found to be well-extracted by the alkyl maltoside detergent along with its interaction partner acrA. Other membrane protein types found from DM extractions include two porins, ompA and ompF, as well as a lipoprotein (nlpB) which was also efficiently extracted by OG.

Two other non-ionic detergents had equal extraction efficiencies with OG and

SDS detergents in terms of extraction of OM proteins belonging to the porin family. Both

Cymal5 and Mega9 extracted ompA, ompF and ompX, all of which were also detected from OG and SDS extractions. In addition Cymal5 was the only detergent identified to effectively solubilize the nucleoside-specific channel (tsx) and the LPS-assembly protein,

100

lptD. Two other lipoproteins were detected from our samples and these two proteins were solely extracted from Mega9-soluble fractions.

3.4.4 Possible Insights on Protein-Detergent Interactions

From the detergent extraction studies, we were also able to obtain some insights on possible protein-detergent interactions that may have an effect on the oligomeric structure and functional integrity of membrane proteins. In general, we have observed differences in the electrophoretic mobility of the same protein in the presence of different detergents. It is known that the mobility of the membrane proteins in comparison with soluble proteins would be different due to several factors. During extraction, membrane proteins are solubilized in the presence of detergents and the binding of detergent molecules have been reported to be the cause of the anomalous migration of these proteins in the SDS-PAGE gel (Rath et al., 2006). This aberrant migration patterns were also seen in our gel profiles where identified proteins have apparent Mr values that are different from their expected size. For example, the four subunits of the ATP synthase migrate at a relatively higher position than their expected size based on the protein markers (Table 3.3., Figure 3.2). In addition, the migration pattern of specific proteins such as ompF, ompA and ompX were different in the presence of different specific detergents (Figure 3.3). These inconsistent mobilities could be attributed to different numbers of detergent molecules that are bound to the membrane protein. Each detergent has specific aggregation numbers, where this number represents the average number of detergent molecules present within a micelle. In a protein-detergent complex, the detergent molecules will associate with the membrane protein and the number of detergent molecules bound could be different for each type of detergent. Thus, we have

101

seen these discrepancies in the migration patterns of the same protein in the presence of various test detergents. Other factors that could influence the migration of membrane proteins is the presence of bound lipopolysaccharide (Pugsley & Schnaitman, 1979), which could also result in diffused or distorted protein bands. In our samples, we have seen slightly skewed ompA protein bands, while a more pronounced band distortion was observed for ompX (Figure 3.3).

Specific detergent types can also affect the oligomeric structure and functional integrity of membrane proteins. From our results, we have observed cases wherein a specific type of protein has distinctive behavior in the presence of particular detergents.

For example, Fos-choline extracts the largest number (four) of protein subunits from the respiratory complex ATPase, but other detergents such as DM and Cymal 5 present interesting migration patterns between the different subunits that were different than the same proteins identified from Fos-choline. Another example is the behavior of ompF in various detergent preparations. OmpF was effectively extracted by five non-ionic detergents as well as the ionic surfactant SDS. However, the functional oligomeric units as well multiple banding patterns were only seen with specific detergents (SDS, DM,

Cymal5), which may suggest positive effects of these detergents in maintaining the stability of the target protein. These differences in the protein profiles of specific proteins in the presence of various solubilizing agents may give insights into interactions that may affect the protein’s functional integrity.

102

3.5 Summary

Various test detergents were screened to solubilize constitutive membrane proteins from bacterial membranes. Using 1D SDS-PAGE gel for separation and MALDI-TOF and database searching for protein identification, a total of 34 constitutive membrane proteins were detected from the total and outer membrane preparations.

Our study provides an initial detergent screen set that could be expanded to facilitate the selection of detergents for optimal extraction of different target membrane proteins. To complement the detergent extraction studies, it will also be worthwhile to carry out functional characterization to monitor the ability of different detergents to maintain the functional integrity of the protein targets.

103

Chapter 4

Structure of the First Fungal Ortholog of Aspartate

β-Semialdehyde Dehydrogenase

4.1 Introduction

The enzyme aspartate β-semialdehyde dehydrogenase (ASADH) catalyzes a critical transformation step that produces the first branch point intermediate in an essential microbial amino acid biosynthetic pathway. This pathway is responsible for the production of four essential amino acids as well as various metabolites that are important in carrying out various cellular functions. The aspartate pathway in bacteria differs slightly with that of fungi. In bacteria, the four essential amino acids threonine, isoleucine, methionine and lysine are all synthesized using aspartate as precursor. Fungal organisms also make use of the same aspartate to produce these amino acids with the exception of lysine which is obtained from α-ketoglutarate. This different route in lysine biosynthesis represents one major phylogenetic divergence between bacteria and fungi

(Moat et al., 2002). The aspartate pathway serves as a good target for the development of antimicrobial agents because it is required for microbial viability. Furthermore, this pathway is absent in mammals which helps in selectively inhibiting the targeted

104

microorganism.

ASADH has been mechanistically characterized (Karsten & Viola, 1991) and representative structures from Gram-negative (Blanco et al., 2003; Hadfield et al., 1999),

Gram-positive (Faehnle et al., 2006) and archael species (Faehnle et al., 2005) have been determined. Based on this structural and mechanistic background, a series of selective inhibitors are being investigated as potential lead compounds for the production of new antibacterial agents. However, no structural information has been obtained on any

ASADHs from fungal sources to guide the development of antifungal agents.

The goal of this research is to obtain a representative structure of ASADH from the fungal family and to learn new details from structural comparisons between ASADHs from diverse species. ASADH from the yeast species Candida albicans (caASADH) was chosen as our target because of increasing reported cases of invasive candidiasis. This kind of fungal infection has been ranked as the third most common bloodstream infection in United States leading to cases of substantial morbidity and mortality (Ostrosky-

Zeichner et al., 2010). Recent developments in the antifungal pipeline identified several antifungal agents that are primarily classified based on their mechanism of action.

Echinocandins and nikkomycin Z were identified to competitively inhibit chitin synthases resulting in fungal cell wall disintegration (Cabib, 1991). The azoles, polyenes and terbinafine have been associated with the disruption of fungal cell membrane. Other classes of compounds such as sordarins and flurocytosine were reported to interfere with protein assembly and DNA synthesis, respectively (Ostrosky-Zeichner et al., 2010).

Despite these developments in the pipeline, there is still an increasing demand for antifungal agents due to the emergence of susceptible populations as well as the

105

occurrence of antifungal resistance. In this study, we exploit the viability of the aspartate semialdehyde dehydrogenase enzyme involved in the fungal aspartate pathway as a candidate target for antifungal development.

4.2 Cloning, Expression and Purification of caASADH

4.2.1 Cloning and Expression

The cloning of the hom2 gene encoding aspartate β-semialdehyde dehydrogenase was previously described (Liu, 2007). Briefly, the hom2 gene was amplified from

Candida albicans genomic DNA using standard polymerase chain reaction methods. The amplified gene was ligated into a pET-28 expression vector containing an N-terminus hexahistidine tag using NdeI/EcoRI endonuclease restriction sites to yield the caASADH expression plasmid. To obtain cell cultures of caASADH, the plasmid was transformed into Escherichia coli strain BL21 cells and grown at 310 K in 4 L flasks containing Luria

Broth medium supplemented with 30 μg/mL of kanamycin. Protein expression was initiated by the addition of 1 mM IPTG and the temperature was lowered to 301 K and maintained at this temperature for five hours, after which cells were recovered by centrifugation at 10,000 rpm for 10 minutes at 4 °C. The harvested cell pellets were stored at 193 K until use for protein extraction.

4.2.2 Purification of Native caASADH

To extract the protein of interest, cell pellets were resuspended in lysis buffer composed of 50 mM Tris-HCl, pH 7.5, 2 mM β-mercaptoethanol and 200 mM ammonium acetate, lysed by ultrasonication using 50% power for a total exposure time of

106

5 minutes. The lysed sample was then centrifuged at 11,000 rpm at 277 K for 30 minutes.

The resulting supernatant (soluble fraction) from this step was used as the enzyme source for purification. The soluble fraction was loaded onto a cobalt immobilized metal affinity

(Co-IMAC) column pre-equilibrated with 10 column volumes of lysis buffer and the target enzyme was eluted by using an imidazole gradient from 0-200 mM. Further purification was carried out with a high resolution anion exchange column (Source 30Q) pre-equilibrated with 40 mM MES, pH 6.5, 1 mM EDTA and 2 mM DTT. The caASADH enzyme was separated from the remaining impurities using an ammonium acetate gradient from 0-500 mM.

Enzymatic activity was monitored at 340 nm following the reverse reaction resulting in an increase in absorbance of NADPH. Protein profiles were visualized using

Coomassie blue stained sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-

PAGE). Dynamic light scattering measurements were done prior to crystallization trials to check for the polydispersity of the enzyme solution.

The enzyme was assessed to be >95% pure based on the SDS-PAGE profiles and was shown by dynamic light scattering (DLS) experiments to exhibit <10% polydispersity, indicative of a highly purified, homogeneous protein (Figure 4.1). The protein was dialyzed into a storage buffer (20 mM HEPES, pH 7.0, 2 mM DTT and 1 mM EDTA) and was concentrated to 20 mg/mL for crystallization by using an Amicon spin concentrator (10 kDa MWCO). Samples were stored at 277 K until use for crystallization.

107

Table 4.1 Specific activities of the native form of fungal ASADH from Candida albicans (caASADH).

Parameters Crude IMAC Anion Exchange [Protein], mg/mL 9.88 2.42 5.18

Total Volume, mL 50 40 8

Total Protein, mg 494 97 41

Total Units (U) 1428 330 260

Specific Activity 2.89 3.41 6.34 (U/mg) % Yield 100 23 18

A B

64 kDa

Figure 4.1 Characterization of native caASADH. (A) SDS-PAGE profile of the caASADH purification and (B) dynamic light scattering measurements of the isolated enzyme indicating it as a monodisperse sample.

108

4.2.3 Purification of the Selenomethionine Form of caASADH

The selenomethionine form of fungal ASADH was also tried for crystallization studies because of the initial difficulty of getting good diffraction quality crystals from the unmodified protein. Expression of the selenomethionine form was done following the methionine pathway inhibition method (Doublie, 1997). Extraction and purification of the selenomethionine form was carried out in a similar manner to that of the native enzyme.

However, higher concentrations of reducing agents like β-mercaptoethanol (5 mM) and dithiothreitol (5 mM) were included in the buffers to minimize oxidation of the SeMet groups.

Pure SeMet enzyme was obtained after two step purification. The enzyme activity of the SeMet form is lower as compared to the native form (Table 4.2). The gel profiles and DLS profiles of the selenomethionine form is similar to those obtained from the native form of caASADH. The protein migrates around the ~64 kDa which may indicate a heat resistant dimer, while the DLS profile also shows a homogeneous sample showing a polydispersity lower than 10%.

109

Table 4.2 Specific activities of the selenomethionine form (SeMet) of fungal ASADH from Candida albicans (caASADH).

Parameters Crude IMAC Anion Exchange [Protein], mg/mL 11.67 1.48 0.67

Total Volume, mL 60 30 24

Total Protein, mg 700 44 16

Total Units (U) 861 143 76

Specific Activity 1.23 3.24 4.73 (U/mg) % Yield 100 17 8

A M Crude IMAC AIEX B

64 kDa

R (nm) MW (kDa) % Pd 4.9 138 9.2

Figure 4.2 Characterization of the selenomethionine form of caASADH. (A). SDS- PAGE profile showing the presence of a ~64 kDa protein and (B) Dynamic light scattering measurements indicating the presence of possible tetrameric form of caASADH with Mr of 138 kDa.

110

4.3 Crystallization of caASADH

4.3.1 Crystallization of the Apoenzyme Form of caASADH

Initial screening of crystallization conditions for the apo form of caASADH was carried out with several commercial screening kits (PEGs, PACT, AmSO4, Classics,

Hampton Screens) and in-house custom screen kit (Alpha Screen, Viola) with a total of

480 conditions. Drops for the screening trays were set-up by mixing equal volumes of 10 mg/mL caASADH and the crystallization solution in a 96-well sitting drop tray incubated at 293 K. Each reservoir contains 70 μL of the crystallization solution.

Screening of the apoenzyme form gave thin rods and needle crystals which mostly grew overnight after set-up. Among all of the tested sparse matrix screens, the

PEGs and the Alpha Screen gave better looking rod crystals. The conditions from these screens served as starting points and were optimized to obtain diffraction quality crystals.

Optimization was carried out by varying the pH, precipitant and protein concentrations.

Expansion of the initial crystallization hits gave diamond and rod apoenzyme crystals.

Diamond crystals were obtained from drops grown under equal volumes of caASADH

(16 mg/mL) and reservoir buffer consisting of 21-24 % PEG 200, 0.1 M MES pH 6.5, and 0.01 M dithiothreitol (DTT). These crystals grew after five to seven days when incubated at 293 K, (Figure 4.3A). When tested for diffraction quality, this crystal form of the apoenzyme diffracted to only 8 Å resolution. Further optimization of the conditions did not result in better quality crystals.

On the other hand, thin rod crystals were initially obtained at 10 mg/mL protein

(native caASADH), 16-18% PEG 2000 monomethyl ether (MME), 0.1 M MES pH 6.5,

0.1 M NaCl, 0.1 M Li2SO4. These crystals usually grow overnight after set-up. Further

111

optimization was carried out by increasing the protein concentration to 16 mg/mL, which resulted in larger rod crystals (Figure 4.3B). As compared with the diamond crystals, the rod crystals diffracted to 2.8 Å but with a somewhat higher mosaicity of ~1.0°. The fast crystal growth may have lead to a less ordered crystal lattice resulting in higher mosaicity. Attempts to get better quality crystals by slowing down the crystal growth were carried out by varying the incubation temperature, the volume of the reservoir buffer, and the addition of oils in the crystallization well. Incubating the trays at 277 K gave smaller and thinner crystals overnight, whereas varying the volume of the crystallization well did not have much effect on slowing down the growth. Similarly, the use of silicon and paraffin oil did not retard the crystal growth. For all the modified methods, crystals still grew overnight after set-up.

Despite the high mosaicity of the apoenzyme crystals, data were collected at the synchrotron (Argonne National Laboratory GM-CAT Beamline). The apoenzyme crystal diffracted at 2.7 Å with a lower mosaicity of 0.6°. However, the diffraction pattern indicated that our sample is not made up of single crystal. Preliminary identification of the space group gave a P1 and C2 classification. However, data processing using both space groups did not give useable results. The number of molecules per asymmetric unit was identified to be ~40. Such a high number may be due to the multiple crystals present in our samples. Based on these results, we were not able to do any further work on the structure determination of the apoenzyme. Instead, we focused our efforts on obtaining better diffraction quality crystals from other enzyme forms of caASADH.

112

A B

Figure 4.3 Crystals of the apoenzyme form of native caASADH. Crystals were obtained from the expansion of initial hits from sparse matrix screening. (A). Diamond crystals grew from drops incubated in 21-24 % PEG 200, 0.1 M MES pH 6.5, and 0.01 M DTT. (B). Rod crystals obtained at 16-18% PEG 2000 monomethyl ether (MME), 0.1 M MES pH 6.5, 0.1 M NaCl, 0.1 M Li2SO4.

113

4.3.2 Crystallization of the Selenomethionine Form of caASADH

Initial screening for conditions of crystal growth using the PEGs Suite (Qiagen) and Alpha screen gave thin rods and needle crystals for the selenium-modified enzyme.

Crystallization hits were obtained from pH range of 5.5 to 7.5 and different molecular weights of polyethylene glycol. Most common additives found for the crystallization hits include salts such as sodium chloride, magnesium chloride and lithium sulfate.

Preliminary optimization of the initial conditions gave crystals diffracting from 3.8 Å to

5.0 Å resolution with mosaicity ranging from 2° to 3 °. These crystals were grown from drops incubated with 15-21% PEG 2000 MME, 0.1 M MES pH 6.5, 0.1 M NaCl, 0.1 M

MgCl2 and 0.01 M DTT. No improvement in diffraction quality was obtained from the selenomethionine form, thus further optimization was not pursued.

4.3.3 Crystallization of the caASADH-Binary Complex with NADP

Because of the difficulty in obtaining high quality apoenzyme crystals from both the native and selenomethionine forms of caASADH, efforts were focused on obtaining diffraction quality crystals of the caASADH-NADP complex. This binary complex was prepared with a final concentration of 10 mg/mL of protein in the presence of 2 mM

NADP. Several initial crystallization conditions were identified from a PEG/Ion Screen

(Hampton Research). Improved clusters of rod crystals were obtained from drops grown in 20% PEG 400 and 0.1 M HEPES, pH 6.5 (Figure 4.4A). After further optimization by varying the protein concentration as well as by screening different additives using the

Additive Screens 1 and 2 (Hampton Research), single crystals grew from a 1:1 mixture of the complex (14 mg/mL protein, 2 mM NADP) and reservoir solution (20% PEG 400, 0.1

114

M HEPES pH 6.5, 0.1 M MgCl2) after several days of incubation at 293 K (Figure 4.4B).

Crystals were harvested and subsequently transferred to a cryogenic solution prepared from the reservoir solution and 25% glycerol, after which the cryoprotected crystal was flash frozen in liquid nitrogen for data collection.

4.4. Structural Determination of caASADH-Cofactor Complex

4.4.1 Data Collection and Processing

Preliminary screening of the diffraction quality of the apoenzyme and the caASADH-NADP complex crystals was conducted in-house with a Rigaku FR-E rotating anode generator equipped with an R-axis IV image plate detector. Subsequent synchrotron X-ray diffraction data for the NADP complex were collected at GM-CAT

Sector 23B beamline at the Advance Photon Source (Argonne National Laboratory). A complete data set was collected from a single frozen crystal of this NADP complex. The images were processed using HKL-2000 (Otwinowski & Minor, 1997) and scaling of the data set was done with the SCALEPACK program. The data collection statistics for the

NADP complex are summarized in Table 4.3. The crystal belongs to an orthorhombic space group C2221 with the asymmetric unit consisting of two monomers, giving a

Matthews coefficient of 2.13 Å3/dalton and a solvent content of ~42%.

