THE ASSEMBLY AND INTERACTIONS OF MREB IN THE MAINTENANCE OF CELL SHAPE IN CAULOBACTER CRESCENTUS

A DISSERTATION SUBMITTED TO THE DEPARTMENT OF AND THE COMMITTEE ON GRADUATE STUDIES OF IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

Natalie Anne Dye May, 2010

© 2010 by Natalie Anne Dye. All Rights Reserved. Re-distributed by Stanford University under license with the author.

This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/bg008yn0701

ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Julie Theriot, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Lucille Shapiro, Co-Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

James Spudich

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Aaron Straight

Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.

iii ABSTRACT This work focuses on the mechanism by which MreB contributes to the maintenance of cell shape in the gram-negative alpha-proteobacterium Caulobacter crescentus. The mreB encodes a that resembles , a eukaryotic cytoskeletal protein. Previously, it was shown that mreB is required to maintain a rod-like shape and localizes to a helical pattern near the cytoplasmic membrane. Here, we show that MreB is associated with regions of active growth in Caulobacter, as mutant strains that mislocalize MreB to the cell poles direct new growth at or near the poles. We present evidence to suggest that MreB contributes to the determination of proper length, width, and curvature through partially distinct mechanisms. The determination of proper width involves the essential MreC and Pbp2, which are encoded in the mreB operon. While MreB and MreC are both required to position the cell wall transpeptidase Pbp2 along the lateral sidewalls and away from midcell, the two do not colocalize and each can maintain its localization in the absence of the other. When MreB is mislocalized to the poles, MreC and Pbp2 do not follow. These data argue against the idea that MreB provides a scaffold-like structure to localize enzymes that directly modify the cell wall. The determination of proper curvature, involves the -like protein, . We identify a putative binding site on MreB for Crescentin or other curvature-mediating factors. We also show that the extent to which the subcellular localization of MreB changes over the cell cycle is correlated with cell size, indicating that MreB is involved in the coordination between elongation and division. In addition, we show that in vitro purified MreB spontaneously forms very stable polymers in the presence or absence of nucleotide. These polymers are globular or amorphous and only filamentous when placed on a highly positively charged surface of Poly-L-lysine. These in vitro data suggest that MreB is likely to be regulated at the disassembly step in the cell and that the cellular environment may influence the structure of MreB polymers. Lastly, we present biochemical evidence to support the existence of a disassembly factor in cytoplasmic Caulobacter extract. Together our data suggest that the maintenance of the crescent-rod cell shape in Caulobacter is the result of a complicated balance between MreB’s dynamic subcellular localization, polymeric structure, and communication with cellular components.

iv

For my always loving and supportive parents

John and Donna Dye

v

ACKNOWLEDGEMENTS

This thesis is the culmination of almost seven years of work. During this time, I have had the pleasure of working with so many truly brilliant, creative, and knowledgeable scientists. I would not be the scientist I am today without the generous help of my mentors and colleagues. I would like to take this opportunity to thank the following persons. First and foremost, I must thank my joint thesis advisors, Julie Theriot and Lucy Shapiro. These two women are phenomenal scientists and individuals, and it has been a privilege and a pleasure to work with them. While they have very different scientific perspectives, styles, and interests, they share an inspiring enthusiasm for scientific research and a genuine care for their students. Thank you both for the guidance, mentorship, motivation and inspiration you have given me during this process. In addition to my advisors, I received considerable guidance from the other members of my committee Aaron Straight and Jim Spudich. I very much appreciate the advice and feedback that I received from each of them over the years. I am also thankful for KC Huang, who agreed to be my chair and has given me significant feedback on this work. Harley McAdams, though not an official committee member or advisor, has also provided an interesting and important point of view on this project. I’d also like to thank John Perrino at the EM facility in the Beckman center here at Stanford for guidance with the electron microscopy. I must acknowledge my collaborators in this project. Zachary Pincus was a graduate student with Julie when I first started this project, and even after he left Julie’s lab to become a postdoc thousands of miles away, he remained an active participant in this research. His contributions to the work presented in Chapters 2 and 3 were significant, and these stories would not have been as mature as they are now without his expertise in computation and statistics. Isabelle Fisher contributed to this project as a high school student. Though lacking the many years of formal training in biology, Isabelle was able to grasp complicated concepts and make me think about this subject in new ways. She was a joy to teach and work with, and I hope to see her continue in science. Enrique de la Cruz and his graduate student Kendra Frederick hosted me in their lab at Yale

vi

University for a short time to work with purified MreB. While this collaboration was brief and did not end up yielding publishable data, it was a valuable experience for me to learn specific techniques from them and get their perspective on this project. I have received countless ideas, feedback, reagents, protocols, and emotional support from the members of the Theriot and Shapiro labs, and as such I would like to thank everyone with whom I overlapped. In particular in the Shapiro/McAdams lab: Zemer Gitai, who was my rotation advisor, Antonio Iniesta, Grant Bowman, Erin Goley, Jerod Ptacin, Esteban Toro, Joseph Chen, Patrick McGrath, and Virginia Kalogeraki. In the Theriot lab: Kinneret Keren, Karine Gibbs, Aretha Fiebig, Guy Ziv, David Hallidan, Greg Allen, Mark Tsuchida, and Steph Weber. A very special thank you goes to Susanne Rafelski, who was my mentor in the Theriot lab when I first started and has remained a dear friend. I owe a huge huge huge debt of gratitude to Matthew Footer. He has taught me so much about biochemistry, protein purification, optics, cooking, and more. Thank you for always being so friendly and generous with your time. I think all of us agree that the lab would fall apart without you! I must also especially thank my friend and labmate Erin Barnhart, who has sat next to me almost my entire time in graduate school. She has been my sounding board, my colleague, my friend, and at times my entertainment . Erin also read almost this entire thesis and gave me thoughtful feedback, all while writing her own thesis. You are a talented, fun, intelligent, courageous, and goofy individual and it’s been a pleasure sharing this experience with you. Thank you. I would also like to thank all of the past and present members of the Biochemistry and Developmental Biology departments here at Stanford for making it a collegial and enjoyable place to work. I must specifically acknowledge the friendly, patient, and helpful support staff in these departments, in particular Joella Ackerman, Karen Butzman, Tara Trim, Todd Galitz, and Shedrick Watts. Special thanks to my classmates and friends in the department Ryan Nottingham, Ian Brennan, Kirstin Milks, Eric Espinosa, and especially Kristina Godek. Kristina has been a dear friend and colleague. Her endurance and dedication to science is an inspiration, and I wish her all the success her hard work deserves. I will miss having her

vii right down the hall. She also provided significant feedback on this thesis. I’d also like to thank my friends Dina Finan, Duane Baxter, Michael Costa, and Corey Meyer, who provided me with much wine and amazing home-cooked meals. In my time at Stanford, I have been a part of a wonderful network of female scientists and engineers, and I would like to thank the following for their considerable advice and support in both professional and personal manners: Megan Young, Jennifer Zamanian, Hyejun Ra, Davie Yoon, Misty Davies, Esther Chen, Liv Walter, Sabina Stefania, and Lisa Moore. Through my involvement with the Alpine club, I became part of a fantastic climbing community, which has enriched my experience here at Stanford. Ironically, during my toughest times in graduate school, I enjoyed the greatest adventures in the mountains. Thank you to all my climbing partners for helping me to achieve balance and maintain my sanity! Special thanks to Sarah Inwood, Sharon Lindsey West, Brian Cox, Mary Hollendoner, Linnea Williams, Ken Duncan, Alex and Tanya Nees, Christina Jeffrey, Joe Cackler, Blase Illuiano, Alina Garbuzov, Emily Rains, and my dear Chris McGuinness. To my roommates, who have seen me at my worst and my best, I give a sincere thank you. Much love to Miriam Wenger, Patricia Yam, Rob Maestretti, Kiersten Lane, and Dan Santos. Last, but most certainly not least, my family: John, Donna and Derek Dye. I cannot express how much the support from my family has meant to me. You are wonderful people, and I feel so lucky to have you in my life. I love you so much.

viii

TABLE OF CONTENTS

Abstract ...... iv Acknowledgements ...... vi List of Tables ...... xvi List of Figures ...... xvii

Chapter 1: Introduction ...... 1

From molecules to cells ...... 2 Thesis Overview ...... 4 Building a wall ...... 5 The cell wall ...... 6 Brick-makers and Brick-layers ...... 7 Growth of the sacculus and maintenance of form ...... 8 Alternating elongation and division ...... 10 Mechanisms for establishing intracellular organization in ...... 12 Finding the middle: Regulating the localization of FtsZ ...... 12 Finding the Pole ...... 14 Cytoskeletal proteins in bacteria ...... 17 The FtsZ homolog ...... 17 Bacterial actin homologs ...... 18 Crescentin, an intermediate filament-like protein ...... 19 The MinD/ParA family of bacterial-specific cytoskeletal proteins ...... 20 The push and the pull of the bacterial ...... 22 A push in the right direction ...... 22 A mitotic-like pull ...... 26 Dynamic scaffolding ...... 27 FtsZ: organizing the division apparatus ...... 28 MamK: lining up magnetosensing organelles ...... 29 The of complexity ...... 29