115

A B

Figure 4.4 Crystals of caASADH in complex with its nucleotide cofactor, NADP. (A). Clusters of rod crystals obtained from drops grown in 20% PEG 400 and 0.1 M HEPES, pH 6.5. (B). Single crystal grew from drops incubated in the presence of 20% PEG 400, 0.1 M HEPES pH 6.5, 0.1 M MgCl2.

116

Table 4.3 Data collection statistics for the caASADH-NADP complex.

Parameters Values Temperature (K) 100

Space Group C2221 Unit Cell dimensions a, b, c (Å) 93.41, 152.18, 97.76 α, β, γ (°) 90, 90, 90 Wavelength (Å) 1.033

Resolution (Å) 2.2 (2.3-2.2) a

Total reflections 162,700

Unique reflections 32,749

Mosaicity (°) 0.94

a Rsymm 0.077 (0.44)

Output 20.2 (2.0) a

completeness (%) 91.5 (60.6) a

Redundancy 5.1 (2.3) a a The numbers in parenthesis indicate the values for the highest resolution shell.

117

4.4.2 Structure Determination, Refinement and Validation

The structure of the binary complex was solved by molecular replacement using the program Phaser (McCoy et al., 2005) and a monomer of the archeal ortholog of

ASADH from Methanococcus jannaschii (mjASADH, PDB 1YS4) (Faehnle et al., 2005).

Sequence comparison of the caASADH enzyme with various ASADH forms indicated that mjASADH has 60% sequence similarity and 40% sequence identity with this fungal

ASADH. The data from 79 to 3.5 Å resolution were utilized for the initial search and a replacement solution was obtained with a final TFZ score of 14. The resulting model was then subjected to twenty cycles of rigid body refinement using the medium resolution data up to 3.5 Å. This refinement gave an initial R value of 0.51. The model was further refined by using the Refmac program (Winn et al., 2001) and restrained refinement from

79 to 2.2 Å resolution, thereby lowering the R value to 0.40. The structure of the caASADH-NADP complex was built by a combination of manual building in Coot

(Emsley & Cowtan, 2004) followed by iterative rounds of restrained refinement.

Validation of the structure was carried out by using Procheck from the CCP4 suite

(Collaborative Computational Project, 1994). Surface calculations were conducted by using PISA (Protein Interfaces, Surfaces and Assemblies) at the www.ebi.ac.uk/msd- srv/prot_int/pistart.html web site.

4.4.3. Structural Refinement

The structure of the caASADH-NADP complex was solved by using the molecular replacement program Phaser, with the monomer of the most homologous

ASADH enzyme from Methanoccoccus janaschii utilized as the search model. A series of manual building and cycles of iterative restrained refinement with the data resolution

118

extended to 2.2 Å produced the final structure of the binary complex with refinement values of 0.23 and 0.28 for the Rwork and Rfree, respectively. The refinement statistics for this complex structure are summarized in Table 4.4.

Table 4.4 Refinement statistics for the caASADH-NADP complex.

Parameters Values

Resolution range (Å) 79-2.2

(2.3-2.2) a Wilson B factor (Å2) 42.4

Number of reflections used 32667

Rwork / Rfree (%) 0.23/0.28 (0.28/0.37) a R.m.s.d. bonds (Å) 0.007

R.m.s.d. angles (°) 1.115

Number of molecules per ASU 2

Number of atoms Protein/ligands/waters 5350/48/147 B factors (Å2) Protein/ligands/waters 48/82/49

Ramachandran plot analysis (%)b 93.6/6.4/0/0

a The numbers in parenthesis indicate the values for the highest resolution shell b most favored/additionally allowed/generously allowed/disallowed

119

The final electron density map allowed the complete building of residues 2-187 and 194-365 for each monomer, with both monomers lacking interpretable density for residues 188 to 193. However seven residues in monomer A and nine residues in monomer B were modeled as alanines due to lack of density to fully model the side chains. The truncated residues are primarily surface amino acids with long side chains such as lysine and arginine.

4.5 Structure of the caASADH-Cofactor Complex

4.5.1 Overall Structure

The asymmetric unit of the caASADH-NADP complex contains two subunits that represent the functional homodimer of this enzyme (Figure 4.5). This observed oligomeric form in the crystal structure is in contrast with the dynamic light scattering measurements where a tetrameric form of the enzyme has been initially observed (Figure

4.1). However most of the solved crystal structures of ASADH from various species indicate the dimer as the functional unit.

As has been observed with the other members of the ASADH structural family each monomeric unit is composed of an N-terminal coenzyme binding domain (residues

2-155, 347-365) and a C-terminal catalytic and dimerization domain (residues 156-346)

(Figure 4.5 and 4.6). The coenzyme binding domain is made up of six beta strands and seven alpha helical segments, while the C-terminal domain consists of six beta strands and eight alpha helices that represent the catalytic site and the dimerization interface. The two monomers are nearly identical in structure with a backbone rmsd of 0.4 Å. This fungal ortholog of ASADH most closely resembles the structure of the archeal

120

mjASADH, but with some important structural differences.

Figure 4.5 Overall structure of caASADH in complex with NADP. The monomer units are highlighted in different colors. Each monomer is composed of the N-terminal coenzyme binding domain (lighter shade) and the C-terminal dimerization and catalytic domain (darker shade). The nucleotide cofactor, NADP is shown as sticks model. Figure adapted from Arachea et al., 2010.

121

4.5.2. Secondary Structure Comparison

The N-terminus of the fungal ASADH starts with a β-strand (β1), followed by an alpha helix (α1) connected to two short beta strands (β2 and β3), completing the first β-α-

β motif of the Rossmann fold (Figure 4.6). The two short beta strands in caASADH replace the single beta strand that is typically seen for the bacterial ASADH structures

(Blanco et al., 2003; Faehnle et al., 2006). This motif is then linked to a 44 residue surface loop (residues 36-79) composed of three short helices (α2, α3, α4) that help to enclose the coenzyme binding domain. This longer surface loop observed in the fungal enzyme is a result of a ten residue insert from G42 to W51 (Figure 4.7), and this insert includes three lysines and an aspartic acid residue that form hydrogen bonds with the peptide backbone of adjacent amino acids.

Figure 4.6 Topological map showing the secondary structure of caASADH. The darker shade indicates the N-terminal coenzyme binding domain while the lighter shade shows the C-terminal catalytic and dimerization domain. Figure adapted from Arachea et al., 2010.

122

Following the surface loop is a second β-α-β motif (β4-α5-β5) of caASADH that is highly conserved among all of the ASADHs. The ASADH enzyme family also contains a third β-α-β motif, but the primary sequence and architecture of this motif is less conserved among the enzymes from different organisms. The fungal enzyme resembles the enzyme from the Gram-positive bacteria Streptococcus pneumonia (spASADH)

(Faehnle et al., 2006) with respect to the general fold of this third motif, with both enzymes containing an unstructured loop instead of the first beta strand, followed by an alpha helix (α6) and short loop leading to the second beta strand (β6) in the motif (Figure

4.6). However, caASADH features a slightly longer unstructured loop than spASADH, containing four additional residues, as well as a longer alpha helix (α6) composed of residues 127 to 141 (Figure 4.7).

The placement of the active site nucleophile in caASADH is also different as compared to its position in the bacterial ASADH structures. In the bacterial enzymes this catalytic cysteine is located at the top of a conserved helix (Blanco et al., 2003), while in the fungal structure the end of β7 connects to a short loop that contains the catalytic nucleophile. This location of C156 in caASADH appears to introduce sufficient flexibility to allow an alternative conformation for this active site nucleophile (Figure 4.

8). After this region the loop connects to a conserved helix (α7) and is then directed towards the first beta strand (β7) of the dimerization domain.

123

Figure 4.7 Sequence alignment of ASADH enzymes from different organisms. Conserved residues are shown in red, while similar sequences are shown in yellow. Figure adapted from Arachea et al., 2010.

124

Another significant difference in the fungal enzyme structure is a missing a helical subdomain in caASADH that is part of the dimerization interface in the bacterial

ASADHs and reaches across to the coenzyme binding domain of the adjacent subunit. In the Gram-negative bacteria this helical subdomain is composed of 44 amino acids organized in a helix-turn-helix, while this region is somewhat shorter (22 to 30 amino acids) in the ASADHs from Gram-positive bacteria (Figure 4.7). In contrast, this region in caASADH has only three residues and, despite being part of a conserved sequence in the ASADH family, there is no interpretable density in this region. These residues are likely part of a surface exposed loop that is observed in mjASADH stretching from residues 184 to 209 (Faehnle et al., 2005). The remainder of the dimerization domain in caASADH has a similar overall fold as the other ASADH enzymes, but with some critical differences in the length of each beta strand, the presence of intervening short helices and insertions (such as C284 to A294) in the surface loops (Figure 4.7), and changes in the identity of key amino acids that provide a communication channel between subunits in the bacterial enzymes.

4.5.3 Active Site Comparison

The functionally important residues present in the active site of ASADHs are conserved throughout the entire enzyme family. Superimposition of the residues in the active site of the fungal ASADH with the bacterial spASADH shows that these conserved residues are each in the same position, except for the active site cysteine nucleophile that is found to partially occupy a unique alternative conformation (Figure 4.8). The distance from the sulfur atom of C156 to the Nδ atom of the base catalyst H256 in the likely

125

productive conformation is 3.25 Å, which is similar to the distances observed in the

Gram-positive spASADH (3.22 Å) and the Gram-negative enzyme from Vibrio cholerae

(vcASADH) (3.1 Å). In the alternative conformation the sulfur to nitrogen distance in caASADH increases to 4.38 Å (Figure 4.8), and the position of H256 shifts in response to the changes in the cysteine conformation. A much lower catalytic efficiency would be expected for caASADH when the active site cysteine is in this alternative conformation, and partial occupancy in this conformation may explain the lower overall Kcat value for caASADH relative to the bacterial enzyme forms. The position of the catalytic cysteine is stabilized by a hydrogen-bonding interaction of its backbone carbonyl with the amide nitrogen of G160, while the position of H256 is stabilized by a hydrogen-bond between its δ-nitrogen and the amide nitrogen of A185, as well as a hydrogen-bond between its backbone carbonyl oxygen and the amide nitrogen of S341. The remaining active site residues, including R112 and K214 that participate in the positioning of the phosphate group, as well as E211 and R249 that are involved in the binding of aspartate semialdehyde, are all found in the same position as in the bacterial ASADH structures

(Figure 4.8) and almost certainly play the same roles in the catalytic cycle of these enzymes.

126

4.38 Å C156

3.25 Å

C136 H256 R112

ASA

K214

R249

Phosphate E211

B S341

G160

C156 4.38 Å

3.25 Å

R112 H256 A185

R249 K214 E211

Figure 4.8 Active site comparison of ASADH. (A) Superimposition of the catalytic residues of fungal ASADH (green) with those of bacterial ASADH isolated from H.influenza (white) showing the phosphate (red) and ASA (yellow) binding sites. (B). Active site residues of fungal ASADH showing Cys156 in an alternative conformation (sticks) and the amino acid residues that form interactions with Cys156 and His256.

127

4.5.4 Differences in Cofactor Binding

The enzyme caASADH was crystallized in the presence of its nucleotide cofactor, but differences in the interpretable electron density for the cofactor were observed between the two subunits. One subunit shows continuous density that allowed modeling of the adenine ring as well as the one of the phosphates that comprised the diphosphate moiety (Figure 4.9A), whereas the other subunit contains only broken density in this binding pocket. Neither subunit exhibited complete density for the nicotinamide ring, so the remainder of the NADP molecule was modeled into the structure using the NADP complex structure from its closest ASADH homolog, mjASADH, as a template.

Examination of the network of interactions between NADP and the amino acids in the coenzyme binding site reveals some similarities and some differences compared to the other ASADHs. The NADP in the caASADH structure is positioned through interactions with three sets of amino acids, S37-S40, T12-S14 and G84-V89, that are located in three different unstructured loops. The hydroxyl side chains of S37, S40 and

T12 form hydrogen bonds with the 2’-phosphate group of NADP (Figure 4.9A). In spASADH the related side chain functional groups form the same interactions (Faehnle et al., 2006). However, in the Gram-negative ASADHs the positions of the serine hydroxyl groups are altered and T12 is replaced by an arginine (Blanco et al., 2003). Despite these differences the 2’-phosphate group is found in nearly the same position in all of the

ASADHs.

A more dramatic difference is seen in the positioning of the adenine ring into a completely different pocket in the Gram-negative and Gram-positive bacterial enzymes

(Figure 4.9B). In the enzymes from Gram-negative bacteria R9 establishes a cation-π

128

interaction with the adenine ring in vcASADH (Blanco et al., 2003; Viola et al., 2008), while the same residue (R10) forms a similar contact with NADP in ecASADH (Hadfield et al., 2001). In addition, there is a hydrogen-bond between the 6-amino group of the adenine and the backbone carbonyl of P193 that is located in the other subunit of the dimer. In spASADH the adenine portion of the cofactor is anchored by the interactions of

S37 and T76 side chains as well as the formation of the cation-π interaction, but now with

R39. Making these new interactions requires a rotation in the positions of both the adenine and ribose rings leading to a displacement of as much as 14 Å (Faehnle et al.,

2006) when compared to their positions in the Gram-negative enzymes. The adenine moiety in the caASADH structure is positioned similarly to that observed in the Gram- positive enzyme (Figure 4.9B), but differences in the identity of the amino acids in this binding pocket alter the nature of the interactions. The adenine ring in caASADH is sandwiched between residues S37, S38, and a loop from G84 to V89 (Figure 4.9A).

Residues D86 to V89 are too far away to form direct interactions with the adenine ring.

The arginine at position 39 in caASADH is likely to form the same cation-π interaction observed in spASADH, however electron density in this new fungal structure is not sufficient to orient the arginine side chain, and this residue has been truncated to an alanine in this model.

Binding of the remainder of the NADP molecule includes an electrostatic interaction between the pyrophosphate moiety and the backbone amide nitrogen of S14.

The 2’-hydroxyl group of the nicotinamide ribose forms hydrogen bonds with a water molecule, which is further stabilized through hydrogen bonds with the carbonyl oxygen of L85 and an adjacent water molecule. The ribose 3’-hydroxyl group interacts with the

129

carbonyl oxygen of G84.

Figure 4.9 NADP binding conformations. (A) Binding conformation of the nucleotide cofactor (yellow) in caASADH, showing the interacting residues in green. (B) Comparison of NADP binding of fungal ASADH(green) against gram-positive ASADH (magenta) and gram-negative ASADH (blue).Figure adapted from Arachea et al., 2010.

130

4.5.5 Intersubunit Communication

A network of hydrogen bonding interactions in the dimerization interface of bacterial ASADHs have been proposed to play a key role in intersubunit communication

(Blanco et al., 2003). Four amino acids were identified to be involved in the information relay process between the active sites in subunits A and B. In vcASADH, the carboxylate side chain of E240 in the active site forms a hydrogen bond with the nearby amide side chain of Q161, which is also hydrogen-bonded to the hydroxyl group of T159 (Figure

4.10). The backbone carbonyl of T159 extends this network by forming a hydrogen bond across the dimer interface with the hydroxyl group of Y160 from the adjacent subunit.

The complimentary network of hydrogen-bonding interactions links the active site E240 from subunit B to Y160 of subunit A. The position of these bridging tyrosines is stabilized by π-stacking and by additional perpendicular π-stacking interactions with phenylalanines (F345) from their respective subunits (Blanco et al., 2003). This linking

4-amino acid bridge between the active sites of the functional dimer is also observed in a

Gram-positive ASADH, with the network of hydrogen bonding interactions from glutamic acid to tyrosine conserved in spASADH (Faehnle et al., 2006). The only difference between vcASADH and spASADH in this region is the replacement of the phenylalanine in vcASADH with a tryptophan in spASADH that preserves the stabilizing

π-stacking interaction.

The caASADH enzyme also displayed hydrogen bonding interactions between the active site D211 and N184, which in turn forms a hydrogen-bond with the hydroxyl group of T182. However, this network of hydrogen-bonding interactions is disrupted by replacement of the tyrosine residue that is present in the bacterial ASADHs and is crucial

131

in forming the communication link between the two subunits with a leucine (L183)

(Figure 4.10). Either leucine or methionine is found at this position in most of the fungal

ASADHs. This replacement also results in a loss of π-stacking interactions observed across the dimer interface, compounded by replacement of the stabilizing phenylalanine residues that flank the tyrosines in vcASADH and in nearly all of the fungal ASADHs by a valine (V338). These replacements effectively interrupt the intersubunit communication channel that links the active sites in the bacterial ASADH dimers. Similar disruption of intersubunit communication is observed in the archael ASADH from M. jannaschii in which the critical tyrosyl residue is replaced by a methionine and the phenylalanyl residue by a threonine (Faehnle et al., 2005).

Previous comparisons among the ASADH enzyme family also showed a correlation between the overall extent of surface contacts between the subunits at the dimer interface and enzyme activity. The enzyme orthologs with the smallest buried dimer surface and the smallest percent of buried surface area were found to have the lowest kcat values (Viola et al., 2008). When this new caASADH structure is compared to these other enzymes it has the lowest subunit surface area (~1800 Å2) and the lowest percent buried surface area (11%), consistent with its very low catalytic activity.

132

Figure 4.10 Dimerization interface of fungal ASADH versus bacterial ASADH. Interactions in the dimerization interface showing the linking 4-amino acid bridge responsible for the intersubunit communication in bacterial V. cholerae ASADH (blue). Replacement of these residues in the fungal enzyme (green) results in a loss of the hydrogen bonding network and pi-stacking interactions leading to lower catalytic activity. Figure adapted from Arachea et al., 2010.

133

4.5.6 Structure-Activity Relationships

While the residues in the active site are highly conserved throughout the ASADH family, the catalytic activity of this fungal ortholog is much lower than the other

ASADHs that have been kinetically characterized. The kcat of this fungal enzyme is only

0.12 ± 0.01 sec-1, more than three orders of magnitude lower than the most active

ASADHs, (Table 4.5). The Gram-negative forms of this enzyme have kcat values ranging from 610 sec-1 for the E. coli enzyme to 58 sec-1 for the enzyme from V. cholera (Moore et al., 2002), while the Gram-positive enzyme from S. pneumonia has a kcat value of 2 sec-1 (Faehnle et al., 2006). A comparison of the structure of the fungal enzyme with the other ASADHs provides some insights into how the catalytic activities of these related enzymes can be affected by subtle differences in structural features. While the identity of the catalytic and substrate binding groups are fully conserved, the catalytic cysteine in caASADH is found to exist partially in a non-productive alternative conformation

(Figure 4.8). Since neither ASA nor an amino acid analog are present in this complex it is possible that the equilibrium between the productive and non-productive orientations of the active site nucleophile could be altered upon substrate binding. It is therefore unclear whether any fraction of the enzyme nucleophile will remain in this non-productive conformation during the catalytic cycle thereby contributing to the lower activity of this fungal ASADH.

The orientation of the ASA substrate binding groups are essentially unchanged between caASADH and the other members of this enzyme family (Figure 4.8), so it is unlikely that the lower activity of the fungal ortholog is a consequence of changes in how this enzyme binds its amino acid substrate. There are, however, substantial differences in

134

Table 4.5 ASADH structure and activity relationship.

Enzyme Sequence Interface Communication kcat Form identity to area Channel (s-1) caASADH (Å2) (%)

caASADH -- 1806 No 0.12 mjASADH 44 1854 No 4 a,b vcASADH2 24 2352 Yes 58a spASADH 28 2819 Yes 2a hiASADH 20 3404 Yes 330a vcASADH1 19 3422 Yes 120a ecASADH 19 3708 Yes 610a a Data taken from Moore et al., (2002) and Faehnle et al., (2006) b Activity measured at 343 K

how caASADH binds its coenzyme substrate. The missing beta strand in the third β-α-β motif in the caASADH structure is similar to what was observed with mjASADH, and the absence of this beta strand was reported to affect coenzyme binding and contribute to the lower enzymatic activity found in the archael ortholog (Faehnle et al., 2005). The archael ASADH also lacks well defined density for the nicotinamide moiety of the bound

NADP, similar to what is seen for this fungal ASADH. This conformational flexibility in the portion of the cofactor that accepts the hydride from the substrate could be a contributing factor to the lower catalytic efficiency.

The loss of intersubunit communication in this fungal ortholog of ASADH is also proposed to play a role in its low catalytic activity. The helical subdomain in the bacterial forms of ASADH that provides contacts between the dimerization domain of one subunit

135

and the coenzyme binding domain of the adjacent subunit is absent in the caASADH structure. In addition, the hydrogen bonding network that connected the two active sites in vcASADH and spASADH is missing in caASADH. The low activity archael ASADH is also missing these intersubunit communication channels present in the bacterial

ASADHs. First, the helical subdomain is truncated in the archael enzymes and can no longer make contact across the dimer interface. Also, the hydrogen bonding network that provides a link between the active sites in the bacterial enzymes is interrupted in mjASADH, with conserved methionines replacing the bridging tyrosines and the stabilizing aromatic phenylalanines replaced with threonines (Faehnle et al., 2005). As a consequence the alternating site reactivity that is seen in bacterial ASADHs (Biellmann et al., 1980) is likely absent in both the fungal and archael forms of ASADH. Requiring the catalytic cycle in adjacent subunits to function out of phase (alternating site reactivity) would not appear to offer a catalytic advantage when compared to each site functioning independently. However, if the protein motions that are involved in substrate binding in one subunit can be coupled to complementary motions needed for product release in the adjacent subunit then an energetic advantage can be achieved. These types of coupled motions have been identified in a detailed analysis of the catalytic cycle of ASADH from

E. coli (Nichols et al., 2004). The loss of this network of hydrogen bonding interactions in caASADH as well as the absence of the helical subdomain will clearly affect intersubunit communication, requiring each subunit to function as an isolated catalytic entity and likely contributing to the lower catalytic efficiency of the fungal ASADHs.

136

4.6 Conclusions

The first structure of a fungal ortholog of ASADH has a similar overall fold and domain organization to the other members of this enzyme family. However, differences in positioning of the active site nucleophile, the orientation of the nucleotide cofactor and the nature of the subunit interface in the functional dimer are proposed to contribute to the lower catalytic efficiency of Candida albicans ASADH.

137

Chapter 5

Crystallographic Screening of ASADH Targets and Inhibitor

Complexes

5.1 Introduction

Our interest in developing inhibitors for the key metabolic enzyme aspartate β- semialdehyde dehydrogenase (ASADH) has encouraged us to target this enzyme from pathogenic organisms. The recent fragment-based kinetic identification of inhibitors of

ASADH (Gao et al., 2010) that were highly selective against gram-positive bacteria (S. pneumonia), gram-negative bacteria (V. cholerae) and fungal forms (C. albicans) provided the framework for the expansion of inhibitor studies by exploring more potent and selective compounds against a particular organism or a family of organisms.

Screening more targets from the same family of organisms will identify any common selectivity features within the same family. With this aim, we have expanded our studies by evaluating drug-resistant strains of the pathogenic organism Staphylococcus aureus. S. aureus together with other clinically important bacteria such as E. coli, M. tuberculosis,

Shigella, Salmonella, H. influenza and N. gonorrhoeae among others has been characterized by single and multiple antibiotic drug resistance by various biological

138

mechanisms (Levy & Marshall, 2004; Martinez et al., 2009). The resistance of these microorganisms to antibiotic therapy poses increasing risks to public health. Thus, the development of more effective antimicrobials that could combat specific or multiple drug resistance is necessary.

The cloning of the ASADH gene from a methicillin-resistant strain of S. aureus

(Mu50) was carried out using the Gateway Cloning Technology. The target enzyme was subsequently purified and was screened for initial crystallization conditions. Inhibitor screening was also carried out (Wang & Viola, unpublished results) and structural studies on inhibitor binding will follow once good diffraction quality crystals are produced.

Obtaining crystal structures of the target enzymes is necessary to aid in determining the mode of inhibitor binding. The crystal structures of the three representative forms of

ASADH (spASADH, vcASADH, caASADH) that were previously screened have each been solved, thus serving as molecular replacement search models in structure elucidation of bound inhibitors. The latest addition to the set of solved ASADH structures is the fungal ortholog isolated from C. albicans (Arachea et al., 2010). This published structure was utilized as a template in structural studies of caASADH-inhibitor complexes by x-ray crystallography. Several inhibitor complexes of caASADH with inhibitors identified from the water-soluble and organic-soluble fragment libraries were prepared by co- crystallization and soaking experiments, tested for diffraction quality and diffraction data were collected at the synchrotron.

This chapter is divided into two parts. The first part presents our efforts in the purification and crystallization of a methicillin-resistant strain of S. aureus, while the second part focuses on the structural studies of fungal enzyme-inhibitor complexes.

139

5.2 ASADH from Staphylococcus aureus

5.2.1 Target Identification

An initial protein search query was performed using the UniProt Consortium database (Jain et al., 2009) to obtain amino acid sequences of ASADH from various strains of S. aureus. The search results showed seventeen isolated strains of S. aureus, which include strains with different antimicrobial resistance or susceptibility. S. aureus is a pathogen that is efficient at developing antimicrobial resistance and strains which were identified to have developed resistance against beta-lactam antibiotics are classified as methicillin-resistant. Vancomycin-resistant strains on the other hand are insensitive to vancomycin, a glycopeptide antibiotic introduced in the 1950s to combat infections caused by S. aureus.

The level of antimicrobial resistance or susceptibility was previously defined from testing strains of S. aureus to obtain information on the minimal inhibitory concentration

(MIC). These interpretive criteria set MIC values as thresholds, above which the organism would be defined as resistant and below which an organism would be classified as either intermediate or susceptible. MIC testing of S. aureus that were reported by the

National Committee for Clinical Laboratory Standards proposed that vancomycin- intermediate strain be defined as S .aureus with an MIC of 8-16 mg/L when tested by recognized methods. If a particular strain has MIC values greater than 32 mg/L, the strain is classified as vancomycin-resistant, while those with MIC values lower than 4 mg/L are defined to be vancomycin-susceptible (Walsh & Howe, 2002)

140

Sequence comparison of the ASADH amino acid sequence from the seventeen strains indicated 97% sequence identity (Figure 5.1). Alterations in the amino acid sequence were seen in 11 positions with either a single or multiple mutations found in each strain as compared with the ASADH sequence of native S. aureus. For example, the vancomycin-resistant MRSA strain Mu50 isolated in 1997 has a glutamic acid residue at position 48 in contrast to a glycine residue at the same position in the native saASADH sequence. This single mutation was also seen in the sequences of an MRSA strain N315 as well as in the vancomycin-intermediate strains Mu3, JH1 and JH9 (Table 5.1).

On the other hand, multiple mutations were identified in MRSA isolates which were either community or hospital acquired. MRSA252 which is an MRSA strain acquired in hospitals showed four mutations in its ASADH sequence, which were different from the single mutation seen from the Mu50 strain. Other strains with multiple alterations in the sequence include MSSA476, MW2, RF122, MN8 and TCH60. The most common feature among these five strains is the replacement of amino acid residues at positions 42, 242 and 245. These positions in the native saASADH sequence are occupied by glutamine, aspartic acid and threonine, respectively. These same residues are then replaced by lysine, glutamic acid and alanine in the sequences of methicillin- resistant strains (Table 5.1). It appears that the strains harboring multiple mutations have at least some common alterations in their sequences. The ASADH sequences from the various strains were also compared with those of three representative forms of ASADH from other species. The comparison suggests that the location of the mutations occur in the variable regions of the ASADH sequence from these representative forms.

141

The differences in the ASADH amino acid sequence from various methicillin- resistant strains of S. aureus provide an interesting feature in studying key elements that may be essential in identifying selective inhibitors that would target a family of organisms or a smaller set from that same family. The comparison of the inhibition pattern against ASADH enzymes of selected methicillin-resistant strains with that of the non-resistant S. aureus will give information on features of inhibitor selectivity. It is however unknown whether these mutations in the ASADH have a direct correlation with antibiotic resistance in the family of S. aureus. Among the strains analyzed, the MRSA strain with vancomycin resistance (Mu50) was selected as our target. A single mutation was chosen to initially test our approach of identifying selective inhibitors against two different strains from the S. aureus family.

142

143

Figure 5.1 Amino acid sequence alignment of various strains of Staphylococcus aureus. Seventeen strains including antibiotic resistant strains of S. aureus were obtained from the UniProt Consortium database (accessed on October 28, 2009). The different strains are given using the database accession code followed by the strain name. High sequence identity of 97% was observed among the different variants. Alterations in the amino acid sequence were seen in 11 positions with variable mutations present in each of the strain. Highly conserved residues are shown in red, similar residues are highlighted in yellow while non-conserved amino acids are indicated in white.

144

Table 5.1 Mutations in the ASADH sequence of S. aureus variants. Amino acid mutations were identified with the various methicillin-resistant variants and these mutations were compared against equivalent positions in representative ASADH families.

Equivalent Positions in Representative ASADH Families

Accession Strain Amino acid Native Gram (-) bacteria Gram (+) bacteria Fungi Strain Description No. Name Mutations saASADH vcASADH spASADH caASADH

Q99U90 Mu50 Glu48 Gly48 Lys48 Asp49 Ala49 MRSA strain with vancomycin resistance Q6GH14 MRSA252 Lys42 Gln42 Pro42 Lys43 Lys43 Gly87 Ala87 Ala87 Ala87 Ala100 Hospital acquired MRSA Glu242 Asp242 Lys283 Lys261 Ala265 strain Ala245 Thr245 Ile286 Ala264 Pro270

Q7A5P8 N315 Glu48 Gly48 Lys48 Asp49 Ala49 MRSA strain

A7X270 Mu3 Glu48 Gly48 Lys48 Asp49 Ala49 Vancomycin-intermediate strain Q6G9G7 MSSA476 Ser9 Ala9 Try8 Ala10 Ala11 Asp25 Asn25 Asp25 Thr26 Pro27 Community-acquired Lys42 Gln42 Pro42 Lys43 Lys43 methicillin-susceptible Glu242 Asp242 Lys283 Lys261 Ala265 strain Ala245 Thr245 Ile286 Ala264 Pro270

Q8NWS6 MW2 Ser9 Ala9 Try9 Ala10 Ala11 Lys42 Gln42 Pro42 Lys43 Lys43 Community-acquired Glu242 Asp242 Lys282 Lys261 Ala265 MRSA strain Ala245 Thr245 Ile286 Ala264 Pro270

145

Table 5.1 (continued) Mutations in the ASADH sequence of S. aureus variants. Amino acid mutations were identified with the various methicillin-resistant variants and these mutations were compared against equivalent positions in representative ASADH families.

Equivalent Positions in Representative ASADH Families

Accession Strain Amino acid Native Gram (-) bacteria Gram (+) bacteria Fungi Strain Description No. Name Mutations saASADH vcASADH spASADH caASADH

Q2YXX5 RF122 Ile170 Val170 Met178 Leu172 --- Glu242 Asp242 Lys283 Lys261 Ala265 Community-acquired Ala245 Thr245 Ile286 Ala264 Pro270 methicillin-resistant strain Pro247 Ala247 Leu288 Ile266 Val272 Ala304 Ser304 Gly348 Ser323 Ser341

A5ISS6 JH9 Glu48 Gly48 Lys48 Asp49 Ala49 Vancomycin-intermediate strain A6UIL5 JH1 Glu48 Gly48 Lys48 Asp49 Ala49 Vancomycin-intermediate strain C2K8G4 MN8 Lys42 Gln42 Pro42 Lys43 Lys43 Gly87 Ala87 Ala87 Ala87 Ala100 MRSA isolate Glu242 Asp242 Lys283 Lys261 Ala265

C2GC86 TCH60 Lys42 Gln42 Pro42 Lys43 Lys43 Gly87 Ala87 Ala87 Ala87 Ala100 MRSA isolate Glu242 Asp242 Lys283 Lys261 Ala265 Ala245 Thr245 Ile286 Ala264 Pro270

146

Equivalent Positions in Representative ASADH Families

Accession Strain Amino acid Native Gram (-) bacteria Gram (+) bacteria Fungi Strain Description No. Name Mutations saASADH vcASADH spASADH caASADH

A6QGU5 Newman No mutation found Antibiotic-susceptible strain Q5HG26 COL No mutation found Methicillin-resistant strain A8Z3X2 TCH1516 No mutation found Community-acquired methicillin-resistant strain Q2FYP0 NCTC No mutation found MRSA isolate

147

5.2.2 Cloning of ASADH from S. aureus

5.2.2.1 Gene Amplification

The gene encoding the target enzyme was amplified from the genomic DNA of S. aureus Mu50 strain (American Type Culture Collection, Manassas, VA) using standard polymerase chain reactions. This target saASADH possess a glutamic acid residue at position 48 as compared with a glycine residue at this same position in native saASADH sequence (Figure 5.1, Table 5.1). Several reaction mixtures with different combinations of DNA template, primer, dNTPs, MgCl2 and DNA polymerase concentrations were prepared and subjected to the following thermocycler parameters: 368K denaturation step for 30 seconds, 328K annealing step for 30 seconds, 345K extension step for 60 seconds for a total of 30 cycles with the final extension step at 345K performed for 10 minutes.

An initial hot start step at 368K for one minute was done at the beginning of the run to allow complete separation of the template genomic DNA. The target ASADH gene was successfully amplified from a reaction mixture consisting of 20-40 ng genomic DNA, 0.4

μM each of forward and reverse primers (Table 5.3), 0.4 mM dNTP mix, 2U Vent polymerase and 1 mM MgCl2. Agarose gels showed the presence of a PCR product at

1000 bp region, indicative of successful amplification of the 990 base pair ASADH gene

(Figure 5.2). The band was excised and gel purified using the QIAQuick Gel Extraction

Kit (Qiagen) following the manufacturer’s protocol. Purified gene product was stored at

253K until used for cloning.

148

A B

Figure 5.2 Agarose gel electrophoresis profiles of the cloning of saASADH. A. The amplified saASADH gene obtained from polymerase chain reaction. B.PCR products generated from the attachment of attB sites to the saASADH gene in preparation to Gateway Cloning (Invitrogen). M: DNA marker; lanes 1-2: amplified saASADH gene resulting from 10 and 20 ng of genomic DNA template, respectively; lanes 3-5: PCR product obtained from the generation of attB sites onto the saASADH gene using 10-30 ng of template DNA.

149

5.2.2.2 Generation of the attB1-saASADH-attB2 PCR Products

Cloning of the saASADH gene to generate entry and expression clones was carried out using the Gateway Cloning Technology (Invitrogen). To perform the cloning reactions, PCR gene products must carry attB sites in order to be suitable substrates in the recombination reactions. These attB sites are site-specific attachment sites which serve as binding sites for recombination to occur between interacting DNA molecules.

The attB sites were incorporated into the saASADH gene that has been previously amplified using the attB adapter PCR method. For this purpose the following primers were designed and primer sequences are summarized in Table 5.3. The attB adapter PCR is divided into two steps. The first step involved the attachment of template specific primers to saASADH gene template and the second step is the incorporation of the attB adapter primers to create a full length attB sequence attached to the target gene.

Table 5.2 Primer designs used for the amplification and cloning of saASADH.

Primer Name Sequence

saASADH gene Forward Primer 5’ ATG ACA AAG TTA GCA GTT GTG GG 3’ saASADH gene Reverse Primer 5’ ATT CGC TCC TTT TAA ACG CAT AA 3’ saASADH attB1 Forward Primer 5’-AAA AAA GCA GGC TTC GAA GGA GAT (template specific primer) AGA ACC ATG ACA AAG TTA GCA GTT -3’ saASADH attB2 Reverse Primer 5’-GTA CAA GAA AGC TGG GTC ATT CGC (template specific primer) TCC TTT TAA ACG-3’ attB1 Adapter Primer 5’- G GGG ACA AGT TTG TAC AAA AAA GCA GGC T-3’ attB2 Adapter Primer 5’- GGG GAC CAC TTT GTA CAA GAA AGC TGG GT-3’

150

For the first step, three reaction mixtures (Mix A) were set up that contain the following components: 10-30 ng of saASADH DNA, 0.2 μM each of forward and reverse template specific primers , 0.4 mM of dNTPs, 2 U of Vent polymerase, 1 mM of magnesium chloride, 5 μL of polymerase buffer and final volume was adjusted to 50 μL.