ix

How does MreB function? ...... 32

Chapter 2:Two independent spiral structures ...... 35

Abstract ...... 36 Introduction ...... 37 Results ...... 39 MreC is helical in Caulobacter ...... 39 MreC does not colocalize with MreB ...... 39 mreC is essential and required for proper cell shape ...... 41 MreC and MreB localize independently of one another ...... 41 Pbp2 forms a spiral-like pattern that partially colocalizes with MreC ...... 43 Pbp2 mislocalizes to the division plane in the absence of either MreB or MreC structures ...... 43 FtsZ depletion prevents A22-induced mislocalization of GFP-Pbp2 ...... 45 Quantitative shape analysis demonstrates that a FtsZ-depletion also prevents A22- induced shape defect ...... 46 Discussion ...... 48 Acknowledgements ...... 50 Author Contributions ...... 51 Materials and Methods ...... 51 Bacterial growth conditions and strain construction ...... 51 Depletions ...... 51 Microscopy ...... 52 PCA shape analysis ...... 52 Supplemental Text ...... 53 Creation of fluorescently-tagged constructs ...... 53 Creation of MreC-depletion strain and assessment of viability ...... 54 A22 treatments ...... 54 PCA shape analysis ...... 54 Statistical Analysis ...... 56 Supplemental figures ...... 60

x

Chapter 3: Modifying morphology with mutations in mreB ..... 65

Abstract ...... 66 Introduction ...... 67 Results ...... 69 Isolating A22-resistant Caulobacter ...... 69 Measuring the shape phenotype of A22-resistant Caulobacter ...... 73 Clustering A22-resistant Caulobacter strains by cell shape ...... 75 A variable width phenotype ...... 80 Determining the cellular localization of MreB mutants ...... 83 Analyzing MreB localization in subcellular regions ...... 86 Distinguishing A22-resistant mutants by subcellular localization patterns ...... 88 Characterizing the cell cycle regulation of the mutant localization patterns...... 90 Correlating MreB localization with cell shape ...... 97 Correlating crescentin localization with curvature ...... 101 Localizing the peptidoglycan transpeptidase Pbp2 in selected mutant strains ...... 102 Changing the distribution of mutant MreB with A22 ...... 103 Discussion ...... 109 A22 resistant Caulobacter strains can have distinct cell shapes and MreB localization patterns ...... 109 Mutant residues cluster around the ATP-binding site ...... 112 The accumulation of MreB at the poles ...... 113 The dynamic subcellular localization of MreB ...... 116 A model for the role of nucleotide in regulating MreB dynamics ...... 116 Polar MreB=polar cell wall synthesis? ...... 118 How do changes in MreB localization affect cell shape?...... 119 Future directions ...... 120 Author Contributions ...... 121 Acknowledgements ...... 121 Materials and Methods ...... 122 Bacterial growth conditions ...... 122 Selection of A22-resistant mutants ...... 122

xi

Imaging the shapes of A22-resistant Caulobacter ...... 123 Fluorescent strains ...... 124 Microscopy ...... 125 Analysis of cell shape ...... 125 Analysis of MreB distributions in mixed populations ...... 126 Analysis of Venus-MreB timelapses ...... 127 Clustering strains by shape ...... 128 Highlighting residues on the crystal structure ...... 128 Using Chernoff faces to display cell shape phenotypes ...... 129 Supplemental Figures………………………………………………………………….. 130

Chapter 4: In vitro assembly of Caulobacter MreB ...... 137

Abstract ...... 138 Introduction ...... 139 Results ...... 143 Attempt to purify native MreB from Caulobacter ...... 143 Recombinant expression and native purification of Caulobacter MreB ...... 145 Denaturing preparation of Caulobacter MreB ...... 148 High speed centrifugation of purified MreB ...... 148 Electron microscopy of MreB polymers: glow discharged grids ...... 151 Electron microscopy of MreB polymers: Poly-L-Lysine-coated grids ...... 156 Dynamic light scattering of MreB polymers ...... 161 Determination of critical concentration for polymerization ...... 164 Purification of GFP-MreB ...... 166 Chemical labeling of MreB to a fluorophore ...... 169 Discussion ...... 173 The assembly state of purified Caulobacter MreB ...... 173 The influence of salt on MreB polymers in vitro ...... 175 The influence of temperature on MreB polymers in vitro ...... 176 The effect of nucleotide on MreB polymers in vitro ...... 177 Materials and Methods ...... 181

xii

Reagents ...... 181 Buffers (in alphabetical order) ...... 181 Cycling of endogenous MreB in Caulobacter extract ...... 182 Cloning ...... 183 Growth of E.coli and induction of His-MreB ...... 183 Native purification of recombinant MreB ...... 184 Denaturing purification of recombinant MreB ...... 185 SDS-PAGE ...... 187 Measurement of protein concentration and concentrating purified MreB ...... 187 Thrombin digestion ...... 187 Purification of His-MreB from Caulobacter ...... 188 Native purification of GFP-MreB ...... 189 Centrifugation ...... 189 Electron Microscopy ...... 190 Dynamic light scattering ...... 191 Right angle light scattering ...... 191 Chemical coupling of MreB to Dylight ...... 192 Preparation of DNA-coated beads ...... 192 Imaging of Dylight-MreB ...... 193 Light microscopy ...... 193

Chapter 5:Fractionation of MreB disassembly activity ...... 195

Abstract ...... 196 Introduction ...... 197 Results ...... 199 Purified MreB responds to Caulobacter extract ...... 199 The identification of prominent proteins in an active MDF fraction (Scheme 1) ... 200 A modified fractionation scheme separates Transaldolase and MDF activities (Scheme 2) ...... 203 Transaldolase and MDF activity ...... 206 Nucleic acid affects the fractionation of, but is not required for, MDF activity ..... 207

xiii

Discussion ...... 211 What is the inhibitor? ...... 211 Suggestions for future work ...... 212 Is this activity relevant for MreB function in vivo? ...... 214 Author Contributions ...... 214 Acknowledgements ...... 214 Materials and Methods ...... 215 Buffers (in alphabetical order) ...... 215 Purification of MreB ...... 216 Preparation of Caulobacter extract ...... 216 Protease treatment ...... 216 Nuclease treatment and Ethidium Bromide spot testing ...... 217 Ammonium sulfate precipitation ...... 217 Determination of protein concentration ...... 218 DOC/TCA precipitation ...... 218 SDS-PAGE ...... 218 MDF activity assay ...... 218 Scheme I protocol ...... 220 Scheme II protocol ...... 221 Scheme III protocol ...... 221 Scheme IV protocol ...... 222 Transaldolase assay ...... 222 Transaldolase knockout ...... 223 Mass Spectrometry ...... 224

Chapter 6: Is MreB cytoskeletal? ...... 225

Abstract ...... 226 Why do we think that MreB is cytoskeletal? ...... 227 Mechanisms for slow, persistent motion ...... 230 A cytoskeletal structure ...... 232 The peptidoglycan cell wall ...... 233

xiv

The nucleoid ...... 235 Concluding remarks ...... 237 Acknowledgements ...... 238

Chapter 7: Concluding Remarks ...... 241

References ...... 247

xv

LIST OF TABLES Chapter 1 Table 1: Functions and mechanisms of selected bacterial cytoskeletal proteins………………………………………………………………………. 21

Chapter 3 Supplementary Table 1: Isolated mreB mutations…………………………… 135 Supplementary Table 2: Number of cells in shape dataset and OD at measurement……………………………………………………………… 136

Chapter 5 Table 1: Mass spectrometry data from active Caulobacter MDF fraction…………...... 224

Chapter 6 Table 1: Characteristics of the different types of structures that could potentially drive the slow, directed motion of MreB……………………… 239

xvi

LIST OF FIGURES Chapter 1 Figure 1: Cytoskeletal movement of DNA in bacteria……………………………. 25 Figure 2: Hypothetical scaffolds in bacteria………………………………………. 31

Chapter 2 Figure 1: MreC forms a helix that is nonoverlapping with MreB structures…… 40 Figure 2: mreC is essential and independent of MreB………………………….. 42 Figure 3: Both MreC and MreB are required for helical GFP-Pbp2 localization… 44 Figure 4: FtsZ-depletion prevents GFP-Pbp2 accumulation at division plane and A22-induced cell shape defect………………………………………………… 45 Figure 5: Model for maintenance of proper cell morphology in Caulobacter……. 50 Supplemental Figure 1: YFP-MreB and mCherry-MreB colocalize……………… 60 Supplemental Figure 2: Established MreC-mRFP1 spirals are stable in A22- treated cells……………………………………………………………………….. 60 Supplemental Figure 3: Established helical pattern of GFP-Pbp2 is stable in A22- treated cells………………………………………………………………………. 61 Supplemental Figure 4: Cell shape can be quantitatively described by using principal components analysis…………………………………………………. 62 Supplemental Figure 5: Secondary and tertiary PCA shape modes demonstrate that A22 affects cell curvature…………………………………………………. 63 Supplemental Figure 6: 2D shape map describes cell shape according to the first two principal modes of variation………………………………………………. 64

Chapter 3 Figure 1: Morphology varies in A22-resistant Caulobacter……………………… 70 Figure 2: A22 can alter the morphology of A22-resistant Caulobacter………….. 71 Figure 3: A22-resistant Caulobacter have similar doubling times……………….. 72 Figure 4: PCA Shape analysis of A22-resistant Caulobacter…………………….. 74 Figure 5: Clustering A22-resistant Caulobacter strains by cell shape……………. 78 Figure 6: Clustering A22-resistant Caulobacter strains by cell shape and the response to A22…………………………………………………………………… 79 Figure 7: The dramatic variable width phenotype exhibited by some A22-

xvii

resistant strains……………………………………………………………….. 81 Figure 8: The subcellular localization of Venus-MreB mutants in the absence of A22 ………………………………………………………………………….. 84-85 Figure 9: Illustration of PCA Fluor Modes………………………………………. 87 Figure 10: Timelapse imaging and analysis of MreB distribution in synchronized wild type and G165A Venus-MreB expressing strains ……………………… 91 Figure 11: Timelapse imaging and analysis of MreB distribution in synchronized V324A and D189G Venus-MreB expressing strains. ……………………….. 92 Figure 12: Timelapse imaging and analysis of MreB distribution in synchronized A325P Venus-MreB………………………………………………………….. 94 Figure 13: Average timelapse data for each strain. ………………………………. 96 Figure 14: Correlating MreB localization with cell shape in the absence of A22... 98 Figure 15: GFP-Crescentin localization in selected MreB mutant strains………... 100 Figure 16: mCherry-Pbp2 localization in the polar MreB mutants………………. 102 Figure 17: The subcellular localization of Venus-MreB mutants in A22………… 104-5 Figure 18: Average localization values for each strain grown in the presence and absence of A22……………………………………………………………….. 107 Figure 19: Correlating MreB localization with cell shape in the presence of A22.. 108 Figure 20: Residues conferring A22-resistance highlighted on the crystal structure of T.maritima MreB1……………………………………………….. 111 Figure 21: A model for the cell-cycle regulation of MreB localization………….. 115 Supplemental Figure 1: Pairwise correlation coefficients between measurements of cell shape…………………………………………………………………... 130 Supplemental Figure 2: Mean PCA Shape Mode values for each strain…………. 131 Supplemental Figure 3: Using PCA Fluor Modes 1 and 2 to describe the distribution of Venus MreB in selected example cells……………………………. 132 Supplemental Figure 4: Mutant residues in conserved motifs of the actin fold…... 133 Supplemental Figure 5: The phenotypes of A22-resistant Caulobacter represented in abstract faces. ………………………………………………… 134