The DNA fragment was amplified using the following thermocycler conditions: 368K denaturation for 30 seconds, 328K annealing for 30 seconds and 341K extension for 60 seconds for a total of 10 cycles. After ten cycles, the reaction was stopped and 10 μL aliquot of each PCR reaction mixture (Mix A) was transferred to three separate 40 μL of a freshly prepared reaction mixture (Mix B) in order to carry out the attachment of the full attB sequence to the gene of interest. The freshly prepared reaction mixtures contain the same components as Mix A with the exception of the attB adapter primers. The samples were then subjected to 20 cycles using the same thermocycler parameters as described above. Agarose gel electrophoresis indicated the attachment of the desired sequences onto the saASADH which is shown by an increase in band migration of the attB1-saASADH- attB2 PCR products as compared to that of the saASADH gene alone (Figure 5.2). The

PCR products were excised from the gel and purified using QIAQuick Gel Extraction Kit

(Qiagen).

5.2.2.3 Generation of the Entry and Expression Clones

To create entry clones, purified attB1-saASADH-attB2 (attB-saASADH) PCR products were then inserted into a donor vector pDONR 221 by a BP recombination reaction. The BP recombination reaction system (2 μL attB-saASADH at 42 ng/μL; 1.5

μL pDONR 221 at 65 ng/μL; 3.0 μL of BP Clonase and 8.5 μL of Tris-EDTA buffer, pH

151

8) was incubated at 298K overnight, after which 2 μL of the BP reaction mixture was transformed into 50 μL of DH5α competent cells (Invitrogen) following the manufacturer’s protocol. Transformants were selected on Luria Broth agar plates supplemented with 50 μg/mL kanamycin. Clones were then sent for DNA sequencing to confirm the identity of the insert (Intergrated DNA Technologies). Sequencing results indicated that the gene insert has 92% sequence identity with that of the target sequence

(Figure 5.3).

The target ASADH gene from the entry clone obtained from the BP recombination reaction was transferred to a destination vector pDEST 42 via an LR recombination reaction. This recombination step resulted in an expression vector with a hexahistidine tag at the C-terminal end. The LR recombination reaction system is as follows: 2 μL of the entry clone at 60 ng/μL, 1 μL of pDEST 42 at 150 ng/μL, 2 μL of

LR Clonase, and TE buffer added to 10 μL. This reaction mixture was incubated at 298K for six hours and was subsequently transformed into DH5α and transformants were screened using LB agar plates supplemented with 100 ug/mL ampicillin. The correct expression clones were confirmed by PCR and positive clones (saASADH-pDEST 42) were stored in 253 K until use.

152

Figure 5.3 DNA sequencing analysis of the saASADH entry clone. The gene insert in the entry clone showed 92% sequence identity with the target DNA sequence.

153

Figure 5.4 Confirmation of the expression clones. Entry and expression clones were generated using Gateway Technology. The identity of the gene insert in the entry clone was verified by DNA sequencing. The same gene insert was recombined with pDEST 42 to create the expression clone. The identity of the expression clones was verified by polymerase chain reaction. Amplification showed an intense band migrating at the expected molecular weight of the gene insert. M: DNA marker; lane 1: amplified saASADH gene from genomic DNA; lanes 2-4: saASADH gene amplified from isolated expression clones in the LR Recombination reaction.

154

5.2.3 Protein Expression

To test for protein expression, the expression clone was transformed into BL21

DE3 cells and was grown on LB agar plates supplemented with 100 μg/mL ampicillin.

Colonies were picked and introduced into four flasks containing 25 mL of LB broth with the same amount of antibiotic as previously described. Expression was induced by the addition of 1 mM IPTG after 2 hrs of incubation at 310K. The flasks were incubated at different post-induction temperatures (310K, 306K, 301K and 298K) and one mL aliquots were taken out from each flask at different time points. Cell pellets were obtained by centrifugation at 13,000 rpm for 5 minutes. Cells were lysed using 50 μL of

Bug Buster Protein Extraction Reagent, followed by 20 minutes incubation at room temperature with gentle shaking. Insoluble cell debris was recovered by centrifugation and the supernatant which contains soluble protein was analyzed for total protein content using the Bradford assay. The samples were subsequently tested for expression of the protein target by SDS-PAGE electrophoresis, with 15 μg of protein samples loaded onto each well. The target protein was further confirmed by Western blot by using an anti- histidine polyclonal antibody to detect the presence of the hexahistidine tag attached to the protein of interest. Gels used for both gel electrophoresis and western blot were run side by side for easy comparison of bands identifying the target protein.

Figure 5.5 shows the total protein levels extracted from soluble fraction of lysed cell extracts grown at different temperatures with five time points taken at each temperature. For each temperature, the highest protein production was observed between

4-8 hours after induction. Further incubation after this amount of time resulted to a decrease in extracted total protein. Comparison of 4 hr and 8 hr time points among the

155

different post-induction temperature showed that highest total protein levels were obtained at 298K, whereas incubation at 301K showed a slight decrease in protein levels.

On the other hand, cell growth at 306K and 310K have similar protein levels and these protein levels are at least twofold lower when compared with the values obtained at lower temperatures (Figure 5.5A)

Expression of the target saASADH was followed using gel electrophoresis and western blot. For all temperatures, the target protein was expressed in the soluble fraction and protein expression was already observed at 2 hrs after induction as seen from the appearance of a protein band at ~37 kDa region indicative of the presence of the monomeric form of the enzyme. Faint bands are also seen at the 64 kDa marker which corresponds to the dimeric form of the enzyme. The expression analysis established that the target protein is well expressed at the four temperatures tested, while optimal total protein expression was obtained at 298K with comparable protein levels obtained at

301K. (Figure 5.5B)

156

Figure 5.5 Protein expression of saASADH. (A). Total protein content analyzed from soluble fractions obtained at different post-induction temperatures. (B). SDS- PAGE and western blot profiles of the soluble fractions showing the expression of the target enzyme saASADH.

157

5.2.4 Stability Studies of saASADH

The cell growth of saASADH was carried out following the same procedure as described in Section 4.2.1 except for the use of 298K as the post-induction temperature.

Cell lysis and protein extraction were conducted in a similar manner using the method outlined in Section 4.2.2 with some modifications in buffer components used in the purification process. Freshly extracted crude samples of saASADH (50 mM Na- phosphate pH 7.5) showed a specific activity of 0.076 U/mg, which is one-fourth the specific activity of fungal ASADH (Table 4.1). This specific activity of saASADH is also very low as compared with the values obtained from crude ASADH fractions of gram- positive bacteria S. pneumonia (18.9 U/mg) and gram-negative bacteria V. cholerae (23.6

U/mg) (Camper & Viola, 2009).

Despite the unusually low enzyme activity, initial purification trials were carried out to check if the activity improves upon removal of impurities to isolate the pure enzyme. Isolation of the pure enzyme was performed using Ni-IMAC column followed by a high resolution anion-exchange column (Source 30Q). Enzyme activity was checked after each purification step and SDS-PAGE gels were run to monitor the purification process. The enzyme activity decreased dramatically as the protein was further separated from the impurities. Specific activity of the fractions obtained from the Ni-IMAC step showed 13-fold lower activity (0.006 U/mg) as compared to the crude enzyme, while fractions eluted from the anion-exchange column showed an even lower specific activity of 0.003 U/mg. Though the enzyme activity assays indicate loss of activity during purification, the gel profiles on the other hand suggested the enrichment of the well expressed protein band which is migrating in the molecular weight region that

158

corresponds to the dimeric form of saASADH. These results suggest that a pure but inactive form of the enzyme is being isolated and that the enzyme is not stable in the lysis and initial purification conditions that were used. Different purification schemes that were previously utilized to isolate other forms of ASADH were also tested to determine is other combinations of purification matrices and buffer conditions would stabilize an active enzyme. However, these various methods tested did not yield a functional enzyme form, suggesting that other parameters such as changes in pH and the addition of commonly used additives to stabilize proteins could be needed to help restore enzyme activity.

To systematically identify the essential component in restoring the enzyme activity of saASADH, freshly extracted crude samples of saASADH resuspended in 100 mM Na-phosphate pH 7.4 were dialyzed against buffers in the pH range from 4 to 8.

Although the values of specific activity increased by at least 100% by incubating at different pH values, these activity levels are still very low as compared to previously characterized ASADH forms, showing that changing the pH did not have a significant effect on enzyme activity. The identity of the buffer salts were also analyzed, but no further improvement in enzyme activity was observed. However, the addition of additives had a better effect on improving the activity. Salts like ammonium acetate, ammonium sulfate, sodium chloride and potassium chloride slightly increased the enzyme activity.

The most significant improvement in enzyme activity was seen in the presence of 15% glycerol in the lysis buffer which resulted in restoration of saASADH enzyme activity by nearly 90-fold (Table 5.3). The activity of saASADH in the presence of glycerol presents a novel case within the set of ASADH enzymes that have been mechanistically and

159

structurally characterized. All of the ASADH forms that have been characterized and reported so far did not require glycerol to obtain a functional form of the enzyme.

Table 5.3 Stability studies of saASADH. A. Effect of pH pH Buffer Specific Activity Fold Increase (U/mg) Control pH 7.4 0.1M Na-phosphate 0.012 --- 4 0.1 M citrate 0.039 3.2 5 0.1 M citrate 0.030 2.5 6 0.1 M MES 0.024 2.0 7 0.1 M Hepes 0.035 2.9 8 0.1 M Hepes 0.037 3.1

B. Effect of identity of buffer salt pH Buffer Specific Activity Fold Increase (U/mg) 7.4 0.1 M Na-phosphate 0.012 --- 7.4 0.1 M Tris-HCl 0.029 2.4 7.4 0.1 M Bis-Tris 0.052 4.3 7.4 0.1 M Hepes 0.025 2.1 7.4 0.1 M imidazole 0.027 2.3

C. Effect of additives pH Additive Specific Activity Fold Increase (U/mg) 7.4 No additive 0.012 --- 0.1 M Na-phosphate 7.4 0.1 M 0.065 5.4 CH3COONH4 7.4 0.1 M NaCl 0.076 6.3 7.4 0.1 M KCl 0.17 14 7.4 0.1 M (NH4)2SO4 0.49 41 7.4 15 % glycerol 1.05 88

160

On the other hand, the use of polyols such as glycerol is widely reported in the literature as structure stabilizing agents. Specific examples include stabilization of tubulin and calcium ATPases of sarcoplasmic reticulum, both of which were reported to lose native structure in aqueous buffers but are stable in concentrated solutions of glycerol even after prolonged storage (Sousa, 1995). Glycerol has also been routinely used in variety of applications such as protein refolding, protein crystallization, cryoprotectant, and as a component in both food and biopharmaceuticals (Vagenende et al., 2009).

To determine the optimal glycerol concentration needed to stabilize the enzyme during prolonged storage conditions different concentrations of glycerol were incubated with the crude saASADH fractions. Specific activity of protein samples stored at 278K was monitored for ten days. Enzyme activity increases with increasing glycerol concentrations, however only 15-20% concentrations are effective in maintaining enzyme activity after ten days of storage. Lower concentrations of glycerol tend to be detrimental to the enzyme, causing loss of activity after only a single day of incubation (Table 5.4).

Table 5.4 Effect of glycerol levels on saASADH enzyme activity.

Storage time Specific Activity (U/mg)

(Days) 20% 15% 10% 5%

1 0.82 0.87 0.58 0.09

3 0.50 0.67 0.06 0.06

5 0.52 0.74 Negative slope Negative slope

10 1.06 0.87 Negative slope Negative slope

161

The positive effect of glycerol on the enzyme activity of saASADH could be attributed to several factors. It is likely that glycerol plays a role in suppressing conformation flexibility in the protein structure leading to a more compact and ordered state. It could also stabilize the protein against denaturation and aggregation. These attributes could be correlated to molecular mechanisms that include steric effects, chemical interactions between the cosolvent and the protein, as well as effects on water structure which could lead to stabilized intramolecular H-bonding interaction and preferential interactions to bury hydrophobic surfaces to form a more compact conformation (Sousa, 1995; Vagenende et al., 2009)

5.2.5 Purification of saASADH

After several trials of protein purification, utilizing several pH values to isolate active saASADH, the target enzyme was purified at pH 7.5 using two chromatographic steps (summarized in Table 5.5). Phosphate buffer that was initially used in the purification trials was substituted with Hepes buffer to minimize salt contamination in the planned crystallization experiments. All buffers used contain 20% glycerol. Crude extracts were loaded onto an Ni-IMAC column and the target enzyme was eluted using an imidazole gradient (25 mM Hepes pH 7.5, 200 mM ammonium acetate, 0-200 mM imidazole). The Ni-IMAC chromatogram showed two peaks, one very sharp peak eluting at 20-50 mM imidazole and a smaller peak seen at higher imidazole concentrations (100-

150 mM). This smaller peak showed enzyme activity, whereas the former is mostly composed of impurities (Figure 5.6A). Active fractions were pooled and dialyzed overnight and further fractionated using a Source 30Q column pre-equilibrated with 25

162

mM Hepes 7.5, 1 mM DTT, 1 mM EDTA, followed by elution of the saASADH using an ammonium acetate gradient with the active enzyme eluting at 200-350 mM of ammonium acetate (Figure 5.6B). The target saASADH was assessed to be 95% pure by gel electrophoresis (Figure 5.6B inset). Dynamic light scattering measurements indicate the presence of a 78 kDa protein indicative of a dimeric form of protein in solution. The protein also exhibits less than 15% polydispersity which suggest a homogeneous sample ready for crystallization trials (Figure 5.6C).

Table 5.5 Specific activities of the ASADH from S. aureus (saASADH).

Parameters Crude IMAC Anion Exchange [Protein], mg/mL 17 1.07 0.71

Total Volume, mL 50 30 8

Total Protein, mg 850 32 5.7

Total Units (U) 918 228 88

Specific Activity 1.08 7.1 15.4 (U/mg) % Yield 100 25 9

163

2010 10 04 IMAC g48001:10_UV2_280nm 2010 10 04 IMAC g48001:10_Conc 2010 10 04 IMAC g48001:10_Fractions June 10 2010 TCH1516 Source30Q Amacetate(1276184703)001:10_UV1_280nm June 10 2010 TCH1516 Source30Q Amacetate(1276184703)001:10_Conc A June 10 2010 TCH1516 Source30Q Amacetate(1276184703)001:10_Fractions B June 10 2010 TCH1516 Source30Q Amacetate(1276184703)001:10_Logbook 2010 10 04 IMAC g48001:10_Logbook

mAU %B mAU %B 3000 100 M Crude IMAC AIX 50.0

50.0 2500 saASADH 80 40.0

2000 40.0

60 1500 30.0 30.0

1000

40 20.0 20.0

500 saASADH

10.0 20 10.0 0

0.0 0 -500 0.0 1 2 3 4 5 6 7 8 9 10 111213 141516 171819 2021 222324 252627 2829 303132 33 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 0 50 100 150 200 250 300 350 ml 0 50 100 150 200 250 300 ml C

R (nm) % Polydispersity Mr (kDa)

3.8 14.0 78

Figure 5.6 Purification of saASADH. Chromatograms of the (A) Ni-IMAC and (B) anion-exchange (Source30Q) purification steps. Purification was monitored using SDS- PAGE (inset) and enzyme activity assays. (C) Dynamic light scattering measurements of saASADH.

164

5.2.6 Screening of Crystallization Conditions for saASADH

5.2.6.1 Sparse-matrix Screening

For crystallization trials, active fractions obtained from the anion-exchange step were dialyzed against 25 mM Hepes pH 7.5, 1 mM DTT, 1 mM EDTA and 20% glycerol and was subsequently concentrated to 10 mg/mL using a spin concentrator (10 kDa

MWCO). Both the apoenzyme and the binary complex of saASADH with its nucleotide cofactor NADP were screened using commercially available sparse matrix screens

(PEGs, PACT, AmSO4, Classics, Hampton Screening) and a custom-made screen (Alpha

Screen, Viola Laboratory) with a total of 480 random conditions. Trays used for screening were set up by mixing 1 μL each of saASADH (10 mg/mL) and well solution and drops were suspended in a reservoir containing 70 uL of the crystallization solution. The trials for the binary complex were set-up in a similar manner but in the presence of 2 mM

NADP in the complex. Trays were maintained at 293K and drops were monitored for several weeks.

After two weeks of incubation, very thin needle clusters were obtained from two sparse-matrix conditions (Figure 5.8). Apoenzyme crystals grew from drops incubated with 20% PEG 550 MME, 0.1 M NaCl, 0.1 MgCl2, 0.1 M CAPSO pH 9.5, whereas the

NADP binary complex crystals were obtained from 30% MPD, 0.1 M sodium acetate buffer pH 4.6, 20 mM CaCl2. Expansion of the initial conditions using the hanging drop method was done by varying pH, precipitant and protein concentrations, however no reproducible crystals were obtained. Micro seeding of freshly prepared drops was also performed but this technique yield crystals with the same morphology as those of the initial hits, all of which are unsuitable even for screening for diffraction quality.

165

Figure 5.7 Sparse matrix screening of saASADH. Thin needle clusters were obtained from two sparse matrix conditions. (A) Apo enzyme crystals grown in the presence of 20% PEG 550 MME, 0.1 M NaCl, 0.1 MgCl2, 0.1 M CAPSO pH 9.5. (B) NADP binary complex crystals grew in drops incubated in 30% MPD, 0.1 M sodium acetate buffer pH 4.6, 20 mM CaCl2.