Chapter 4 Figure 1: Cycling of MreB in Caulobacter extract……………………………….. 144 Figure 2: Purified MreB forms a large complex………………………………….. 147 Figure 3: Electron microscopy of natively-purified MreB in salt on standard glow

xviii

discharged grids………………………………………………………………. 150 Figure 4: Electron microscopy of natively-purified MreB in salt on grids coated with Poly-L-lysine……………………………………………………………. 153 Figure 5: Electron microscopy of refolded MreB in salt on grids coated with Poly-L-Lysine………………………………………………………………… 154 Figure 6: Electron microscopy of refolded MreB in salt on grids coated with Poly-L-Lysine………………………………………………………………… 155 Figure 7: Non-filamentous particles of refolded and natively purified MreB in salt on PLL-coated grids……………………………………………………… 158 Figure 8: Electron microscopy of natively-purified MreB in the absence of salt on PLL-coated grids………………………………………………………….. 159 Figure 9: Electron microscopy of refolded MreB in the absence of salt on PLL- coated grids…………………………………………………………………… 160 Figure 10: Dynamic light scattering of MreB polymers………………………….. 163 Figure 11: Right angle light scattering of MreB polymers………………………... 165 Figure 12: Purified GFP-MreB forms polymers that appear as puncta in light microscopy and bind to DNA-coated beads………………………………….. 168 Figure 13: Dylight-labeled MreB binds PLL and condenses DNA………………. 172

Chapter 5 Figure 1: A biochemical assay for MreB disassembly……………………………. 201 Figure 2: Fractionation Scheme 1………………………………………………… 202 Figure 3: Fractionation Scheme 2………………………………………………… 205 Figure 4: Transaldolase is not required for MDF activity………………………… 207 Figure 5: Fractionation Scheme 3………………………………………………… 208 Figure 6: Fractionation Scheme 4…………………………………………………. 210

Chapter 6 Figure 1: Models for the persistent motion of single molecules of MreB……….. 231

xix

xx

Chapter 1 Introduction

Natalie Dye

Sections of this chapter were published in the Trends in Cell Biology, 2007 [1]

P a g e | 1

FROM MOLECULES TO CELLS

Even the simplest of cells are capable of growing, changing form, replicating, undergoing directed migration, and interacting with their environment. In all kingdoms of life, one can find examples of cells that maintain asymmetry and internal compartmentalization across numerous division cycles. This cellular organization is vitally important for the physiology and survival of many organisms. Additionally, the overall size and shape of the cell can be important and regulated phenotypes, highlighting that mechanisms for controlling growth exist on both a local and cellular scale, as well as on an organismal scale for multicellular life forms. The ability of the cell to spatially and temporally coordinate the numerous biochemical reactions that contribute to these activities is truly amazing. How does an organized, dynamic, three-dimensional cell arise from the essentially one-dimensional information that is encoded in DNA? Ultimately a series of molecular interactions must be capable of generating and maintaining a cell that is orders of magnitude larger and persists for orders of magnitude longer than each individual component. In eukaryotic cells, dynamic intracellular organization relies heavily on the self-organizing cytoskeleton of actin, and molecular motor proteins. These proteins are essential for such cell-level behaviors as motility, cell division, and vesicular transport. Above a certain critical concentration, actin and tubulin each spontaneously self-assemble into long filamentous homopolymers that can be several microns in length. They can each bind high energy nucleoside triphosphates (ATP or GTP, respectively), and the nucleotide state affects both the thermodynamics and kinetics of assembly, such that in A/GDP-polymers that are less stable than A/GTP-bound polymers. As a result, interesting dynamic behaviors arise, including treadmilling, during which subunits of a polymer at steady state preferentially add at one end and subtract from the other, and dynamic instability, during which filaments spontaneously switch between periods of assembly and disassembly, even if the monomer concentration remains constant. Other proteins can also use this nucleotide cycle to either stabilize or destabilize these polymers spatially and temporally. Because microtubules and actin filaments are inherently polar (with different rates of assembly and disassembly at each end) and interacting proteins can differentially bind

2 | P a g e

CHAPTER 1 to these different ends, these filaments are capable of communicating long range positional information to other components of the cell. Cytoskeletal components are also capable of performing work by actively translocating other molecules and generate force both locally at the level of a single filament and globally at the level of an entire crosslinked meshwork or bundle of filaments. By controlling the presence and localization of proteins that regulate cytoskeletal polymerization and interactions, the cell can generate different types of cytoskeletal patterns that have various consequences for intracellular organization and cellular behavior. In the last two decades it has become recognized that bacterial cells are also capable of a remarkable level of intracellular organization. In general there are no membrane-enclosed compartments in bacterial cells; nonetheless, proteins, lipids, plasmids, and genomic loci have been observed to occupy specific subcellular locations dynamically throughout the cell cycle [2-6]. The mechanisms by which bacterial cells achieve this intracellular organization remain largely unclear. The cytoskeleton has long been considered to be a eukaryotic invention. Actin and tubulin are extremely well conserved across , but no homologs could be identified in prokaryotic organisms by primary sequence searches. Recently, however, proteins with similar structural folds to those of actin and tubulin have been identified in bacteria [7-10]. With this discovery, we have to wonder whether even bacterial cells also rely heavily on dynamic polymeric proteins related to those in eukaryotic cells for cell-level organization. The differences and similarities between the mechanisms used by cells that are so evolutionarily distant from one another could be valuable for understanding the origins and evolution of cellular and multicellular life.

P a g e | 3

THESIS OVERVIEW

This thesis describes the work that I have done to contribute to our understanding of the role of the actin homolog MreB in the determination of cell shape in the Gram negative bacteria Caulobacter crescentus. In this introductory chapter, I will briefly introduce what was known prior to this work about the mechanisms for establishing intracellular organization and cell shape in bacteria. I will also introduce the bacterial cytoskeleton and what is known about the mechanisms that these cytoskeletal proteins use to contribute to cellular processes. In the next chapter, I will describe how MreB collaborates with other proteins encoded in its operon to maintain a proper rod-like cell shape. In Chapter 3, I will describe how I and colleagues used A22, a drug that specifically targets MreB, to probe the dynamics of MreB and select for mutants in mreB. In Chapter 4, I will describe preliminary characterization of the in vitro biochemistry of Caulobacter MreB polymerization. In Chapter 5, I will describe our attempt to use a biochemical assay for MreB disassembly in vitro to identify regulatory factors from Caulobacter cell extract. In Chapter 6, I will describe alternative ways to explain the existing data on MreB that do not invoke the polymerization of MreB into filamentous homopolymers in cells. Chapter 7 is a concluding chapter that synthesizes the work performed in this thesis and provides suggestions for future work in this field.

4 | P a g e

CHAPTER 1

BUILDING A WALL

The shape of the cell is one read-out of its dynamic intracellular organization. In unwalled cells, those lacking a rigid macromolecular structure such as peptidoglycan (bacteria), cellulose (plants) or chitin (fungi), the dynamic protein-based cytoskeleton network provides mechanical support and rigidity to the cell and therefore plays a critical role in the determination of cell shape. In walled cells, the enzymes that synthesize the force-bearing structure must coordinate their activities in space and time. In eukaryotes, the mechanism for this coordination is still unclear but is known to involve the actin- and -based cytoskeleton. As noted above, in bacteria, the mere presence of protein cytoskeletal components has only recently been suggested, and how these components contribute to the processes regulating cell shape in walled bacteria remains completely unknown. The study of bacterial cell shape has a rich history, however. Because this thesis will focus heavily on bacterial cell shape, I must review some of this history. In this section, I will briefly introduce what was known previously about the nature of the bacterial cell wall and the molecular determinants of cell shape, focusing on the particular proteins, ideas, and experiments that are relevant to this thesis. Much of the relevant research on this topic (particularly in Gram-negative organisms) has focused on the model organism because of its genetic tractability, fast growth, and fairly common cell shape. As such, I will focus on E.coli here, even though my thesis work was in Caulobacter. Both E.coli and Caulobacter cells have been described as rods, approximated as cylinders with hemispherical caps. Rod-like cell shapes are common in the bacterial kingdom, although many other interesting shapes have been described [11]. Because this section is intended as an introduction and not a comprehensive review, I have greatly over-simplified decades of research and indeed entire careers. More detailed information can be found in Refs [12-17]. Unless otherwise cited, the references for the data I describe below can be found within those articles and books.