166

5.2.6.2 Systematic Evaluation of Conditions

Because of difficulty obtaining reproducible hits from sparse matrix screening, a systematic approach of evaluating crystallization conditions was employed. In this approach, the apoenzyme and saASADH-NADP binary complex were screened against representative compounds from major classes of precipitants on one dimension, while varying pH conditions as the second parameter. Classes of precipitants evaluated include long and short chain polymers (PEG 4000 and PEG 400), salts (ammonium sulfate and

NaCl), and organic solvents like ethanol and methylpentanediol. A 24-well tray was utilized for this purpose and a minimal crystallization matrix was used using different concentrations of the precipitant while varying pH conditions from pH 4-9 with increments of 0.5 pH units. Concentration values range from 5-30% v/v for PEGs and ethanol, 10-65% v/v for MPD, 1- 4 M for NaCl and 0.8-3.2 M for ammonium sulfate.

Buffers used for the various pH range include citrate buffer (pH 4-5.5), MES (pH 6 –

6.5), Hepes (pH 7-7.5), Tris (pH 8- 8.5) and Bicine (pH 9). Three drop sizes were prepared with different protein to reservoir volume ratio, followed by incubation at 293K using the hanging drop vapor diffusion method.

The systematic screens from ammonium sulfate and NaCl showed clear drops at all pH values tested. On the other hand, amorphous precipitation was observed at pH 4 to

7.5 for the PEG 4000 and PEG 400 screen. The organic precipitant ethanol also showed heavy and brown precipitate in all concentrations and pH values that were screened, whereas MPD mostly exhibited clear drops. Despite the exhaustive and systematic search to discover initial conditions to effect crystal growth for both the apo enzyme and NADP forms of saASADH, no hits were identified that could serve as good starting points for

167

crystallization.

5.2.6.3 Incorporation of Glycerol in the Crystallization Trials

Glycerol played a major role in restoring the enzyme activity of saASADH and concentrations below 10% dramatically decreased enzyme activity. For all the crystallization trials performed for this enzyme, the inclusion of glycerol in the mother liquor was not tested. Since the enzyme requires at least 15% of glycerol to remain active, it is possible that the addition of crystallization solution onto the protein drop during crystal screening reduces the amount of glycerol present in the protein sample.

This may possibly lead to partial denaturation and conformational flexibility of saASADH during the course of the crystallization trials as evidenced by the formation of amorphous precipitate in most drops.

To minimize this possible problem on sample denaturation, screening trials were set up with the addition of 20% glycerol in the crystallization well. For this purpose, two sparse-matrix screens (Classics Screen, Alpha Screen), two systematic screens (PEG4000 and ammonium sulfate as precipitants, pH range of 4-9) as well as two published crystallization conditions for ASADH enzymes were utilized (PEG 3350, 0.2 M

CH3COONH4 and PEG 8000, 0.2 M CH3COONH4 , pH 5-8.5). Both the apoenzyme and the binary complex were rescreened. After performing all these crystallization trials, no encouraging results were obtained from all screens even after six months of crystal monitoring.

An important bottleneck encountered while performing crystallization trials is to find the right combination of conditions that would effect crystal growth and to overcome this barrier by at least identifying preliminary conditions that could be further optimized.

168

The presence of glycerol in the saASADH samples adds another variable in the crystallization experiments. Enzyme activity assay have established the essential role of glycerol in retaining catalytic activity of saASADH. It is however, unknown whether the presence of glycerol will likely contribute in obtaining crystal hits. All of the ASADH enzymes that have been characterized and reported so far were isolated in an active form in the absence of glycerol, though there were cases when glycerol was used as a cryoprotectant in freezing ASADH crystal for low temperature data collection (Arachea et al., 2010; Faehnle et al., 2006; Blanco et al., 2003).

The use of glycerol in protein structure stabilization and crystallization has been reviewed by Rousa (1995). In his review, Rousa cited earlier examples of proteins that require the presence of glycerol in crystallization condition. One of them is the T7 RNA polymerase which was reported to crystallize with 25-30% v/v glycerol and that glycerol stabilized the crystalline form of the T7 RNAP relative to the amorphous state. A similar behavior was observed for crystallization of chimeric T7/T3 RNAP. Other examples of proteins that need glycerol for effective crystallization include the 40 kDa fibronectin fragment, penicillin G acylase, HIV transcriptase and DNA polymerase II (Rousa, 1995)

In all the mentioned cases, glycerol improved crystal size, quality and reduced mosaicity resulting to diffraction quality crystals. Future trials will be needed to systematically explore the inclusion of glycerol as a variable in screening for crystal growth.

169

5.3 Enzyme-Inhibitor Complexes of Fungal ASADH (caASADH)

5.3.1 Identified Inhibitors from Fragment Libraries

Identification of inhibitors of fungal ASADH isolated from Candida albicans was carried out by using a small-molecule fragment library screening approach. Experimental details and core structure elucidation of inhibitors are described elsewhere (Gao et al.,

2010). Briefly, the inhibitor screening from two fragment libraries found that phenyl and halo acid derivatives selectively inhibit the fungal form of the enzyme as compared to its bacterial counterparts, despite the highly conserved active site residues of the ASADH family. This set of inhibitors exhibit low millimolar binding affinity constants (Ki) and ligand efficiency (LE) values ranging from 0.37 to 0.83 (Table 5.6). Ligand efficiency represents the binding free energy per non-hydrogen atom, and these values are employed as a screening tool for fragment hit prioritization. Successful examples of initial fragment hits and further optimized drug leads were tabulated by Orita et al., (2009) and include compounds with average LE values of 0.40 ±0.10 and 0.38 ±0.08 for fragment hits and for drug leads, respectively.

170

Table 5.6 Inhibitors of fungal caASADH (Candida albicans) identified from fragment library screening (from from Gao et al., 2010) Inhibitor Chemical Formula Ki (mM) Ligand Efficiency

A. DMSO soluble fragment library - 3,4-Dihydroxyphenylacetate (HO)2-C6H3CH2COO 0.63 ± 0.14 0.37 - Phenylacetate C6H5CH2COO 0.83 ± 0.28 0.42 - 4-Aminophenylacetate H2N-C6H4CH2COO 2.2 ± 0.2 0.33 - 4-Nitrophenylacetate O2N-C6H4CH2COO 3.6 ± 0.7 0.26 - a 4-Hydroxyphenylacetae HO-C6H4CH2COO NI -- - 3-Aminophenylacetate H2N-C6H4CH2COO 0.61 ± 0.13 0.40 - 3-Hydroxyphenylacetate HO-C6H4CH2COO 0.49 ± 0.15 0.41 - a Phenyllactate C6H5CH2C(OH)COO NI -- - Phenylpropionate C6H5(CH2)2COO 0.63 ± 0.08 0.40 + - a Phenylglycine C6H5CH(NH3 )COO NI --

B. DMSO soluble fragment library - 3-Bromopyruvate CH2Br-C(O)COO 0.34 ± 0.07 0.68 - 3-Fluropyruvate CH2F-C(O)COO 0.14 ± 0.02 0.76 - 3-Iodopyruvate CH2I-C(O)COO 0.18 ± 0.02 0.73 - Chloroacetate CH2Cl-COO 4.6 ± 0.6 0.64 - Bromoacetate CH2Br-COO 3.2 ± 0.6 0.68 - Chlorodifluroacetate CF2Cl-COO 0.59 ± 0.05 0.63 - 3-Bromopropionate CH2Br-CH2COO 0.64 ± 0.05 0.73 - 3-Chloropropionate CH2Cl-CH2COO 0.24 ± 0.01 0.83 - 4-Bromobutyrate CH2Br-(CH2)2COO 0.16 ± 0.01 0.74 - 5-Bromovalerate CH2Br-(CH2)3COO 0.11 ± 0.03 0.68 - 6-Bromohexanoate CH2Br-(CH2)4COO 0.21 ± 0.02 0.56 - - Bromosuccinate OOC-CH2Br-CH2COO 0.55 ± 0.07 0.50 aNo inhibition observed at 10 mM

171

5.3.2. Crystallization of Enzyme-Inhibitor Complexes

5.3.2.1 Co-crystallization Approach

Protein-ligand complexes were formed either as a binary complex or as a ternary complex in the presence of its nucleotide cofactor, NADP or the cofactor analogue, 2’5’-

ADP. Both the soaking and co-crystallization approach were utilized in the preparation of the inhibitor complexes.

For the co-crystallization approach, purified caASADH (30-40 mg/mL) was incubated with 10-30 mM inhibitor concentrations for at least an hour to form the binary complex, followed by centrifugation and filtration (0.1 μm) to achieve a homogenous solution. The ternary complex was prepared in a similar manner but with the inclusion of

5 mM NADP or 2’, 5’-ADP in the complex. Binary and ternary co-crystal formation was induced by hanging drop vapor diffusion in a 24-well Linbro plate containing the previously optimized apo condition: 14-24% PEG 2000 MME, 0.1 M Li2SO4, 0.1 M

NaCl, 0.1 M MES pH 6.5. Preliminary crystallization trials indicated that only the complexes derived from the DMSO-soluble library produced thin rod co-crystals which were further optimized to obtain suitable size for in-house diffraction quality screening

(Figure 5.8D-F). Improved rod crystals grew from drops comprised of 2 μL complex: 2

μL well solution as well as from drops made up of 1 μL complex: 1 μL well solution: 1

μL water.

In contrast, protein-ligand complexes prepared with inhibitors from the water- soluble fragment library typically gave flaky and “grass-like” crystal morphologies

(Figure 5.8A) that diffract with poor resolution. New crystallization conditions at a similar pH range (pH 6-6.5) were tested by changing the identity of the precipitant (PEG

172

2000 MME versus PEG 3350, PEG 8000) and by the replacement of the NaCl and

Li2SO4 salts with either Mg(OAc)2, KOAc and CaCl2 as additives. With these few changes in conditions, more compact and better looking binary and ternary complex crystals (Figure 5.8B-C) were isolated from drops grown in the presence of 10-18% PEG

3350, 0.1 M KOAc, 0.1 M MES pH 6.

The co-crystals from the different complexes suitable for crystallographic screening were first protected with a cryogenic solution consisting of 20% ethylene glycol (or 25% glycerol), well solution with ligands used for the co-crystallization, and was subsequently flash frozen and maintained in liquid nitrogen until testing for diffraction quality.

173

Figure 5.8 Protein-inhibitor complexes of fungal caASADH from co-crystallization experiments. (A) Flaky crystals of protein complexes obtained from co-crystallization with water-soluble fragments. (B) Improved rod crystals of ternary complex caASADH- 3-chloropropionate-NADP and (C) caASADH-3-bromopyruvate-NADP obtained from 10-18% PEG 3350, 0.1 M KOAc, 0.1 M MES pH 6. Representative images of co-crystals prepared with DMSO-soluble inhibitors. (D) caASADH-phenylacetate binary complex, and ternary NADP (E) and 2’5’ADP (F) complexes grown in 14-24% PEG 2000 MME, 0.1 M Li2SO4, 0.1 M NaCl, 0.1 M MES pH 6.5.

174

5.3.2.2 Soaking Approach

Two preformed crystal morphologies of caASADH apoenzyme were utilized for the soaking studies with the inhibitors. Rod crystals from 14-24% PEG 2000 MME, 0.1

M Li2SO4, 0.1 M NaCl, 0.1 M MES pH 6.5, as well as diamond-shaped crystals grown in 15-18% PEG 400, 0.1 M MES pH, 0.01 M DTT (Figure 5.9) were incubated with phenylacetate and analogs at 10 mM concentrations for 30 minutes. Soaked crystals were transferred to a cryoprotectant solution and then dipped into liquid nitrogen.

Assessment of diffraction quality of the binary complexes from soaking experiments indicated that the crystals are of poor quality. Crystal cracking and dissolution were observed for both crystal forms upon incubation with inhibitors. The incubation time was decreased to 5 minutes and among the inhibitors tested, only the 3,

4-dihydroxyphenylacetate survived the soaking conditions. The cracking and dissolution of the apo crystals upon inhibitor binding could be attributed to protein conformational changes that are detrimental to the crystal lattice-packing interactions. Moreover, the lattice packing in the apo crystals may have formed interactions that can limit the access of inhibitor to the active site resulting in crystals of unsuitable diffraction quality.

Figure 5.9 Apo crystals of caASADH used for inhibitor soaking studies.

175

5.3.3 Data Collection and Structure Refinement

Screening of diffraction quality of the complexes was carried out on an in-house

R-AXIS IV image plate detector mounted on a Rigaku FR-E rotating anode X-ray generator. Co-crystals which diffracted with at least 3 Å resolution were brought to

Advanced Photon Source (Argonne National Laboratory) and diffraction data sets were collected on the GM/CA-CAT beamline Sector 23B. The data were processed with HKL-

2000 (Otwinowski & Minor, 1997) and scaled with SCALEPACK. Structural calculations were done using the CCP4 (Collaborative Computational Project, 1994) program suite with the REFMAC (Winn et al., 2001) program utilized for the refinements. The experimental model was initially refined as a rigid body using low resolution data at 3.5

Å, followed by restrained refinement with the data obtained in the highest resolution shell. Difference electron density maps were acquired and analyzed using COOT (Emsley

& Cowtan, 2004) for the presence of unoccupied electron density to identify possible sites of inhibitor binding.

5.3.4 Assessment of Structural Data and Ligand Binding

A total of six co-crystal structures were crystallographically analyzed for unaccounted electron density within the active site as well as in other parts of the model.

All structures characterized belong to C2221 space group except for the 3,4- dihydroxyphenlyacetate which crystallized in P1 space group and no useable solution was obtained for this complex. All of the collected diffraction data have high completeness (overall data) and refinement Rwork and Rfree values are typically 0.23 and

0.27, respectively. Despite obtaining 2-2.5Å resolution data of the putative enzyme- inhibitor complexes, no conclusive electron density was found that would allow modeling

176

of the inhibitor or even a fragment of the inhibitor. However, in this same resolution range the binding of NADP and 2’5’ADP is seen as evidenced by continuous electron density in the cofactor binding site which suggest that the cofactor concentration used was sufficient to observed binding interactions in our structure. The bromo- and chloro atoms in the pyruvate ligands were initially expected to display noticeable electron densities owing to the nature of the halo atoms, however no observed blobs could be correlated to either bromine or chlorine in the structures. Also, no electron density blobs anywhere in the experimental model could be conclusively used to account for the substituted phenylacetate inhibitors. The only blob that was seen so far in the vicinity of the active site represents a sulfate group which was derived from the crystallization condition. This sulfate group occupies the same position as that of the phosphate group in the active site of ASADH elucidated from previously solved structures. From these results, further optimization is necessary to get representative structures of fungal

ASADH with bound inhibitors.

The absence of bound inhibitors in the active site of caASADH could be due to various reasons. First, the conditions used for enzyme kinetic screening is different from the conditions used to form enzyme-inhibitor complexes for crystallographic screening.

These changes in pH conditions as well as the presence of precipitation may have an effect on inhibitor binding. Second, the low binding affinity of the fungal ASADH inhibitors means less involved interactions with protein residues due to smaller fragments, limiting the number of contacts involved in favorable interactions. These inhibitors may be present in the active site, however binding in different orientations would lead to either partial or the complete absence of electron densities.

177

Table 5.7 Diffraction data and refinement statistics of inhibitor complexes of caASADH.

Ligands Space Group Resolution Rsymm I/σ Completeness Redundancy Rw/Rfree Rmsd Rmsd Unit cell (Å) Bonds Angles (Å, °)

3-chloropropionate C2221 2.00 0.060 20.7 93.2 4.8 0.24/0.27 0.006 1.08 + NADP 93.7, 153.3, 98.4 (2.07-2.00) (0.454) (2.1) (68.1) (3.0) 90,90,90

Phenylacetate C2221 2.25 0.056 20.9 95 4.1 0.22/0.26 0.008 1.24 + NADP 93.5, 152.1, 98.2 (2.33-2.25) (0.226) (3.4) (79.5) (2.2) 90,90,90

3-bromopyruvate C2221 2.10 0.053 20.8 92.5 4.0 0.23/0.27 0.008 1.21 +NADP 93.9, 154.0, 98.3 (2.18-2.10) (0.337) (2.94) (67.3) (3.1) 90,90,90

4-nitrophenylacetate C2221 2.40 0.066 12.6 82.5 2.8 0.23/0.27 0.008 1.23 + 2’5’-ADP 94.8, 156.2, 99.3 (2.49-2.40) (0.299) (2.40) (38.7) (1.4) 90,90,90

Phenylacetate C2221 2.40 0.070 14.7 98.5 4.3 0.23/0.28 0.012 1.26 + 2’5’-ADP 94.1, 155.5, 99.0 (2.49-2.40) (0.201) (5.4) (94.1) (3.4) 90,90,90 3-4-dihydroxy P1 2.40 0.051 11.4 93.5 1.9 ------phenylacetate 105.9, 104.5, 180.0 (2.49-2.40) (0.438) (1.25) (65.0) (1.5) 75.1, 75.4, 84.9

178

5.3.5 Comparison with Inhibitor Binding Studies in other ASADH Forms

Ideally, the identified inhibitors should bind at the active site, however no bound inhibitors were observed from the structural studies of caASADH. However, co-crystal structures of enzyme-inhibitor complexes with bound inhibitors have been obtained from bacterial forms of ASADHs isolated from S. pneumonia (spASADH) and V. cholerae

(vcASADH). Inhibitors belonging to the family of amino acid and derivatives were found to bind at the active site of spASADH (Pavlovsky et al., 2011, manuscript submitted), whereas benzophenone and analogs were detected at the catalytic site of vcASADH, albeit with broken electron densities (Potente, 2010). One specific example of a bound

ASADH inhibitor is 2,3-diaminopropionic acid, which was found in the active site of spASADH thru the formation of favorable ion-pair interaction between the carboxylate group and the guanidinium group of Arg245. This interaction is further stabilized by H- bonding contacts between the 2-amino group and the carboxylate of a nearby Glu220 residue. Other examples of enzyme-inhibitors complexes from this bacterial ASADH family have also been characterized and were the focus of other studies (Pavlovsky et al.,

2011; Potente, 2010) and will not be further discussed. These cocrystal structures with bound ASADH inhibitors provide a structural framework for inhibitors that could be further optimized to enhance protein-inhibitor interactions that will result in the development of more potent ASADH inhibitors.