P a g e | 5

The peptidoglycan cell wall Most bacteria have a layer of peptidoglycan that lies just outside of the cytoplasmic membrane. This material provides structural support to the cell and holds the shape of the cell. In fact, the peptidoglycan layer can retain the shape of the cell even after it is separated away from the rest of the molecular components [13, 15]. Thus, this layer is often referred to as the cell wall. A cell wall that has been isolated from the rest of the cell is called the ―sacculus.‖ The term ―peptidoglycan‖ is a general name for a macromolecule that comprises both glycans and peptides, and the specific form of peptidoglycan in bacterial cell walls has also traditionally been referred to as ―murein.‖ Murein (or the peptidoglycan in bacterial cell walls) consists of glycan polymers that are crosslinked by short peptides (3-5 amino acids). The glycan polymers are typically made up of alternating units of N-acetylglucosamine (NAG) and N-acetylmuramic acid (NAM). The short peptides are covalently linked to the NAM units. The amino acid composition of these peptides can vary considerably by number and type, depending on the bacterial species. Often atypical amino acids are present in the peptidoglycan, include D enantiomer forms. In cells, the peptides of one glycan chain are covalently linked to those of another neighboring glycan chain at a dibasic amino acid of one of the peptide side chains (meso-diaminopimelic acid in E.coli), such that a continuous layer of peptidoglycan is made that envelopes the entire cell. The composition of the peptide side chains, as well as the extent and type of crosslinking can be measured using HPLC to separate products of muramidase-treated isolated sacculi. The precise organization of the glycan chains in the peptidoglycan of live cells is still unclear. Recent cryoelectron tomography experiments have demonstrated that the thickness of the peptidoglycan is 5-8nm in plunge-frozen cells of the Gram-negative organisms E.coli and Caulobacter [18]. Given known distributions of glycan lengths in cells, this thickness indicates that the glycan chains probably lie parallel to the cytoplasmic membrane and that there is only a single layer of these crosslinked glycan polymers. Other models have been suggested, however, particularly for Gram-positive organisms with thicker cell walls [19, 20]. Within that single layer of Gram-negative peptidoglycan, the glycan strands are thought to lie roughly perpendicular to the long axis of the cell, placing the peptide bonds of the crosslinked side chains in parallel to the long

6 | P a g e

CHAPTER 1 axis, although again, there is not very much direct evidence to support this idea [18, 21- 23].

Brick-makers and Brick-layers The synthesis of the cell wall requires numerous enzymes that lie in three compartments of the cell (, membrane and periplasm). Both genetic and biochemical methods (mostly in E.coli) have been used to identify the specific molecular components for each step. The synthesis of the basic repeating unit of peptidoglycan, a disaccharide of NAM and NAG, occurs in the cytoplasm. The peptide side chain is added stepwise to the NAM, and then this unit is covalently coupled to a lipid carrier (Lipid I). The last cytoplasmic step is the addition of the NAG molecule and is catalyzed by the enzyme MurG. The resulting molecule, Lipid II, is transported across the cytoplasmic membrane to the periplasmic side, where the sugar-peptide moiety is incorporated into the existing peptidoglycan with another set of enzymes and the lipid carrier molecule is recycled. In the context of cell shape, perhaps the most relevant reactions are the ones that occur in the periplasm, as they are involved directly in the growth and destruction of this shape-determining structure: polymerization of glycan strands (transglycosylation), crosslinking of glycans (transpeptidation), removal of old strands and crosslinks, and maturation of the murein. If we consider the growth of the peptidoglycan to be analogous to the building of a brick house, the disaccharide units would be the bricks, the cytoplasmic enzymes would be the ―brick-makers,‖ and the periplasmic enzymes would be the ―brick-layers‖. While the house cannot be built without bricks, it is the precise arrangement and number of bricks that determines the shape of the house. Of course this is not a perfect analogy, as the peptidoglycan is not a fixed structure but a dynamic one that is constantly growing and turning over. Many of the enzymes that are involved in peptidoglycan synthesis in the periplasm were identified by their ability to bind penicillin and are thus named ―Penicillin binding proteins‖ or PBPs. Because they were identified biochemically, PBPs were named according to molecular weight. This naming convention has become quite confusing as homologous proteins in two different species may be named differently if

P a g e | 7

the molecular weight ordering changes. I will refer to the E.coli names throughout unless otherwise specified. There are two types of enzymes capable of transglycosylation: those that are capable of only transglycosylation and those that can catalyze both transglycosylation and transpeptidation, the crosslinking of peptides of neighboring glycan strands. Most transglycosylation occurs through these bifunctional enzymes, which are also referred to as high molecular weight penicillin binding proteins. In E.coli, there are three such bifunctional enzymes, Pbp1A, Pbp1B, and Pbp1C, which are partially redundant. While these enzymes are capable of both reactions, their transpeptidase activity is not sufficient for complete viability and proper shape, as there are (in E.coli) two monofunctional transpeptidases (Pbp2 and Pbp3) that are also required for proper growth and form. There also exist many enzymes that remove existing peptides and glycans from this structure, including lytic transglycosylases, endopeptidases, amidases, and carboxypeptidases, although these will be largely ignored in this thesis.

Growth of the sacculus and maintenance of form Bacterial cells are able to maintain a remarkably constant shape as they double in size and divide [24], indicating that the activities of all these synthetic enzymes must be coordinated on a cellular scale. Furthermore, because the cell wall resists a great amount of turgor pressure, the synthesis and destruction of peptidoglycan must be carefully spatially and temporally regulated. For these reasons, it has been proposed that many of these enzymes bind together in a large complex [13, 25]. Analogous to other large enzymatic complexes such as the replisome, a ―holoenzyme‖ for peptidoglycan synthesis would contain all of the enzymes required for the incorporation of a new strand: transglycosylases to make the new strand, transpeptidases to attach it to the existing structure, and autolytic enzymes to break the existing bonds and remove old strands. In this way, the enzymatic functions could be regulated and coordinated. Indeed, there is biochemical data to suggest that many of the periplasmic ―brick-layers‖ can bind to one another [25-27]. Nevertheless, there is still no direct evidence to suggest that these enzymes are always in a complex with one another in the cell or that their activities are truly coupled. To explain how a structure that is constantly under tension can be

8 | P a g e

CHAPTER 1 replicated, a ―make before break‖ hypothesis was proposed: first, new peptidoglycan segments are made just underneath the old layer and attached to the structure; then the old segments are degraded and the new segment expands to fill the space [13, 14]. Thus only when new pieces are made and incorporated into the existing structure are the old pieces removed; growth and cleavage are coupled in space and time. Not only do the ―brick-layers‖ have to synthesize a structure that is constantly resisting turgor pressure, but they have to act in such a way to maintain a constant shape. In short, while bacterial growth and division has been studied for decades, it is still mysterious how the peptidoglycan is so faithfully formed into its shape. Some have argued that peptidoglycan synthesis is templated, such that the activity of the hypothetical holoenzyme is guided by the existing glycan strands [13, 15, 25]. If the glycans are copied exactly, preserving the existing length distribution and number of crosslinks, the shape is maintained. The idea that the peptidoglycan and cell shape is structurally inherited seems reasonable. Indeed, rod-shaped E.coli cells that have been spheroplasted to remove the cell wall will immediately regrow a spherical sacculus [28]. Thus, without pre-existing information on the structure of the peptidoglycan, the cells do not immediately generate a proper cell shape de novo. However, these cells do eventually adapt a rod-like shape, implying that the cells do actually have dedicated mechanisms for morphogenic control [28]. A conceptually very different model, the Surface Stress Theory, was proposed by Arthur Koch in the early 1980s [14, 29, 30]. This theory argues that cell shape emerges from the biophysical properties of the peptidoglycan and the physiochemical properties of the synthetic enzymes when they are in a structure that is resisting constant turgor pressure. Central to this theory is the idea that hydrostatic pressure forces cell expansion, increasing the stress on the bonds of the peptidoglycan and lowering the activation energy for hydrolysis and transpeptidation. Cell shape and dimensions are determined by the ability of the cell to restrict regions of growth or change the biochemical properties of the murein. If, for example, the new pole of the cell is made inert after division, a rod cell emerges strictly from the biophysical properties. The concepts underlying this theory were greatly influenced by the physics of soap bubble expansion. While these models provide an interesting and important foundation, neither is

P a g e | 9

able to completely explain the phenotypes of certain morphogenic mutants, reversible transitions in shape due to environmental conditions, or the shapes of more complicated bacterial organisms, such as budding and prosthecate bacteria [31-33]. Furthermore, current models for the maintenance of cell shape must now incorporate the recent discovery of the putative bacterial cytoskeletal elements, which will be the focus of this thesis.

Alternating elongation and division In order to understand how the cell controls its shape, it is important to know the timing and subcellular localization of growth and destruction of the peptidoglycan cell wall. Early work using radioactivity pulse-chase, conditional morphological mutants, and pharmacological inhibition of specific synthetic enzymes indicated that there are two distinct modes of growth (elongation and division) catalyzed by distinct molecular machineries that at least partially compete with one another [34, 35]. Some enzymes can participate in both processes, but others are specific to one or the other. In particular, the transpeptidases Pbp2 and Pbp3 are specific to elongation or division, respectively [35, 36]. When Pbp2 is inhibited pharmacologically or genetically perturbed, the cells become round, whereas when Pbp3 is perturbed, the cells become filamentous. Likewise, MreB and FtsZ are thought to be fairly specific to elongation and division, respectively [35, 37- 41]. The balance between these two processes is important for maintaining a proper cell shape in E.coli [34]. In the early 1990s, Miguel de Pedro and colleagues developed a method for differentially labeling old and new peptidoglycan in isolated E.coli sacculi that relies on the ability of cells to incorporate exogenous D-Cysteine specifically into the cell wall [42]. Using pulse-chase experiments, they were able to show that new peptidoglycan incorporates into the existing cell wall along the lateral sides of the cylindrical shape and at midcell as the cells divide. The poles of the cell remain relatively inert to new peptidoglycan insertion. Later, Errington and colleagues developed a different technique for labeling new peptidoglycan in Gram-positive organisms that can be used to follow growth in live cells [43]. Their technique is use a fluorescent-derivative of the antibiotic Vancomycin (VancFL), which binds the terminal D-Ala-D-Ala of nascent pentapeptide-

10 | P a g e

CHAPTER 1 glycan chains. In B.subtilis, the terminal D-Ala is removed shortly after the new glycan chain is incorporated into the existing meshwork, so VancFL should only label new peptidoglycan. A similar technique using a different antibiotic, Ramoplanin has also been used [44]. Unfortunately, this technique is problematic in Gram-negative cells, which have an outer membrane that is impermeable to these large glycolipid molecules, or in species that have a high percentage of pentapeptide-containing glycan strands (for example, Caulobacter [45]). Nonetheless, in B.subtilis, they found that new peptidoglycan incorporates along the lateral sides and at midcell and not at the poles, just as de Pedro et al found in E.coli. Both methods have now been used to show that the growth along the sides requires MreB homologs, and the growth at midcell requires FtsZ [43]. Interestingly, there are (of course) exceptions: a rod-shaped species that does not contain an mreB homolog was found to synthesize new peptidoglycan at the poles rather than the sides [43]. The mechanisms for how growth is restricted to specific subcellular locations and how MreB and FtsZ contribute to these processes are not known. Recently fluorescent fusions have also been made to many of the peptidoglycan synthetic enzymes in E.coli and B.subtilis as a proxy for determining the sites of new cell wall growth in live cells [46, 47]. Many of the synthetic enzymes are present in fairly low copy in the cell, however [48], so it is very difficult to determine accurate localization patterns. In addition, these experiments are complicated by the fact that the sites of localization and function may not be completely identical.

P a g e | 11

ESTABLISHING INTRACELLULAR ORGANIZATION IN BACTERIA

In order to restrict growth of the outer layers of the bacterial cell to particular subcellular regions (cylindrical sides and midcell but not the poles), the cell must have mechanisms for controlling either the subcellular localization or activity of the synthetic enzymes. The mechanism for this particular process is unknown, but progress has been made in understanding how specific molecular components of other processes in the cytoplasm become localized. In particular, we now know of dedicated mechanisms for distinguishing between the poles, the midcell division plane, the nucleoid, and a growing forespore. Because in general bacterial cells are small, and diffusion of small molecular components is fast on this scale, active transport mediated by cytoskeletal proteins is not always required for proper redistribution of cellular components. In fact diffusion is central to many of the mechanisms for intracellular localization that have been described. Here, I will review some of the mechanisms that bacterial cells use to distinguish particular regions of the internal cytoplasm. More information can be obtained from these excellent recent reviews [5, 6, 49, 50].

Finding the middle: Regulating the localization of FtsZ Most bacterial species divide by binary fission. Many of the components of the division machinery in the bacterial model organisms have been identified genetically [51, 52]. The specific function of each one of these proteins has not been fully elucidated, but it is known that these proteins arrive at the middle of the cell at different times [51, 52]. The first protein to arrive and the only protein that is required for the recruitment of all other downstream components is FtsZ, a tubulin homolog (See ―Cytoskeletal Proteins in Bacteria‖ Section). FtsZ forms linear polymers that are linked to the cytosolic surface of the inner membrane by other proteins and required to recruit the downstream components of the division machinery [51, 52]. To produce two daughter cells of similar sizes, a dividing cell must be able to assemble FtsZ and the rest of the division machinery at or near the middle of the cell, and to ensure that the daughter cells each receive a full copy of the chromosome, the division machinery must be assembled in the space in between two segregated chromosomes and not inappropriately over the chromosomes. The

12 | P a g e

CHAPTER 1 question of how FtsZ is specifically placed at the cell middle has been studied extensively in E.coli and B.subtilis and is now reasonably well understood in these organisms [2, 49, 52, 53]. Many of the required for accurate division site placement were also identified genetically. In E.coli, a group of genes, minC, minD, and minE, were found to have a ―minicell‖ phenotype: instead of dividing in the middle, these mutant cells would occasionally divide near one of the poles to produce one large cell and one ―mini‖ cell, which is much smaller than normal and usually without a full copy of the chromosome [54]. Work in numerous labs over many years has led to the following model for the function of these proteins in E.coli (reviewed in [49, 52, 55, 56]). MinC is a direct inhibitor of FtsZ polymerization. MinD is a Walker-type ATPase that self-associates in an ATP-dependent manner (See Cytoskeletal Proteins in Bacteria section). MinD binds to the membrane through an amphipathic helix in an ATP-dependent manner and recruits MinC to the membrane. MinE disassociates the MinCD complex from the membrane by activating the ATPase activity of MinD. MinE also cooperatively self-assembles into a ring-like structure. Together, this system of three proteins when confined to the cylindrical geometry of the E.coli cell will set up a pattern of localization that oscillates from pole to pole with period of approximately a minute [57-60]. The result of this oscillation is that the time-averaged concentration of MinC, the inhibitor of FtsZ polymerization, is highest at the poles and lowest at midcell. Thus, the Min system allows FtsZ to dynamically sense how far it is from the poles, providing the cell with a kind of distance sensor that relies solely on the rates of diffusion, ATP hydrolysis, and cooperative binding of the proteins to one another and the membrane. The oscillatory pattern of this system has been modeled computationally by several groups using theory for the emergence of dynamic chemical oscillations first developed by Alan Turing’s in 1952 (the list is long and still growing; here are selected examples: [61-66]). Oscillatory patterns between MinD and MinE have also been reconstituted in vitro with purified components on a supported lipid bilayer [67]. Controlling the localization of an FtsZ inhibitor seems to be a general way in which bacteria find midcell, although different organisms have evolved slightly different mechanisms for achieving this spatial control. In B.subtilis, there is no MinE component.

P a g e | 13

MinD binds specifically to DivVIA, which localizes to the poles [68-71]. Because there is no MinE, this system does not oscillate. In Caulobacter, there are no close homologs of any of the Min proteins, but the Walker-type ATPase, MipZ, plays an analogous role [53]. Like MinC, MipZ is a direct inhibitor of FtsZ polymerization. In cells, MipZ binds to ParB, which binds near the chromosomal origin of replication, at one pole of the cell. At this time, early in the cell cycle, FtsZ is localized at the opposite pole. When the chromosomal origins segregate, MipZ follows so that a complex of DNA, ParB and MipZ is present at both poles. Thus, the concentration of the inhibitor MipZ is highest at the poles and lowest in the middle. In this organism, division site placement is directly coupled to chromosome segregation. In E.coli and B.subtilis, other mechanisms exist for preventing FtsZ assembly over the nucleoid (involving the proteins SlmA [72] or Noc [73], respectively).

Finding the Pole There are now numerous examples of proteins that specifically localize to one or both poles of a bacterial cell. One example, DivVIA, was mentioned in the last section. Additional examples include the chemoreceptor array in E.coli and Caulobacter, polarly localized histidine kinases in Caulobacter, and virulence factors in Shigella flexneri and Listeria monocytogenes [2, 5, 50]. Proper localization is usually critical for the function of these polar proteins and mislocalization has important consequences for cell behavior. For example, ActA is a virulence factor that is present at the surface of Listeria cells specifically at one pole [74, 75]. Listeria uses ActA to promote actin-based motility upon invasion of human epithelial cells [76, 77]. The polar localization of ActA is critical for proper motility of the cells and therefore the cell’s ability to spread from one host cell to another [74, 75, 78, 79]. In fact, proteins are not the only molecular components that are able to specifically localize to the cell pole. It is now recognized that the bacterial chromosome is highly organized, such that specific chromosomal loci occupy particular subcellular regions of cell [80-82]. In E.coli and B.subtilis, it has been shown that the phospholipid cardiolipin is concentrated at the cell poles [3, 83-85]. Our understanding of the mechanisms by which these components find their way to the pole and are retained at the poles is still incomplete and research in this area is

14 | P a g e

CHAPTER 1 currently very active. In theory, a reaction-diffusion based mechanism, like the one regulating FtsZ localization, could underlie asymmetric polar development as well, although such a mechanism has not yet been identified. From the data that are currently available it is clear that there is more than one mechanism by which molecular components localize to the cell pole, varying by organism and the specific component. In particular, it is still unclear how the most upstream component of a pathway becomes localized. This component must be able to independently sense the difference between the poles and the rest of the cell; sometimes it must even be able to tell the difference between the two poles. The poles of the bacterial cell are of course geometrically distinct from the rest of the cell, possessing higher negative curvature than the lateral parts of the cell. It is possible that there are mechanisms that allow molecular components to detect that geometry, allowing the cell to generate polarity de novo. Indeed, very recent work in E.coli and B.subtilis suggests that there are molecules capable of using geometric cues to find distinct intracellular regions of the cell [86-88]. As mentioned above, the peptidoglycan at the poles of the bacteria is also inert to new synthesis. Peptidoglycan synthesis at division involves a set of enzymes that are specific for the division process and different from those that are required for synthesis along the sidewall. After division, this new peptidoglycan undergoes a maturation process that renders the new pole inert for new cell wall synthesis. Because this maturation occurs relatively slowly with time, there should be a detectable difference between older and newer cell poles, although no such chemical differences have yet been clearly identified. If a polarity cue is able to bind to peptidoglycan and discriminate between the peptidoglycan at the poles and that along the sidewall, it would be able to recognize the poles and preferentially localize itself and downstream components to these sites. Similarly, if a protein is able to detect the molecular differences between old and new peptidoglycan, it could be asymmetrically localized. No molecules have yet been identified with this activity, however. Lastly, these chemical differences at the poles highlight the fact that bacterial cells are not really born symmetric; therefore a polar protein’s intracellular localization usually only needs to be propagated through division, rather than arise de novo at every cell cycle. For example, it is possible that a protein is placed as a localization determinant

P a g e | 15

during division at the site that will become the new pole. If that protein is fixed at that site in some way, the daughter cell would then be born asymmetric and already capable of distinguishing its poles. In this way, the inheritance of structurally asymmetric cells could be an epigenetic determinant of intracellular localization of molecular components in a bacterial cell. Once the primary determinants of a subcellular region are localized, downstream components can be localized to these sites through a diffusion and capture mechanism, whereby the molecular component diffuses freely throughout the cell until it is bound and restricted to a particular site by binding to a partner protein that is already localized. Perhaps the first demonstration of such a mechanism in bacterial cells was the protein SpoIVFB in Bacillus subtilis [89]. Alternatively, the localization of a protein can be determined at the level of synthesis. If a protein is made in one particular location, its concentration will remain high at that location as long as its diffusion away from that site is limited or the protein is actively removed from other locations in the cell. The polar localization of IcsA, a virulence protein of the bacterial species Shigella flexneri, is thought to be maintained in this way [90, 91].