179

Lys223

Asn127 Glu220

Arg99

Arg245

Cys128 His252

NADP

Figure 5.10 Example of bound inhibitor in the active site of spASADH. The inhibitor 2, 3-diaminopropionate was found bound in the active site of ASADH isolated from the gram-positive bacteria S. pneumoniae. Shown are the interactions between the inhibitor functional groups and the catalytic residues. Figure adapted from Pavlovsky et al., 2011, manuscript submitted)

180

5.4 Summary and Future Work

The ASADH enzyme from S. aureus was successfully amplified and recombinantly expressed in E. coli cells. Active and stable enzyme was only isolated in the presence of 20% glycerol. Crystallization trials in the presence of absence of glycerol were evaluated and this resulted to obtaining thin needle apo and NADP complex crystals that require further optimization to obtain diffraction quality crystals. A variety of enzyme forms could also be tested to see whether these forms could produce crystals that will be used for structural studies. Other enzyme forms such as saASADH-inhibitor complexes as well as selenomethionine form of saASADH could be analyzed for crystallization trials to see if these forms will crystallize better as compared to the apoenzyme.

Enzyme-inhibitor cocrystals from fungal ASADH were analyzed by x-ray crystallography to determine inhibitor binding. However, structural characterization of these complexes did not show any bound inhibitors in the active site, in contrast to the data obtained from kinetic screening. It is recommended that the conditions used for crystallographic screening such as the addition of PEG2000 or other precipitants be evaluated while doing the inhibitor screening. Also, the other sets of inhibitors that were not tested due to lack of good diffracting crystals could be further exploited. It will also be worthwhile to try to cocrystallize the bacterial ASADH inhibitors with the fungal enzyme to check the binding of these inhibitors. The binding or non-binding of these bacterial ASADH inhibitors in the active site will further confirm the features of selectivity between the different ASADH families.

181

References

Akiyama, Y. (2009) Quality control of cytoplasmic membrane proteins in Escherichia coli, J Biochem 146, 449-454.

Arachea, B. T., Liu, X., Pavlovsky, A. G., and Viola, R. E. (2010) Expansion of the aspartate beta-semialdehyde dehydrogenase family: the first structure of a fungal ortholog, Acta Crystallogr D Biol Crystallogr 66, 205-212.

Benoit, S., Abaibou, H., and Mandrand-Berthelot, M. A. (1998) Topological analysis of the aerobic membrane-bound formate dehydrogenase of Escherichia coli, J Bacteriol 180, 6625-6634.

Biellmann, J. F., Eid, P., and Hirth, C. (1980) Affinity labeling of the Escherichia coli aspartate-beta-semialdehyde dehydrogenase with an alkylating coenzyme analogue. Half-site reactivity and competition with the substrate alkylating analogue, Eur J Biochem 104, 65-69.

Bill, R. M., Henderson, P. J., Iwata, S., Kunji, E. R., Michel, H., Neutze, R., Newstead, S., Poolman, B., Tate, C. G., and Vogel, H. (2011) Overcoming barriers to membrane protein structure determination, Nat Biotechnol 29, 335-340.

Blanco, J., Moore, R. A., and Viola, R. E. (2003) Capture of an intermediate in the catalytic cycle of L-aspartate-beta-semialdehyde dehydrogenase, Proc Natl Acad Sci U S A 100, 12613-12617.

Blanco, J., Moore, R. A., Faehnle, C. R., and Viola, R. E. (2004) Critical catalytic functional groups in the mechanism of aspartate-beta-semialdehyde dehydrogenase, Acta Crystallogr D Biol Crystallogr 60, 1808-1815.

Blanco, J., Moore, R. A., Faehnle, C. R., Coe, D. M., and Viola, R. E. (2004) The role of substrate-binding groups in the mechanism of aspartate-beta-semialdehyde dehydrogenase, Acta Crystallogr D Biol Crystallogr 60, 1388-1395.

Blanco, J., Moore, R. A., Kabaleeswaran, V., and Viola, R. E. (2003) A structural basis for the mechanism of aspartate-beta-semialdehyde dehydrogenase from Vibrio cholerae, Protein Sci 12, 27-33.

182

Brandt, U. (2006) Energy converting NADH : Quinone oxidoreductase (Complex I), Annu Rev Biochem 75, 69-92.

Byrne, B., Abramson, J., Jansson, M., Holmgren, E., and Iwata, S. (2000) Fusion protein approach to improve the crystal quality of cytochrome bo(3) ubiquinol oxidase from Escherichia coli, Biochim Biophys Acta 1459, 449-455.

Cabib, E. (1991) Differential inhibition of chitin synthetases 1 and 2 from Saccharomyces cerevisiae by polyoxin D and nikkomycins, Antimicrob Agents Chemother 35, 170-173.

Caffrey, M. (2000) A lipid's eye view of membrane protein crystallization in mesophases, Curr Opin Struct Biol 10, 486-497.

Camper, D. V., and Viola, R. E. (2009) Fully automated protein purification, Anal Biochem 393, 176-181.

Capaldi, R. A., and Aggeler, R. (2002) Mechanism of the F(1)F(0)-type ATP synthase, a biological rotary motor, Trends Biochem Sci 27, 154-160.

Cardineau, G. A. and Curtiss, R. (1987) Nucleotide sequence of the Asd gene of Streptococcus mutans: identification of the promoter region and evidence for attenuator-like sequences preceding the structural gene. J. Biol. Chem. 262, 3344- 3353.

Cecchini, G., Maklashina, E., Yankovskaya, V., Iverson, T. M., and Iwata, S. (2003) Variation in proton donor/acceptor pathways in succinate : quinone oxidoreductases, FEBS Lett 545, 31-38.

Cherezov, V., Rosenbaum, D. M., Hanson, M. A., Rasmussen, S. G., Thian, F. S., Kobilka, T. S., Choi, H. J., Kuhn, P., Weis, W. I., Kobilka, B. K., and Stevens, R. C. (2007) High-resolution crystal structure of an engineered human beta2- adrenergic G protein-coupled receptor, Science 318, 1258-1265.

Cohen, G. N. (1985) Aspartate kinases I, II, and III from Escherichia coli, Methods Enzymol 113, 596-599.

Corsini, L., Hothorn, M., Scheffzek, K., Sattler, M., and Stier, G. (2008) Thioredoxin as a fusion tag for carrier-driven crystallization, Protein Sci 17, 2070-2079.

Cowan, S. W., Schirmer, T., Rummel, G., Steiert, M., Ghosh, R., Pauptit, R. A., Jansonius, J. N., and Rosenbusch, J. P. (1992) Crystal structures explain functional properties of two E. coli porins, Nature 358, 727-733.

Cox, R. J., Gibson, J. S., and Hadfield, A. T. (2005) Design, synthesis and analysis of inhibitors of bacterial aspartate semialdehyde dehydrogenase, Chembiochem 6,

183

2255-2260.

Cox, R. J., Gibson, J. S., and Mayo Martin, M. B. (2002) Aspartyl phosphonates and phosphoramidates: the first synthetic inhibitors of bacterial aspartate- semialdehyde dehydrogenase, Chembiochem 3, 874-886.

Diedrich, D. L., Stein, M. A., and Schnaitman, C. A. (1990) Associations of Escherichia coli K-12 OmpF trimers with rough and smooth lipopolysaccharides, J Bacteriol 172, 5307-5311.

Doublie, S. (1997) Preparation of selenomethionyl proteins for phase determination, Methods Enzymol 276, 523-530.

Doyle, D. A., Morais Cabral, J., Pfuetzner, R. A., Kuo, A., Gulbis, J. M., Cohen, S. L., Chait, B. T., and MacKinnon, R. (1998) The structure of the potassium channel: molecular basis of K+ conduction and selectivity, Science 280, 69-77.

Emsley, P., and Cowtan, K. (2004) Coot: model-building tools for molecular graphics, Acta Crystallogr D Biol Crystallogr 60, 2126-2132.

Faehnle, C. R., Blanco, J., and Viola, R. E. (2004) Structural basis for discrimination between oxyanion substrates or inhibitors in aspartate-beta-semialdehyde dehydrogenase, Acta Crystallogr D Biol Crystallogr 60, 2320-2324.

Faehnle, C. R., Le Coq, J., Liu, X., and Viola, R. E. (2006) Examination of key intermediates in the catalytic cycle of aspartate-beta-semialdehyde dehydrogenase from a gram-positive infectious bacteria, J Biol Chem 281, 31031-31040.

Faehnle, C. R., Ohren, J. F., and Viola, R. E. (2005) A new branch in the family: structure of aspartate-beta-semialdehyde dehydrogenase from Methanococcus jannaschii, J Mol Biol 353, 1055-1068.

Faham, S., and Bowie, J. U. (2002) Bicelle crystallization: a new method for crystallizing membrane proteins yields a monomeric bacteriorhodopsin structure, J Mol Biol 316, 1-6.

Faham, S., Boulting, G. L., Massey, E. A., Yohannan, S., Yang, D., and Bowie, J. U. (2005) Crystallization of bacteriorhodopsin from bicelle formulations at room temperature, Protein Sci 14, 836-840.

Fountoulakis, M., and Gasser, R. (2003) Proteomic analysis of the cell envelope fraction of Escherichia coli, Amino Acids 24, 19-41.

Galan, J. E., Nakayama, K. and Curtiss, R. (1990). Cloning and characterization of the Asd gene of Salmonella typhimurium: use in stable maintenance of recombinant plasmids in Salmonella vaccine strains. Gene, 94, 29-35.

184

Gao, G., Liu, X., Pavlovsky, A., and Viola, R. E. (2010) Identification of selective enzyme inhibitors by fragment library screening, J Biomol Screen 15, 1042-1050.

Gevaert, K., Van Damme, J., Goethals, M., Thomas, G. R., Hoorelbeke, B., Demol, H., Martens, L., Puype, M., Staes, A., and Vandekerckhove, J. (2002) Chromatographic isolation of methionine-containing peptides for gel-free proteome analysis: identification of more than 800 Escherichia coli proteins, Mol Cell Proteomics 1, 896-903.

Hadfield, A., Kryger, G., Ouyang, J., Petsko, G. A., Ringe, D., and Viola, R. (1999) Structure of aspartate-beta-semialdehyde dehydrogenase from Escherichia coli, a key enzyme in the aspartate family of amino acid biosynthesis, J Mol Biol 289, 991-1002.

Hadfield, A., Shammas, C., Kryger, G., Ringe, D., Petsko, G. A., Ouyang, J., and Viola, R. E. (2001) Active site analysis of the potential antimicrobial target aspartate semialdehyde dehydrogenase, Biochemistry 40, 14475-14483.

Harb, O.S. and Kwaik, Y.A. (1998) Identification of the aspartate-β-semialdehyde dehydrogenase gene of Legionella pneumophila and characterization of a null mutant. Infect. Immun. 66, 1898-1903.

Heginbotham, L., Odessey, E., and Miller, C. (1997) Tetrameric stoichiometry of a prokaryotic K+ channel, Biochemistry 36, 10335-10342.

Hennecke, G., Nolte, J., Volkmer-Engert, R., Schneider-Mergener, J., and Behrens, S. (2005) The periplasmic chaperone SurA exploits two features characteristic of integral outer membrane proteins for selective substrate recognition, J Biol Chem 280, 23540-23548.

Hunte, C., and Michel, H. (2002) Crystallisation of membrane proteins mediated by antibody fragments, Curr Opin Struct Biol 12, 503-508.

Hunte, C., Koepke, J., Lange, C., Rossmanith, T., and Michel, H. (2000) Structure at 2.3 A resolution of the cytochrome bc(1) complex from the yeast Saccharomyces cerevisiae co-crystallized with an antibody Fv fragment, Structure 8, 669-684.

Hwang, B. Y., Varadarajan, N., Li, H., Rodriguez, S., Iverson, B. L., and Georgiou, G. (2007) Substrate specificity of the Escherichia coli outer membrane protease OmpP, J Bacteriol 189, 522-530.

Iwata, S. (2003) Methods and Results in the Crystallization of Membrane Proteins. International University Line, La Jolla, CA

Jain, E., Bairoch, A., Duvaud, S., Phan, I., Redaschi, N., Suzek, B. E., Martin, M. J.,

185

McGarvey, P., and Gasteiger, E. (2009) Infrastructure for the life sciences: design and implementation of the UniProt website, BMC Bioinformatics 10, 136.

Johansson, L. C., Wohri, A. B., Katona, G., Engstrom, S., and Neutze, R. (2009) Membrane protein crystallization from lipidic phases, Curr Opin Struct Biol 19, 372-378.

Karsten, W. E., and Viola, R. E. (1991) Chemical and kinetic mechanisms of aspartate- beta-semialdehyde dehydrogenase from Escherichia coli, Biochim Biophys Acta 1077, 209-219.

Kihara, A., Akiyama, Y., and Ito, K. (1996) A protease complex in the Escherichia coli plasma membrane: HflKC (HflA) forms a complex with FtsH (HflB), regulating its proteolytic activity against SecY, Embo J 15, 6122-6131. Kim, J., Lee, Y., Kim, C., and Park, C. (1992) Involvement of SecB, a chaperone, in the export of ribose-binding protein, J Bacteriol 174, 5219-5227.

Kolbe, M., Besir, H., Essen, L. O., and Oesterhelt, D. (2000) Structure of the light-driven chloride pump halorhodopsin at 1.8 A resolution, Science 288, 1390-1396.

Koronakis, V. (2003) TolC--the bacterial exit duct for proteins and drugs, FEBS Lett 555, 66-71.

Lai, E. M., Nair, U., Phadke, N. D., and Maddock, J. R. (2004) Proteomic screening and identification of differentially distributed membrane proteins in Escherichia coli, Mol Microbiol 52, 1029-1044.

Landau, E. M., and Rosenbusch, J. P. (1996) Lipidic cubic phases: a novel concept for the crystallization of membrane proteins, Proc Natl Acad Sci U S A 93, 14532-14535.

Lange, C., and Hunte, C. (2002) Crystal structure of the yeast cytochrome bc1 complex with its bound substrate cytochrome c, Proc Natl Acad Sci U S A 99, 2800-2805.

LaPorte, S. L., Forsyth, C. M., Cunningham, B. C., Miercke, L. J., Akhavan, D., and Stroud, R. M. (2005) De novo design of an IL-4 antagonist and its structure at 1.9 angstrom, Proc Natl Acad Sci U S A 102, 1889-1894. le Maire, M., Champeil, P., and Moller, J. V. (2000) Interaction of membrane proteins and lipids with solubilizing detergents, Biochim Biophys Acta 1508, 86-111.

Levy, S. B., and Marshall, B. (2004) Antibacterial resistance worldwide: causes, challenges and responses, Nat Med 10, S122-129.

Liu, X. (2008) Structure and regulation of aspartate pathway enzymes and deuteration effects on protein structure. Ph.D Dissertation, University of Toledo

186

Lyon, G. J., and Novick, R. P. (2004) Peptide signaling in Staphylococcus aureus and other Gram-positive bacteria, Peptides 25, 1389-1403.

Ma, D., Cook, D. N., Alberti, M., Pon, N. G., Nikaido, H., and Hearst, J. E. (1993) Molecular cloning and characterization of acrA and acrE genes of Escherichia coli, J Bacteriol 175, 6299-6313.

Martinez, J. L., Fajardo, A., Garmendia, L., Hernandez, A., Linares, J. F., Martinez- Solano, L., and Sanchez, M. B. (2009) A global view of antibiotic resistance, FEMS Microbiol Rev 33, 44-65.

Matern, Y., Barion, B., and Behrens-Kneip, S. (2010) PpiD is a player in the network of periplasmic chaperones in Escherichia coli, BMC Microbiol 10, 1-17.

McCoy, A. J., Grosse-Kunstleve, R. W., Storoni, L. C., and Read, R. J. (2005) Likelihood-enhanced fast translation functions, Acta Crystallogr D Biol Crystallogr 61, 458-464.

Miroux, B., and Walker, J. E. (1996) Over-production of proteins in Escherichia coli: Mutant hosts that allow synthesis of some membrane proteins and globular proteins at high levels, J Mol Biol 260, 289-298.

Moat, A.G., Foster, J.W. and Spector, M.P. (2002) Biosynthesis and Metabolism of Amino Acids, in Microbial Physiology. Wiley-Liss, Inc.

Mohanty, A. K., and Wiener, M. C. (2004) Membrane protein expression and production: effects of polyhistidine tag length and position, Protein Expres Purif 33, 311-325.

Molloy, M. P., Herbert, B. R., Slade, M. B., Rabilloud, T., Nouwens, A. S., Williams, K. L., and Gooley, A. A. (2000) Proteomic analysis of the Escherichia coli outer membrane, Eur J Biochem 267, 2871-2881.

Molloy, M. P., Phadke, N. D., Maddock, J. R., and Andrews, P. C. (2001) Two- dimensional electrophoresis and peptide mass fingerprinting of bacterial outer membrane proteins, Electrophoresis 22, 1686-1696.

Moore, R. A., Bocik, W. E., and Viola, R. E. (2002) Expression and purification of aspartate beta-semialdehyde dehydrogenase from infectious microorganisms, Protein Expr Purif 25, 189-194.

Newstead, S., Ferrandon, S., and Iwata, S. (2008) Rationalizing alpha-helical membrane protein crystallization, Protein Sci 17, 466-472.

Nichols, C. E., Dhaliwal, B., Lockyer, M., Hawkins, A. R., and Stammers, D. K. (2004) High-resolution structures reveal details of domain closure and "half-of-sites- reactivity" in Escherichia coli aspartate beta-semialdehyde dehydrogenase, J Mol

187

Biol 341, 797-806.

Orita, M., Ohno, K., and Niimi, T. (2009) Two 'Golden Ratio' indices in fragment-based drug discovery, Drug Discov Today 14, 321-328.