16 | P a g e

CHAPTER 1

CYTOSKELETAL PROTEINS IN BACTERIA

In the last decade, numerous proteins have been discovered that seem to have cytoskeletal functions and to play an important role in organizing the bacterial cell. Here, I will briefly introduce the types of cytoskeletal elements that have been identified in bacteria. More information can be found in reference [92].

The FtsZ tubulin homolog The gene ftsZ (filamentous temperature sensitive) was identified genetically as having an essential role in cell division [37, 93, 94]. The crystal structure of FtsZ is remarkably similar to that of tubulin, even though the two genes share very little identity at the level of primary sequence [7, 95]. It was also shown that FtsZ, like tubulin, can self-associate into filamentous polymers in a GTP-dependent manner, but the polymers of FtsZ are rather unlike the arrangement of microtubules [96-99]. The critical concentration of FtsZ in GTP is very low, and rate of polymerization is rapid, indicating that perhaps there is no need for nucleation factors [98]. In the presence of GTP and low Magnesium, FtsZ forms straight or gently curved filaments that are only about 4nm wide, corresponding to a single subunit, whereas GDP-bound FtsZ filaments are highly curved, suggesting that GTP hydrolysis drives the transition between straight and curved filaments [100-102]. This nucleotide response resembles that of microtubules, but FtsZ polymers do not undergo dynamic instability. Many have speculated that the transition from straight to curved filaments could be used to generate force for constriction of the cell at division, though other models involving filament sliding have also been proposed (reviewed in [103]). In vivo, FtsZ localizes underneath the membrane in a tight zone near midcell that appears as a band in epifluorescence microscopy or a ring in three-dimensions [104]. Recently, filaments of FtsZ have been visualized by cryo-electron tomography (cryoET) in whole, unlabeled, plunge-frozen cells, indicating that FtsZ can polymerize both in vivo and in vitro [105]. Coupled with the structural resemblance to tubulin, these data indicate that FtsZ is a bacterial cytoskeletal protein. The ftsZ gene is highly conserved in bacteria, and it is even present in the genomes of mitochondria and chloroplasts.

P a g e | 17

Bacterial actin homologs The first indication that bacteria could have proteins that were evolutionarily related to actin was in 1992 [106]. Hexokinase, Hsp70, and actin had been crystallized and their structures seemed to be remarkably similar given the lack of significant primary sequence similarity. These proteins have a structure comprised of four subdomains surrounding a nucleotide binding cleft. In 1992, Bork et al identified motifs of high sequence conservation in these three proteins and then identified other proteins containing these motifs, including the prokaryotic proteins MreB, ParM (then called StbA), and FtsA. The proteins were predicted to have, ―subdomains with the same tertiary structure as the ATPase subdomains Ia and IIa of hexokinase, actin and Hsp70, a very similar ATP binding pocket, and the capacity for interdomain hinge motion accompanying functional state changes” [106]. Since then, the crystal structures of MreB [8], ParM [10], and FtsA [9] have been solved, and indeed it was confirmed that these prokaryotic proteins have actin-like folds. ParM is a plasmid-encoded protein that is required for the segregation of plasmid R1 in E.coli [107-110]. It has been shown to form filaments, both in vitro as a purified protein [10, 111] and in cells (see ―A push in the right direction‖ section, [109, 112]). While it forms filaments comprised of two protofilaments like actin, these filaments have the opposite handedness of those actin and very different biochemical properties [111, 113]. FtsA is a widely conserved chromosomally encoded protein that is involved in cell division, but its ability to assemble into filamentous polymers is controversial. One homolog of FtsA has been shown to form filamentous polymers when purified in vitro, but most other homologs are not thought to polymerize [114, 115]. It is known to associate with the tubulin homolog, FtsZ, and participate in division [51, 92]. Thus, FtsA is usually not thought of as a cytoskeletal protein itself, though it plays a role in cell division by interacting with FtsZ. MreB is also chromosomally encoded, and it is well conserved in bacteria with non-spherical shapes [43, 116]. It has been known for decades that MreB functions in cell shape, as genetic mutations in mreB cause normally rod-like E.coli to transition to round cells [38]. When it was crystallized in 2001, the T.maritima MreB homolog was shown to form filamentous polymers in vitro in a nucleotide-dependent manner [8]. These

18 | P a g e

CHAPTER 1 polymers tend to be large aggregates of several filaments bundled together, forming a structure that can be microns in length (much longer than the length of a bacterial cell). In that same year, Errington and colleagues were the first to demonstrate that GFP-labeled homologs of MreB in B.subtilis form helical patterns subjacent to the membrane [116]. Such a localization pattern had never before been observed in bacteria. Coupled with a well-established role for MreB in the maintenance of cell shape in rod-like bacteria, the existence of what appeared to be a continuous filamentous structure inside cells was suggestive of MreB being a true cytoskeletal filament. MreB has been implicated through genetic experiments in the processes of chromosome organization and segregation [40, 117-119], polar development [117, 120], and cell wall synthesis [43, 121, 122], but the mechanisms by which MreB contributes to these processes is largely unknown.

Crescentin, an intermediate filament-like protein Unlike MreB and FtsZ, the discovery of intermediate-filament-like proteins in bacteria was not made by comparing the crystal structures of the prokaryotic and eukaryotic forms. In fact no crystal structure of any full-length intermediate filament has yet been solved. However a purified bacterial protein has been shown to assemble into IF-like filaments [123]. In a visual screen for mutants in cell shape, Jacobs-Wagner and colleagues discovered that a loss of function in the gene creS causes normally curved Caulobacter cells to become completely straight, indicating that this gene is required for curvature in this organism[123]. This result alone does not indicate that Crescentin is a cytoskeletal protein, but further work on the biochemistry of this protein demonstrated that this protein could spontaneously self-assemble into long helical filaments. Fluorescently labeled Crescentin is localized to the inside curve of wild type Caulobacter cells, consistent with a role for inducing curvature, though individual filaments have not been yet visualized at high resolution with cryoET. The mechanism by which Crescentin induces curvature is not yet fully understood, though it is likely to involve asymmetric growth of the peptidoglycan on opposite sides of the cell [124].

P a g e | 19

The MinD/ParA family of bacterial-specific cytoskeletal proteins In bacteria, members of a superfamily of Walker ATPases have cytoskeletal properties and regulate the localization of both DNA and protein. Members of this family all share the deviant Walker A motif, GXGGXHK[TS], within the P-loop ATP binding pocket [125]. MinD and ParA are two members of this family that were characterized early and represent two groups that are distinct in sequence and function. MinD regulates the position of the cytokinetic FtsZ-ring by controlling the localization of a FtsZ inhibitor, MinC (described above and reviewed in [49, 52, 92]). The ParA subgroup functions in the segregation of DNA [110]. Three subclasses of the ParA group can be made based on protein sequence: chromosomal, and plasmid Type Ia and Type Ib [126]. To date, five members of the MinD/ParA superfamily have been shown to polymerize into filaments in vitro in a nucleotide-dependent manner: MinD (E.coli), Soj (B.subtilis), ParA (plasmid pB171), ParF (plasmid TP228), and SopA (F plasmid) [127- 132]. All but Soj were shown to form long thin protofilaments assembled into large, bundled structures with a frayed appearance at one end. Soj polymers do not resemble these bundled structures, and it is not yet known whether this difference is due to an inherent difference between the homologs or variation in the experimental conditions. In vivo, MinD/ParA homologs tend to form large scale filamentous structures that follow a helical path in the cell, often with oscillatory dynamics [59, 128, 130, 133-135]. E.coli MinD, for example, forms a helix at one cell pole that dissociates and reforms at the opposite pole approximately every 20s, a process dependent on a regulatory protein, MinE [58, 59]. Mutations that affect nucleotide hydrolysis do not prevent the formation of such these helical structures but do affect their dynamic behavior.