Ostermeier, C., Harrenga, A., Ermler, U., and Michel, H. (1997) Structure at 2.7 A resolution of the Paracoccus denitrificans two-subunit cytochrome c oxidase complexed with an antibody FV fragment, Proc Natl Acad Sci U S A 94, 10547- 10553.

Ostermeier, C., Iwata, S., Ludwig, B., and Michel, H. (1995) Fv fragment-mediated crystallization of the membrane protein bacterial cytochrome c oxidase, Nat Struct Biol 2, 842-846.

Ostrosky-Zeichner, L., Casadevall, A., Galgiani, J. N., Odds, F. C., and Rex, J. H. (2010) An insight into the antifungal pipeline: selected new molecules and beyond, Nat Rev Drug Discov 9, 719-727.

Otwinowski, Z., and Minor, W. (1997) Processing of X-ray diffraction data collected in oscillation mode, Method Enzymol 276, 307-326.

Ouyang, J., and Viola, R. E. (1995) Use of structural comparisons to select mutagenic targets in aspartate-beta-semialdehyde dehydrogenase, Biochemistry 34, 6394- 6399.

Pages, J. M., and Bolla, J. M. (1988) Assembly of the OmpF porin of Escherichia coli B. Immunological and kinetic studies of the integration pathway, Eur J Biochem 176, 655-660.

Pautsch, A., and Schulz, G. E. (1998) Structure of the outer membrane protein A transmembrane domain, Nat Struct Biol 5, 1013-1017.

Pavlovsky, A.G, Liu, X., Faehnle, C.F., Potente, N. and Viola, R. E. (2011) Structural characterization of selective inhibitors against homologous ASA dehydrogenase family, manuscript submitted.

Perkins, D. N., Pappin, D. J., Creasy, D. M., and Cottrell, J. S. (1999) Probability-based protein identification by searching sequence databases using mass spectrometry data, Electrophoresis 20, 3551-3567.

Potente, N. (2010) High resolution structural studies of vc-ASADH. Masters Thesis, University of Toledo

Pugsley, A. P., and Schnaitman, C. A. (1979) Factors affecting the electrophoretic mobility of the major outer membrane proteins of Escherichia coli in polyacrylamide gels, Biochim Biophys Acta 581, 163-178.

188

Prive, G. G. (2007) Detergents for the stabilization and crystallization of membrane proteins, Methods 41, 388-397.

Prive, G. G., and Kaback, H. R. (1996) Engineering the lac permease for purification and crystallization, J Bioenerg Biomembr 28, 29-34.

Ragkousi, K., Eichenberger, P., van Ooij, C., and Setlow, P. (2003) Identification of a new gene essential for germination of Bacillus subtilis spores with Ca2+-dipicolinate, J Bacteriol 185, 2315-2329.

Rath, A., Glibowicka, M., Nadeau, V. G., Chen, G., and Deber, C. M. (2009) Detergent binding explains anomalous SDS-PAGE migration of membrane proteins, Proc Natl Acad Sci U S A 106, 1760-1765.

Rocque, W. J., Coughlin, R. T., and McGroarty, E. J. (1987) Lipopolysaccharide tightly bound to porin monomers and trimers from Escherichia coli K-12, J Bacteriol 169, 4003-4010.

Rosenbaum, D. M., Cherezov, V., Hanson, M. A., Rasmussen, S. G., Thian, F. S., Kobilka, T. S., Choi, H. J., Yao, X. J., Weis, W. I., Stevens, R. C., and Kobilka, B. K. (2007) GPCR engineering yields high-resolution structural insights into beta2- adrenergic receptor function, Science 318, 1266-1273.

Sadowski, P.G., Groen, A.J., Dupree P., and Lilley K.S. (2008) Sub-cellular localization of membrane proteins, Proteomics, 8, 3991-4011.

Schafer, H., Nau, K., Sickmann, A., Erdmann, R., and Meyer, H. E. (2001) Identification of peroxisomal membrane proteins of Saccharomyces cerevisiae by mass spectrometry, Electrophoresis 22, 2955-2968.

Schirmer, T. (1998) General and specific porins from bacterial outer membranes, J Struct Biol 121, 101-109.

Schulz, G. E. (2000) beta-Barrel membrane proteins, Curr Opin Struct Biol 10, 443-447.

Shevchenko, A., Tomas, H., Havlis, J., Olsen, J. V., and Mann, M. (2006) In-gel digestion for mass spectrometric characterization of proteins and proteomes, Nat Protoc 1, 2856-2860.

Sleigh, S. H., Tame, J. R., Dodson, E. J., and Wilkinson, A. J. (1997) Peptide binding in OppA, the crystal structures of the periplasmic oligopeptide binding protein in the unliganded form and in complex with lysyllysine, Biochemistry 36, 9747-9758.

Smyth, D. R., Mrozkiewicz, M. K., McGrath, W. J., Listwan, P., and Kobe, B. (2003)

189

Crystal structures of fusion proteins with large-affinity tags, Protein Sci 12, 1313- 1322.

Sousa, R. (1995) Use of glycerol, polyols and other protein structure stabilizing agents in protein crystallization, Acta Crystallogr D Biol Crystallogr 51, 271-277.

Sperandeo, P., Lau, F. K., Carpentieri, A., De Castro, C., Molinaro, A., Deho, G., Silhavy, T. J., and Polissi, A. (2008) Functional analysis of the protein machinery required for transport of lipopolysaccharide to the outer membrane of Escherichia coli, J Bacteriol 190, 4460-4469.

Stenberg, F., Chovanec, P., Maslen, S. L., Robinson, C. V., Ilag, L. L., von Heijne, G., and Daley, D. O. (2005) Protein complexes of the Escherichia coli cell envelope, J Biol Chem 280, 34409-34419.

Stoll, V. S., Manohar, A. V., Gillon, W., MacFarlane, E. L., Hynes, R. C., and Pai, E. F. (1998) A thioredoxin fusion protein of VanH, a D-lactate dehydrogenase from Enterococcus faecium: cloning, expression, purification, kinetic analysis, and crystallization, Protein Sci 7, 1147-1155.

Surrey, T., Schmid, A., and Jahnig, F. (1996) Folding and membrane insertion of the trimeric beta-barrel protein OmpF, Biochemistry 35, 2283-2288.

Tamm, L. K., Arora, A., and Kleinschmidt, J. H. (2001) Structure and assembly of beta- barrel membrane proteins, J Biol Chem 276, 32399-32402.

Terpe, K. (2003) Overview of tag protein fusions: from molecular and biochemical fundamentals to commercial systems, Appl Microbiol Biotechnol 60, 523-533.

Tikhonova, E. B., and Zgurskaya, H. I. (2004) AcrA, AcrB, and TolC of Escherichia coli Form a Stable Intermembrane Multidrug Efflux Complex, J Biol Chem 279, 32116-32124.

Vagenende, V., Yap, M. G., and Trout, B. L. (2009) Mechanisms of protein stabilization and prevention of protein aggregation by glycerol, Biochemistry 48, 11084-11096. van Heijenoort, J. (2001) Recent advances in the formation of the bacterial peptidoglycan monomer unit, Nat Prod Rep 18, 503-519.

Viola, R. E. (2001) The central enzymes of the aspartate family of amino acid biosynthesis, Acc Chem Res 34, 339-349.

Viola, R. E., Faehnle, C. R., Blanco, J., Moore, R. A., Liu, X., Arachea, B. T., and Pavlovsky, A. G. (2011) The Catalytic Machinery of a Key Enzyme in Amino Acid Biosynthesis, J Amino Acids, 2011, 11 pp

190

Viola, R. E., Liu, X., Ohren, J. F., and Faehnle, C. R. (2008) The structure of a redundant enzyme: a second isoform of aspartate beta-semialdehyde dehydrogenase in Vibrio cholerae, Acta Crystallogr D Biol Crystallogr 64, 321-330.

Vogel, H., and Jahnig, F. (1986) Models for the structure of outer-membrane proteins of Escherichia coli derived from raman spectroscopy and prediction methods, J Mol Biol 190, 191-199.

Vogt, J., and Schulz, G. E. (1999) The structure of the outer membrane protein OmpX from Escherichia coli reveals possible mechanisms of virulence, Structure 7, 1301-1309. von Heijne, G. (1996) Principles of membrane protein assembly and structure, Prog Biophys Mol Biol 66, 113-139. von Heijne, G. (2007) The membrane protein universe: what’s out there and why bother?” Journal Intern Med, 261, 543-557.

Wadsten, P., Wohri, A. B., Snijder, A., Katona, G., Gardiner, A. T., Cogdell, R. J., Neutze, R., and Engstrom, S. (2006) Lipidic sponge phase crystallization of membrane proteins, J Mol Biol 364, 44-53.

Walian, P., Cross, T. A., and Jap, B. K. (2004) Structural genomics of membrane proteins, Genome Biol 5, 215.

Wallin, E., Tsukihara, T., Yoshikawa, S., von Heijne, G., and Elofsson, A. (1997) Architecture of helix bundle membrane proteins: an analysis of cytochrome c oxidase from bovine mitochondria, Protein Sci 6, 808-815.

Wallis, R., and Drickamer, K. (1999) Molecular determinants of oligomer formation and complement fixation in mannose-binding proteins, J Biol Chem 274, 3580-3589.

Walsh, T. R., and Howe, R. A. (2002) The prevalence and mechanisms of vancomycin resistance in Staphylococcus aureus, Annu Rev Microbiol 56, 657-675.

Waxman, D.J., Amanuma, H., and Strominger, J.L. (1982) Amino acid sequence homologies between Escherichia coli penicillin-binding protein 5 and class A beta-lactamases, FEBS Lett, 139, 159-163

Weiner, J. H., and Li, L. (2008) Proteome of the Escherichia coli envelope and technological challenges in membrane proteome analysis, Biochim Biophys Acta 1778, 1698-1713.

Weininger, U., Jakob, R. P., Kovermann, M., Balbach, J., and Schmid, F. X. (2010) The prolyl isomerase domain of PpiD from Escherichia coli shows a parvulin fold but is devoid of catalytic activity, Protein Sci 19, 6-18.

191

Winn, M. D., Isupov, M. N., and Murshudov, G. N. (2001) Use of TLS parameters to model anisotropic displacements in macromolecular refinement, Acta Crystallogr D Biol Crystallogr 57, 122-133.

Wohri, A. B., Johansson, L. C., Wadsten-Hindrichsen, P., Wahlgren, W. Y., Fischer, G., Horsefield, R., Katona, G., Nyblom, M., Oberg, F., Young, G., Cogdell, R. J., Fraser, N. J., Engstrom, S., and Neutze, R. (2008) A lipidic-sponge phase screen for membrane protein crystallization, Structure 16, 1003-1009.

Wu, C. C., and Yates, J. R., 3rd. (2003) The application of mass spectrometry to membrane proteomics, Nat Biotechnol 21, 262-267.

Wu, T., Malinverni, J., Ruiz, N., Kim, S., Silhavy, T. J., and Kahne, D. (2005) Identification of a multicomponent complex required for outer membrane biogenesis in Escherichia coli, Cell 121, 235-245.

Yankovskaya, V., Horsefield, R., Tornroth, S., Luna-Chavez, C., Miyoshi, H., Leger, C., Byrne, B., Cecchini, G., and Iwata, S. (2003) Architecture of succinate dehydrogenase and reactive oxygen species generation, Science 299, 700-704.

Zahedi, R. P., Sickmann, A., Boehm, A. M., Winkler, C., Zufall, N., Schonfisch, B., Guiard, B., Pfanner, N., and Meisinger, C. (2006) Proteomic analysis of the yeast mitochondrial outer membrane reveals accumulation of a subclass of preproteins, Mol Biol Cell 17, 1436-1450.

Zhan, Y., Song, X., and Zhou, G. W. (2001) Structural analysis of regulatory protein domains using GST-fusion proteins, Gene 281, 1-9.

Zhou, Y., Morais-Cabral, J. H., Kaufman, A., and MacKinnon, R. (2001) Chemistry of ion coordination and hydration revealed by a K+ channel-Fab complex at 2.0 A resolution, Nature 414, 43-48.

192

Appendix A

Mass Spectrometric Analysis and Database Searching

A1. Identification of KcsA Sample missed Protein hits Mass Score expect matches cleavages 4-142-1 0 GLND_NITWN (1) 104312 54 1.5 5 (Buffer) 0 ISPG_ECO24 (29) 40658 37 71 3 1 MSBA_XYLFA (1) 64993 52 1.8 6 1 KCSA_STRCO (5) 17683 48 5 4 2 RS19_MYCS5 (1) 10280 54 1.3 4 2 KCSA_STRCO (12) 17683 39 39 4 4-142-2 0 LOLB_ECO7I (1) 23626 59 0.38 4 (SDS) 0 KCSA_STRCO (27) 17683 41 24 3 0 YEIT_ECO57 (31) 45092 40 35 3 1 PYRF_FLAJ1 (1) 30582 49 4 4 1 RL9_ECO24 (7) 15759 47 6.5 3 1 LOLB_ECO7I (34) 23626 45 9.6 4 2 ISPG_THEEB (1) 44512 52 1.8 5 3 ATPA_DESMR 54645 54 1.3 6 3 RL9_ECO24 (32) 15759 34 1.4e+02 3 4-142-3 0 APT_TREPA (1) 21005 79 (Sig) 0.0045 4 (Triton) 0 KCSA_STRCO (4) 17683 48 5.4 3 0 LOLB_ECO7I (10) 23626 45 9.8 3 0 ATPB_ECO24 (13) 50352 45 10 4 1 RUVB_BURM7 (1) 39257 65 0.11 5 2 RUVB_THIDA (1) 37276 65 0.1 6 3 RIMM_ALISL (1) 20085 70 (Sig) 0.034 5 3 KCSA_STRCO (7) 17683 52 1.8 5 4-142-4 0 MIXTURE 1 (1) 199 (Sig) 4.1e-15 20 (DM) ATPB_ECO24 50352 170 (Sig) 3.3e-12 12 TNAA_ECO24 53146 84 (Sig) 0.0014 8 0 ATPB_ECO24 (2) 50352 170 (Sig) 3.3e-12 12 0 TNAA_ECO24 (3) 53146 84 (Sig) 0.0014 8 1 MIXTURE 1 (1) 156 (Sig) 8.2e-11 23 ATPB_ECO24 50352 131 (Sig) 2.6e-8 13 TNAA_ECO24 53146 76 (Sig) 0.0088 10 1 MIXTURE 2 (2) 141 (Sig) 2.6e-9 22 ATPB_ENT38 50183 116 (Sig) 8.2e-7 12 TNAA_ECO24 53146 76 (Sig) 0.0088 10 1 ATPB_ECO24 50352 131 (Sig) 2.6e-08 13 2 ATPB_ECO24 (1) 50352 112 (Sig) 2.1e-06 13 3 MIXTURE 1 (1) 109 (Sig) 4.1e-06 24 ATPB_ECO24 50352

193

TNAA_ECO24 53146 3 MIXTURE 2 (2) 109 (Sig) 4.1e-06 24 ATPB_ENT38 50183 TNAA_ECO24 53146 3 ATPB_ECO24 (3) 50352 95 (Sig) 0.0001 13 4-142-5 0 ATPB_ECO24 (1) 50352 73 (Sig) 0.017 6 (OG) 0 LOLB_ECO7I (50) 23626 60 0.36 5 1 RL19_BARHE (1) 16504 80 (Sig) 0.0035 7 1 KCSA_STRCO (6) 17683 56 0.09 6 1 TNAA_ECO24 (8) 53146 55 1.1 8 1 ATPB_ECO24 (31) 50352 54 1.2 7

2 KCSA_STRCO (1) 17683 71 (Sig) 0.025 7 4-142-5 2 TNAA_ECO24 (6) 53146 51 2.6 10 3 KCSA_STRCO (1) 17683 84 (Sig) 0.0012 8 3 TNAA_ECO24 (34) 53146 46 7.3 10 4-142-6 0 ATPB_ECO24 (1) 50352 66 0.082 5 0 KCSA_STRCO (27) 17683 54 1.5 4 0 ATPB_ECOHS (31) 50325 52 2.3 4 1 KCSA_STRCO (1) 17683 62 0.2 5 1 LOLB_ECO7I (17) 23626 42 20 4 1 ATPB_ECO24 (40) 50352 39 41 5 2 KCSA_STRCO (1) 17683 74 (Sig) 0.012 7 2 YAIW_ECOLI (43) 40563 41 28 7 3 KCSA_STRCO (1) 17683 90 (Sig) 0.00033 8 3 TRAN_ECOLI (19) 66949 42 21 6 3 LOLB_ECO7I (35) 23626 38 57 4 4-142-7 0 HSLV_DESAP (1) 18867 34 1.2e+02 2 1 QUEC_RICCK (1) 25353 39 43 3 1 YCIN_ECO57 (3) 9380 32 1.8e+02 2 2 QUEC_RICCK (1) 25643 36 88 3 2 YCIN_ECO57 (16) 9380 30 3.5e+02 2 3 QUEC_RICCK (1) 25643 35 1.1e+02 3 3 YCIN_ECO57 (3) 9380 29 4.5e+02 2 4-142-8 0 HRCA_BURP8 (1) 37293 44 12 3 0 YCIN_ECOL6 (3) 9380 37 63 2 0 YJAA_ECOLI (8) 14438 36 92 2 0 DNAT_ECO7I (12) 19457 33 1.5e+02 3 1 METN_CHLAB (1) 38182 50 3.4 4 1 YCIN_ECO57 (28) 9380 31 2.4e+02 2

2 DNLJ_RICAH (1) 77779 52 2.1 7 2 ARLY_ECO27 (44) 50620 27 6.3e+02 6 4-142-9 0 KCSA_STRCO (1) 17683 69 (Sig) 0.043 4 0 LOLB_ECO7I (3) 23626 64 0.12 4 0 LOLB_ECO24 (10) 23594 45 11 3 0 ATPB_ECO24 (42) 50352 43 18 4 1 KCSA_STRCO (1) 17683 57 0.7 5 1 LOLB_ECO7I (6) 23626 52 2.1 4 1 TRAP_ECOLI (43) 22295 37 62 3 2 RPOB_CLOP1 (1) 139044 59 0.45 9 2 KCSA_STRCO (4) 17683 53 1.6 5 2 LOLB_ECO7I (6) 23626 48 5.3 4 2 TNR2_ECOLX (13) 21316 42 20 5 2 PHYDA_ECO27 (29) 51649 38 50 4

194

A2. Identification of Constitutive Proteins from Total Membrane.

MS Sample # Band ID Protein Hit Score Expect Matches Mr (Da) (% cov) 9-165-1 165-1 M9 FDOG_ECOLI 65 (S) 0.0074 11 (12) 113291 9-165-2 165-1 Fos HTRL_ECOLI 36 5.5 4 (15) 33669 9-165-3 165-3,4 SDS Mixture 1 106 (S) 5.7e-07 29 RNE_ECOLI 70 (S) 0.0023 11 (18) 118357 EFTU1_ECO24 62 (S) 0.016 8 (28) 43430 BGAL_ECOLC 61 (S) 0.017 11 (16) 117366 9-165-4 165-3 Fos Mixture 1 168 (S) 3.6e-13 23 BGAL_ECO24 127 (S) 4.6e-09 16 (19) 117366 EFTU1_ECO24 65 (S) 0.0068 7 (22) 43430 9-165-5 165-5 Buffer NUOG_ECO57 60 (S) 0.025 9 (14) 101226 9-165-6 165-5 M9 ODP1_ECO57 57 (S) 0.042 5 (8) 99954 9-165-7 165-5 LDAO Mixture 1 162 (S) 1.4e-12 26 ODP1_ECO57 141 (S) 1.8e-10 17 (28) 99954 NUOG_ECO57 54 0.087 9 (13) 101226 9-165-8 165-6 Buffer Mixture 1 125 7.2e-09 30 EFG_ECO24 93 (S) 1.2e-05 15 (27) 77706 NUOG_ECOL6 84 (S) 8.3e-05 17 (19) 101216 9-165-9 165-6 OG EFG_ECO24 63 (S) 0.012 7 (16) 77706 9-165-10 165-6 M9 Mixture 1 161 (S) 1.8e-12 23 EFG_ECO24 122 (S) 1.4e-08 14 (32) 77706 DHSA_ECO57 86 (S) 5.5e-05 10 (19) 65019 9-165-11 165-6 Fos EFG_ECO24 91 (S) 1.9e-05 11 (23) 77706 9-165-12 165-8 M9 DNAK_ECOLI 75 (S) 0.00078 9 (17) 69130 9-165-13 165-8 LDAO DNAK_ECOLI 158 (S) 3.6e-12 13 (30) 69130 9-165-14 165-8 Tween DNAK_ECOLI 125 (S) 7.2e-9 11 (23) 69130 9-165-15 165-10 Buffer NUOCD_ECOLI 61 (S) 0.018 6 68425 9-165-16 165-10 OG Mixture 1 145 (S) 7.2e-11 26 DHSA_ECO57 118 (S) 3.6e-08 15 (30) 65019 PPID_ECOL6 76 (S) 0.0006 12 (21) 68108 9-165-17 165-10 Cy5 DHSA_ECO57 83 (S) 0.00012 8 (15) 65019 9-165-18 165-10 M9 DHSA_ECO57 91 (S) 1.7e-05 8 (16) 65019 9-165-19 165-10 Fos DHSA_ECO57 74 (S) 0.00091 7 (13) 65019 9-165-20 165-11 OG Mixture 1 107 (S) 4.6e-07 19 90612 YAET_ECO24 85 (S) 7.6e-05 11 (18) 9-165-20 165-11 OG DHSA_ECO57 58 (S) 0.032 8 (13) 65019 9-165-21 165-12 LDAO 1. OPPA_ECOLI 70 (S) 0.0023 8 (19) 60977 2. PCKA_ECO24 57 (S) 0.043 7 (17) 59868 9-165-22 165-13 Buffer ATPA_ECO24 183 (S) 1.1e-14 13 (32) 55420 9-165-23 165-13 LDAO ATPA_ECO24 185 (S) 7.2e-15 14 (34) 55420 9-165-24 165-14 OG Mixture 1 105 (S) 7.2e-07 18 ATPA_ECO24 91 (S) 2e-05 10 (26) 55420

9-165-24 165-14 OG GLPD_ECOLI 76 (S) 0.00063 10 (19) 56889 9-165-25 165-14 Cy5 Mixture 1 161 (S) 1.8e-12 22 GLPD_ECOLI 127 (S) 4.6e-09 14 (28) 56889 ATPA_ECO24 91(S) 1.9e-05 10 (27) 55420 9-165-26 165-14 Fos Mixture 1 147 (S) 4.6e-11 22 GLPD_ECOLI 119 (S) 2.9e-08 14 (29) 56889 ATPA_ECO24 86(S) 6e-05 10 (27) 55420

9-165-27 165-15 Buffer Mixture 1 196 (S) 5.7e-16 32 ATPB_ECO24 140 (S) 2.3e-10 16 (51) 50352 TNAA_ECO24 89 (S) 2.7e-05 16 (35) 53146

9-165-29 165-15 Cy5 ATPB_ECO24 78 (S) 0.0004 6 (21) 50352

9-165-30 165-15 M9 ATPB_ECO24(0) 78 (S) 0.00032 6 (22) 50352 TNAA_ECO24(1) 66 (S) 0.0064 8 (19) 53146 9-165-31 165-15 LDAO Mixture 1 193 (S) 1.1e-15 27 ATPB_ECO24 126 5.7e-09 14 (41) 50352 TNAA_ECO24 101 1.8e-06 13 (30) 53146

195

9-165-32 165-15 Fos ATPB_ECO24 104 9.1e-07 8 (27) 50352 9-165-33 165-17 OG Mixture 1 93 (S) 1.1e-05 12 HFLK_ECO57 74 (S) 0.00091 7(17) 45517 165-17 OG DACC_ECOLI 63 (S) 0.011 5 (20) 43639 9-165-34 165-16 M9 RBSB_ECOLI 50 0.25 3(15) 30931 9-165-35 165-17 M9 AHPC_ECO57 65 (S) 0.0081 4 (33) 20864 9-165-36 165-18 Buffer CYSK_ECOLI 58 (S) 0.041 4 (21) 34526 9-165-37 165-18 M9 KDGK_ECOLI 85(S) 7.2e-05 7 (25) 34231 9-165-38 165-19 LDAO 1.KDGK_ECOLI 61 (S) 0.019 5 (21) 34231 2.HSLO_ECOLI 46 0.56 4 (17) 32862 9-165-39 165-20 M9 EFTS_ECOLI 69 0.0029 5(21) 30520 9-165-40 165-21 Cy5 1.RBSB_ECOLI 77 (S) 0.00048 6 (26) 30931 2. ATPG_ECOLI 42 1.5 4 (17) 31673 9-165-41 165-22 M9 1.KDUD_ECOLI 56 (B) 0.051 4 (23) 27169 2.BLAT_ECOLX 36 5.5 4 (17) 31669 9-165-42 165-22 LDAO ATPG_ECO27 76 (S) 0.00052 6 (24) 31673 9-165-43 165-23 OG Mixture 1 103 (S) 1.1e-06 10 OMPP_ECOLI 89 (S) 2.7e-05 6 (26) 35477 DHSB_ECOLI 57 (S) 0.047 4 (21) 27390

9-165-44 165-24 M9 Mixture 1 106 (S) 5.7e-07 10 DHSB_ECOLI 82 (S) 0.00013 6 (28) 27390 SODM_ECO57 68 (S) 0.0033 4(37) 23065

9-165-45 165-24 LDAO SODM_ECO57 60 (S) 0.021 4(27) 23065

196

A3. Identification of Constitutive Proteins from Outer Membrane.

Sample Protein Mass score expect matches (% Cov) 9-121-0 ISPE 31156 39 (NS) 2.9 3 (16) 9-121-1 ATPA (1) 55420 113 1.1e-07 12 (31) (SDS) OMPF (2) 39309 60 0.025 7 (27) 9-121-2 OMPF (1) 39309 85 7.6e-05 8 (25) (SDS) 9-121-3 MIX (1) 107 4.6e-07 18 (SDS) OMPF (1, 2) 39309 86 5.2e-05 9 (38) ATPB (1,3) 50352 65 0.0071 9 (31) TOLC (4) 53708 56 0.057 8 (27) 9-121-4 OMPF (1) 39309 59 0.028 4 (21) (SDS) 9-121-5 OMPF (1) 39309 86 6.3e-05 7 (22) (SDS) 9-121-6 OMPA (1) 37294 92 1.3e-05 7 (28) (SDS) 9-121-7 FADL (1) 48512 69 0.0029 6 (19) (SDS) 9-121-8 YFIO (1) 27870 94 1e-05 9 (41) (SDS) 9-121-9 DPS (1) 18684 108 3.6e-07 10 (61) (SDS) 9-121-10 YDIU (1) 54580 46 (NS) 0.57 6 (17) (SDS) 9-121-11 MIX (1) 78 0.00033 10 (SDS) OMPX (1,2) 18649 63 0.012 5 (33)

9-121-11 YBJP (1,3) 19211 59 0.026 5 (39) (SDS) 9-121-12 CITD (1) 10799 38 (NS) 3.5 3( 30) (SDS) 9-121-13 TOLC (1) 53708 102 1.4e-06 11 (29) (DM) 9-121-14 OMPF (1) 39309 71 0.0019 7 (27) (DM) 9-121-15 OMPF (1) 39309 48 (NS) 0.35 5 (20) (DM) AAEA (2) 34751 46 (NS) 0.52 5 (21)

DCUS (3) 60629 45 (NS) 0.76 6 (13)

NUOG 101226 42 (NS) 1.3 7 (9)

9-121-16 MIX 130 2.3e-09 15 (DM) OMPF (1,2) 39309 88 4e-05 7 (40) ATPA (1,3) 55420 81 0.00017 8 (25) 9-121-17 OMPF (1) 39309 90 2.5e-05 7 (38) (DM) 9-121-18 OMPF (1) 39309 52 (NS) 0.13 4 (21) (DM) 9-121-19 ATPB (1) 50352 173 1.1e-13 13 (42) (DM) 9-121-20 ACRA (1) 42229 110 2.3e-07 11 (36) (DM) 9-121-21 NLPB (1) 36878 102 1.4e-06 10 (40) (DM) 9-121-22 OMPA (1) 37294 95 7.2e-06 9 (38) (DM) 9-121-23 FADL (1) 48509 55 (B) 0.066 5 (18)

197

(DM)

9-121-24 YEEA (1) 40162 42 (NS) 1.6 4 (11) (DM) CDH (2) 28628 34 (NS) 8.9 3 (14)

9-121-25 RL5 (1) 20347 57 0.044 8 (42) (DM) TSX (2) 33568 45 (NS) 0.71 6 (32)

PAL (4) 18870 41 (NS) 1.7 6 (35)

9-121-26 RL9 (1) 15759 58 0.036 6 (42) (DM) YBJP (2) 19211 37 (NS) 4.3 4 (32)

PAL (5) 18870 35 (NS) 6.8 5 (34)

9-121-27 CVAA (1) 47315 41 (NS) 1.9 4 (12) (DM) 9-121-28 YEEA (1) 40162 32 (NS) 14 3 (7) (DM) 9-121-29 YEEA (1) 40162 31 (NS) 14 3 (7) (OG) 9-121-30 9-121-31 OMPF (1) 39309 102 1.4e-06 8 (26) (OG) 9-121-32 MIX 126 5.7e-09 19 (OG) OMPF (1,2) 39309 102 1.4e-06 8 (26)

(OG) NLPB (1,3) 36878 74 0.001 9 (36) 9-121-33 OMPA (1) 37294 98 3.9e-06 7 (28) (OG) 9-121-34 FADL (1) 48509 42 (NS) 1.6 4 (13) (OG) 9-121-35 YBJP (1) 19211 48 (NS) 0.39 5(31)

(OG) MIPA (3) 27813 38 (NS) 3.4 4 (31)

(OG) TSX (4) 33568 33 (NS) 11 4 (24)

9-121-36 OMPX (1) 18649 72 0.0013 5 (33) (OG) 9-121-37 MBEA (1) 58005 43 (NS) 1.3 6 (12) (OG) 9-121-38 OMPF (1) 39309 53 (NS) 0.11 4 (19) (Cy5) 9-121-39 MIX 144 9.1e-11 20 Cy5 ATPA (1,2) 55420 100 2.6e-06 11 (29)

OMPF (1,3) 39309 94 8.5e-06 10 (47)

9-121-40 OMPF (1,) 39309 138 3.6e-10 11 (54) (Cy5) 9-121-41 MIX 102 1.4e-06 16 Cy5 OMPF (1,2) 39309 76 0.00052 7 (40)

LPTD (1,3) 89836 65 0.0066 9 (14)

9-121-42 ATPB (1) 50352 191 1.8e-15 16 (51) (Cy5) 9-121-43 ACRA (1) 42229 85 7.1e-05 8 (31) (Cy5) 9-121-44 OMPA (1) 37294 117 4.6e-08 8 (35) (Cy5) 9-121-45 NOHB (1) 20474 76 0.00057 6 (56) (Cy5) FADL (2) 48509 54 (B) 0.1 77 (22)

198

9-121-46 TSX (1) 33568 89 3.1e-05 6 (32) (Cy5) 9-121-47 OMPX (1) 18649 89 3e-05 6 (52) (Cy5) YBJP (2) 19211 33 (NS) 11 5 (39)

9-121-48 OMPF (1) 39309 40 (NS) 2.5 3 (15) (M9) YBIP (2) 59901 35 (NS) 7.4 3 (10)

9-121-49 OMPF (1) 39309 59 0.029 4 (21) (M9) 9-121-50 OMPA (1) 37294 106 5.7e-07 8 (30) (M9) MHPA (2) 62492 64 0.01 6 (15)

9-121-51 MEPA (1) 30525 41 (NS) 2 4 (27) (M9) 9-121-52 MIX 74 0.00087 12 (M9) YBJP (1,2) 19211 60 0.021 6 (42)

PAL (1,3) 18870 59 0.029 6 (44)

9-121-53 OMPX (1) 18649 50 (NS) 0.025 4 (28) (M9) 9-121-54 ISPE (1) 31138 38 (NS) 3.4 4 (21) (LDAO FTSH (2) 70763 35 7.6 5 (11)

9-121-55 YAET (1) 90612 44 0.93 4 (9) 9-121-56 MIX 127 4.6e-09 20 LDAO ATPB (1,2) 50352 85 6.9e-05 10 (31)

ATPA (1,3) 55420 82 0.00014 10 (27)

9-121-57 YIBA (1) 32318 66 0.0063 7 (25)

NLPB (2) 36878 33 (NS) 12 4 (18)

9-121-58 YHGF (1) 85357 54 (NS) 0.095 5 (9) LDAO MSCM (2) 46631 36 (NS) 5.9 3 (11)

ATPG (3) 31673 32 (NS) 14 5 (18)

9-121-59 ISPE (1) 31138 41 (NS) 1.9 4 (21) (Fos) OMPF (3) 39309 28 (NS) 34 3 (15) 9-121-60 ACRA (1) 42229 57 0.043 7 (24) (Fos) NUOF (3) 49783 40 (NS) 2.1 7 (18)

9-121-61 ATPG (1) 31673 65 0.0076 6 (21) (Fos) 9-121-62 ARND (1) 33360 42 (NS) 1.5 3 (17) (Fos) 9-121-63 DNAT (1) 19385 44 (NS) 0.91 4 (29) (Fos) 9-121-64 TRBG 9295 42 (NS) 1.4 3 (25) (Fos) 9-121-65 PCOA 67322 52 (NS) 0.13 5 (12) (Tween) 9-121-66 HCAD 44329 44 (NS) 0.95 5 (18) (Tween) 9-121-67 NUOG 101226 68 0.0033 9 (12) (Tween)

199

9-121-68 FTSH 70763 72 0.0014 10 (21) (Tween) 9-121-69 MIX 100 2.6e-06 10 Tween ATPB (1,2) 50352 71 0.0019 5 (22)

ATPA (1,3) 55420 64 0.0093 5 (17)

9-121-70 NUOF (1) 49783 52 (NS) 0.15 7 (14) (Tween) 9-121-71 CVAA 47315 52 (NS) 0.14 6 (17) (Tween) 9-121-72 PRIM (1) 65915 58 0.4 7 (14) (Tween) EFTU1 (2) 43430 35 (NS) 7.2 4 (14)

9-121-73 ATPG (1) 31673 44 0.98 4 (18) (Tween) 9-121-74 YNEG (1) 13767 46 (NS) 0.59 4 (21) (Tween) METN (2) 38055 40 (NS) 2.2 4 (20)

PAL (3) 18870 34 (NS) 8.7 4 (31)

9-121-75 DHSB (1) 27390 49 (NS) 0.28 5 (23) (Tween) 9-121-76 9-121-77 RS1 (1) 61237 108 3.6e-07 13 (27) (Buffer) 9-121-78 CH601 (1) 57467 129 2.9e-09 11 (27) (Buffer) 9-121-79 ATPA (1) 55420 140 2.3e-10 12 (31) (Buffer) 9-121-80 ATPB (1) 50352 129 2.9e-09 11 (33) (Buffer) 9-121-81 CVAA (1) 47315 44 (NS) 0.93 4 (12) (Buffer) EFTS (2) 30520 37 (NS) 4.6 3 (10)

9-121-82 RHO (1) 47033 47 (NS) 0.48 4 (13) (Buffer) 9-121-83 OMPA (1) 37294 47 (NS) 0.46 5 (21)

Buffer RBSB (2) 30931 35 (NS) 6.9 4 (18)

ATPG (3) 31673 32 (NS) 14 4 (18)

9-121-84 TRAG (1) 102523 37 (NS) 4.5 5 (7) (LDAO) PET (2) 139858 32 (NS) 13 5 (6)

200