20 | P a g e

CHAPTER 1

Table 1: Functions and mechanisms of selected bacterial cytoskeletal proteins*

Protein name Eukaryotic Organisms Function Mechanism Refs** homolog AlfA Actin B.subtilis, on Segregate Unknown [136] plasmid pLS32 plasmid DNA FtsA Actin Many species Cell division Unknown Reviewed in [49, 52, 92] MamK Actin Magnetotactic Position Scaffold [137-139] species magnetosomes MreB Actin Nearly all rod- Maintain cell Possible [39, 40, 43, shaped species shape, segregate scaffold 116, 117, chromosome, 120, 140- regulate protein 143] localization ParM Actin E.coli on Segregate Insertional [109, 111, Plasmid R1 plasmid DNA polymerization 144, 145] ratchet FtsZ Tubulin Nearly all Define the plane Scaffold Reviewed species of division and in [49, 52, constrict the cell 92] CreS Intermediate C.crescentus Induce cell Unknown [123] filaments only (so far) curvature MinD None Many species, Prevent FtsZ Scaffold for a [59, 132], including polymerization FtsZ inhibitor, reviewed in E.coli, at the poles MinC [49, 52, 92] B.subtilis ParA, None Many species, Segregate Possible [110, 126, chromosome including chromosome depolymerizatio 146, 147] -encoded V.cholerea and n ratchet in C.crescentus Vibrio, others unknown ParA, None All types of Segregate Unknown [126, 130, plasmid- species, on plasmid DNA 135, 148] encoded plasmids

*The proteins listed above have been categorized according to their to eukaryotic cytoskeletal proteins. In white are actin homologs, in yellow, the tubulin homolog; in green, the intermediate filament homolog; in purple, the MinD/ParA superfamily, which does not have any counterpart in the eukaryotic cytoskeleton. **The references listed provide recent insights into the molecular mechanism of these proteins in contributing to cellular organization. More information on the bacterial cytoskeletal elements that are not discussed here can be found in a recent review [92].

P a g e | 21

THE PUSH AND THE PULL OF THE BACTERIAL CYTOSKELETON

Given that passive diffusion alone can allow cellular components to rapidly sample a small bacterial cell, it has been somewhat surprising to find that many of the bacterial cytoskeletal proteins are thought to be involved in actively positioning proteins and DNA (see Table 1). Here, I will review the general mechanisms that are emerging for how bacterial cytoskeletal proteins influence cellular organization.

A push in the right direction One basic way to influence the localization of cellular components is to physically move them through the cytoplasm to a specific location. In eukaryotic cells, there are motor proteins that directly transport vesicles, mRNA and proteins along tracks of actin or tubulin. But actin and tubulin themselves can also function as motors to propel objects through the cytoplasm [149]. In bacteria, the plasmid-encoded actin homolog, ParM, is thought to actively force two clusters of plasmid DNA apart, using its ability to polymerize (Fig 1A) [109, 111, 144, 145]. The plasmid R1 exists in low copy number in the Escherichia coli cytoplasm and would be quickly lost from the population without an active mechanism for partitioning. There are three components encoded by the plasmid that are dedicated to its partitioning: the actin-like ParM, the DNA-binding protein ParR, and a specific site on the DNA, parC [107]. These three form a complex which is necessary and sufficient for the segregation of the DNA in vivo [107-109]. The regulation of par locus expression and the integration of plasmid replication with the host cell cycle are reviewed elsewhere [150]. In the cell, plasmid molecules cluster together so that the DNA appears as tight foci when labeled inside cells [150, 151]. Early in the cell cycle, there is only one focus, which colocalizes with foci of ParM and ParR at midcell [109, 151]. As the cell cycle progresses, the plasmid is replicated by host machinery and split into two foci. These foci then move to opposite poles prior to division and appear connected by a line of ParM [109, 151]. The formation of these ParM filaments in vivo requires the ParR-parC complex [148]. A mutant ParM with decreased ATPase activity still forms filaments in vivo, but they appear more stable than wild type, do not support plasmid segregation, and

22 | P a g e

CHAPTER 1 often inhibit cell division, suggesting that the dynamic nature of these filaments is essential for their function [108, 148, 151]. In vitro, ParM forms filaments with strikingly similar ultrastructure to eukaryotic actin filaments [10, 148]. Filament formation in vitro can be stimulated by the DNA- binding protein ParR, together with parC [148]. This result was first interpreted as ParR accelerating the nucleation of ParM polymerization on DNA, given that the nucleation step is a kinetic hurdle in the polymerization of eukaryotic actin. However, when the rate of ParM nucleation was directly measured in vitro, it was found to be 300 times faster than that of eukaryotic actin [111]. The critical concentration for polymerization of ParM in vitro was also found to be much lower than its in vivo cellular concentration [111, 144, 145, 148]. Finally, ParM filaments in vitro were found to be very unstable, constantly alternating between phases of elongation and rapid disassembly, a process called dynamic instability [111, 144, 145]. In light of these in vitro findings, it seems likely that ParR/parC regulates ParM filament stability, rather than nucleation. This data suggests an in vivo model whereby ParM filaments are bound at both ends to two different plasmid clusters through ParR, which protects them from their inherent instability and allows them to elongate to sufficient lengths to separate the plasmids to opposite ends of the cell [111, 145]. The polymerization of ParM in between ParR would provide the force needed to separate the plasmid clusters [111, 144, 145]. Recently, this model was confirmed with an in vitro reconstitution of plasmid segregation with purified ParM, ParR, and parC coupled to microspheres [144]. In the presence of ATP and ParR, ParM forms dynamic astral filaments on the surface of the parC-coated beads. In addition, it is possible to see ParM filaments connecting two or more beads, forming a spindle-like structure similar to that observed in vivo, which elongates to push the beads apart. It is even possible to witness ―search and capture‖ events, where filaments extending off the surface of one bead reach out and contact a neighboring bead. The elongation of ParM filaments in between beads then results in their separation. This in vitro reconstitution experiment demonstrate that ParM polymerization, coupled to ParR/parC, can directly exert force on an object, mirroring actin polymerization driving the intracellular movement of Listeria monocytogenes as well as

P a g e | 23

whole cell motility in eukaryotic cells [149]. Furthermore, the reconstitution experiments provide additional support for the model that ParR-binding to ParM stabilizes the filaments in between two segregating pieces of DNA. Long filaments of ParM were only observed in between two beads bound by ParR. If the spindle was severed with laser ablation, the newly generated free end of the filament, which was not bound to ParR, rapidly depolymerized [144]. It is not yet known whether other plasmids use a partitioning mechanism similar to the ParM-pushing mechanism. However, a newly discovered actin homolog from Bacillus subtilis, AlfA, required for the segregation of plasmid pLS32, resembles ParM in its intracellular localization pattern and dynamics [136]. This in vivo data suggests that AlfA may also be able to push plasmids through the cytoplasm like ParM, though this hypothesis remains to be tested.

24 | P a g e

CHAPTER 1

FIGURE 1

P a g e | 25

A mitotic-like pull Just as polymerization can generate a pushing force, depolymerization can generate a pulling force. In eukaryotic cells, for example, the energy released upon depolymerization of microtubules can be harnessed by a complex of proteins attached to chromosomes to drive their segregation [152]. Recent evidence from the bacteria Vibrio cholerae suggests an analogous pulling mechanism may exist in as well [146]. V. cholerae has two chromosomes, each encoding their own partitioning genes and exhibiting distinct dynamic behaviors upon segregation [153, 154]. The mechanism of segregation of Chromosome II is not yet known, but recent work indicates that Chromosome I may be pulled to the pole by a ParA homolog (Fig 1B) [146]. While ParA and ParM are similarly named (the ―par‖ referring to a role in partitioning), they are very different at both the sequence and structural levels. At the beginning of the cell cycle, ParA of Chromosome I (ParAI) localizes to the oldest pole of the cell with a DNA-binding protein ParB and the origin of replication [146]. ParAI is also found in the opposite half of the cell in a haze, which can be resolved into filamentous structures with deconvolution microscopy. Once the origin duplicates, one copy stays at the old pole, along with foci of ParB and ParAI. The other ParB-origin complex moves across the cell at the edge of the filamentous ParAI structure. As the migration of the chromosome progresses, the ParAI zone shrinks toward the new pole, until finally condensing into a polar focus. In the absence of ParAI, the origin does not exhibit the same dynamic behavior [146]. The two ParB-origin complexes can be seen ―floating‖ in the cytoplasm, no longer restricted to the pole. This data suggests that not only does ParAI have a role in directing one of the origins to the new pole, but also in keeping the other origin at the old pole. Interestingly, in the absence of ParAI, the two origins are still able to move apart from one another, but they do so inefficiently and bidirectionally, rather than asymmetrically as in the wild type cell. It is possible that a partitioning mechanism that directs chromosome II movement is able to partially compensate for the absence of ParAI. Unlike ParM, ParAI does not localize in between the two separating DNA loci; instead it localizes in between one of the duplicated loci and the new pole of the cell (Fig 1). It seems clear, then, that ParAI and ParM act through distinct mechanisms. In V.

26 | P a g e

CHAPTER 1 cholerae, the data is consistent with a model in which ParAI is pulling the duplicated origin in toward the new pole, similar to the way that microtubules direct the motion of chromosomes in eukaryotic cells. It has not yet been shown that this particular ParA homolog can self associate into filaments in vitro, but it is capable of making filamentous structures in vivo. Additionally, mutations in the ATPase domain of ParAI abolish its dynamic behavior and ability to position the origin, which is consistent with a role for depolymerization in mediating segregation [146]. Visualization of ParAI filamentous structures in vivo with high resolution electron microscopy would confirm the existence of a cytoskeletal structure and help to clarify the mechanism of chromosome movement. Interestingly, in C. crescentus, clusters of filaments tethered to the pole of the cell have been seen that are of unknown protein composition [155]. C. crescentus and V. cholerae both localize a chromosomal origin to the pole [80, 153, 154]. Additionally, C. crescentus ParA is essential for chromosome segregation and shares considerable homology with V. cholerae ParAI [147]. Therefore it is possible these two ParA homologs use similar mechanisms and that the electron micrographs have captured the C. crescentus ParA in the act of separating the chromosome. To confirm this hypothesis it will be necessary to see if the ―polar filaments‖ are missing in a ParA-depleted C. crescentus strain. Recently, a ParA homolog in Rhodobacter sphaeroides, PpfA, was found to be required for segregating clusters of a chemotaxis protein, TlpT [156]. This discovery is the first example of a ParA homolog mediating the segregation of protein, rather than DNA. The mechanism of this segregation is unknown. It is likely to be distinct, however, from that of V. cholerae ParAI, given their different localization patterns. TlpT clusters segregate to the quarter-cell positions, rather than at the poles [156]. The pattern of segregation actually resembles that of V. cholerae Chromosome II [153, 154], as well as that of certain plasmids in E. coli, such as F and pB171, which are also regulated by related by filament-forming ParA homologs [128, 130, 135].

Dynamic scaffolding Cytoskeletal proteins do not have to actively move proteins or DNA through the cytoplasm in order to influence cellular organization. A higher order cellular structure

P a g e | 27

formed by cytoskeletal proteins can act as a scaffold to direct the localization of other cellular components. Just one of many possible examples from eukaryotic cells is the microtubule network in plants, which forms a scaffold for cell wall synthesis, directing the placement of the synthetic enzymes and the orientation of the cellulose microfibrils [157]. Theoretically, cells can form scaffolds with anything capable of forming a higher order structure. The polymerization of a cytoskeletal protein, however, is coupled to nucleotide hydrolysis. Therefore, the assembly and disassembly of cytoskeletal scaffolds is rapid, reversible, and can be easily regulated with modulators of polymerization. This dynamic property allows the structure to be intimately integrated with the cell cycle and responsive to its intracellular environment. In bacteria, there are at least two examples of such dynamic cytoskeletal scaffolds (Figure 2).

FtsZ: organizing the division apparatus The tubulin homolog, FtsZ, is a scaffold for several proteins involved in cytokinesis. Its filaments form a ring-like structure that defines the division plane [158]. In the absence of FtsZ, all of the other proteins involved in division fail to correctly localize and the cell cannot initiate constriction (reviewed in [52, 92]). This Z-ring structure is stabilized and bound to the membrane largely by the protein FtsA, another actin homolog [159-161]. Numerous different molecules are then recruited to this site. In E. coli, FtsA is thought to directly recruit downstream effectors to the site of the FtsZ ring [52, 162, 163], whereas in B. subtilis, the newly discovered proteins SepF and YlmF are also involved [159, 164, 165]. FtsZ may also be directly exerting force on the membrane during constriction, though there is currently little direct evidence to support this idea. The FtsZ ring, perhaps more than any other cytoskeletal structure, demonstrates the importance of dynamic behavior in a scaffold. Photobleaching experiments have shown that the Z-ring of E. coli and B. subtilis is continually remodeling with a half-time of 8-9 seconds [166]. Additionally, low levels of filaments can be seen rapidly traversing the cell along a helical path [167]. Mutations that affect the GTPase activity of FtsZ alter its dynamic behavior and do not fully support cytokinesis [166]. This dynamic behavior of FtsZ allows it to quickly sample the cell and optimally position itself between the replicated chromosomes. Specific mechanisms for integrating Z-ring assembly with

28 | P a g e

CHAPTER 1 chromosome segregation in E. coli, B. subtilis, and C. crescentus have been characterized [49, 52, 53]. While each organism accomplishes this task differently, they all link a sensor for the intracellular position of DNA to a regulator of FtsZ polymerization (See ―Finding the middle‖). For example, in C. crescentus, the origin of replication recruits a direct inhibitor of FtsZ polymerization, MipZ [53]. When the origins duplicate and segregate to the poles, MipZ travels with them and FtsZ is forced to polymerize at midcell, the area of lowest inhibitor concentration. This mechanism ensures a dynamic coupling between chromosome segregation and cytokinesis. The FtsZ ring is also responsive to changing intracellular conditions. In B. subtilis, the initiation of sporulation causes the Z-ring to be moved from a midcell to a polar site [168]. In E. coli, DNA damage induces the expression of a specific FtsZ inhibitor, SulA, that arrests cell division [169]. Again, the rapid turnover of FtsZ filaments allows these responses to occur within seconds.

MamK: lining up magnetosensing organelles The actin-like protein MamK forms a linear scaffold in magnetotactic bacteria

[137]. These bacteria form crystals of either Fe3O4 or Fe3S4 within membrane-bound clusters that line up along the long axis of the cell (for review, [170]). With both cryoelectron tomography and fluorescence microscopy, MamK can be seen forming filaments right under the magnetosensing organelles in Magnetospirillum magneticum [137]. In the absence of MamK, no filaments are observed, the organelles disperse throughout the cytoplasm or aggregate, and the organism is no longer magnetotactic [137]. The polymerization of MamK has not yet been studied in vitro. In addition, its dynamic behavior in vivo is only beginning to be investigated [139]. Presumably the structure must be able to disassemble at division to allow separation of the two daughter cells, but this process has not yet been observed in live dividing cells. It is also not known whether MamK is involved in segregating the magnetosensing organelles at division.

The evolution of complexity Both prokaryotes and eukaryotes seem capable of using cytoskeletal polymers to actively and dynamically organize their intracellular space. Archeal homologs of FtsZ

P a g e | 29

and MreB have been found as well, though very little is known about their in vivo function. Nonetheless, the presence of this activity in eukaryotic and prokaryotic organisms argues that there is a fundamental need for dynamic polymers in cellular life and an ancient origin of polymer motors and scaffolds. Dynamic cytoskeletal polymers were thought to be vitally important in the evolution of eukaryotic cells, enabling them to grow orders of magnitude larger in size and accomplish more complicated tasks than the morphologically simpler bacteria and archaea [171]. Now that it is clear that prokaryotic organisms have cytoskeletal proteins that are fundamentally similar in structure and activity to eukaryotic actin and tubulin, how is it that the eukaryotes have evolved a cellular biology that is so much more complex? Currently, there is too little data on the bacterial cytoskeleton to definitively answer this question. As it stands now, there does seem to be a difference in the functional diversity between eukaryotic and prokaryotic . Eukaryotic actin and tubulin are responsible for the specific localization of thousands of molecules, and are involved in many different mechanical processes. In order to accomplish this feat, eukaryotes have evolved a huge repertoire of binding partners and regulatory proteins for actin and tubulin. In contrast, prokaryotic cytoskeletal elements have so far been shown to be specialized for one particular function, be it the segregation of a particular DNA locus or the localization of one particular protein. MamK, for example, is only involved in lining up the magnetosensing clusters and cannot substitute for E. coli MreB in cell wall synthesis [138, 139]. The one key exception to this observation is C. crescentus MreB, which has been implicated in chromosome segregation, protein localization, and cell wall synthesis [39, 40, 117, 120, 142, 143]. There is not yet sufficient evidence to show that MreB has a direct role in all of these processes. The effect MreB has on polar positioning of proteins may be secondary to its role in directing cell wall growth [138]. It will be important to determine whether MreB is able to directly participate in more than one cellular process and, if so, capable of using different types of mechanisms. The ability to use the same cytoskeletal proteins in many different processes may have been critically important in the evolution of the complex eukaryotic organisms that

30 | P a g e

CHAPTER 1 we see today [171]. Motor proteins that move along cytoskeletal filaments, such as and , may have also been a key development. Such proteins have not yet been found in bacteria and might not exist. They may have been a recent development that evolved to improve the efficiency of intracellular motility and organization and thereby allow for more complex functions [171].

FIGURE 2

P a g e | 31

HOW DOES MREB FUNCTION?

Bacteria are ubiquitous and important components of nearly every biological ecosystem. Experiments that reveal aspects of their cell biology could eventually have important medical, environmental, and technological applications as well as contribute to our understanding of the evolution of complex biological life. The gene mreB is essential for many important bacterial species. Understanding how it contributes to bacterial growth and division will provide insight into the fundamental cell biology of these organisms and potentially the evolution of the eukaryotic cytoskeleton. Furthermore studying MreB could provide further insight into a question that is central to all cell biology: how cells use one-dimensional genetic information to create an organized, dynamic, three-dimensional cell. In this work, I chose to focus on the role of MreB in cell shape determination. There is clear evidence to show that the primary and perhaps most well conserved function for MreB is in the generation of non-spherical cell shape, and thus I chose to focus my efforts on how MreB functions in this particular process. I chose to work in the Gram-negative organism Caulobacter crescentus (hereafter simply called Caulobacter) for several reasons. Caulobacter is genetically tractable and has a dimorphic life cycle that has been extensively studied. To date this is the only organism that has been shown to have representatives from all classes of the bacterial cytoskeletal proteins, allowing us to explore the potential interactions between MreB and other cytoskeletal components. Lastly, MreB is known to change its cellular localization pattern over the course of the cell cycle [39, 117]. Since Caulobacter cultures are synchronizable, I reasoned that factors that control the localization and dynamic behavior of MreB could potentially be more easily studied in this organism than other model systems. While we know that MreB is required for perpetuating a rod-shape in many model bacterial organisms, including Caulobacter, and is structurally similar to the eukaryotic cytoskeletal protein actin, many questions remain. If MreB forms filaments in cells like actin, what do these filaments look like and what are the biochemical properties of assembly in vitro? Are these filaments forming a dynamic scaffold to localize peptidoglycan synthesis machinery as FtsZ does for components of the division

32 | P a g e

CHAPTER 1 machinery [49, 52]? Are these filaments generating force to push molecular components around the cytoplasm as ParM pushes plasmid DNA [144] or to locally affect the properties of the and/or peptidoglycan to stimulate growth as perhaps FtsZ does at division [172]? How does ATP binding and hydrolysis contribute to MreB’s role in cell shape and the dynamics of MreB localization? What are the other proteins that act with MreB? I was inspired by these questions at the beginning of my thesis work. The work that I will describe here begins to address the mechanism by which MreB contributes to cell shape.

P a g e | 33

34 | P a g e

Chapter 2 Two Independent Spiral Structures Control Cell Shape in Caulobacter

Natalie A. Dye, Zachary Pincus, Julie A. Theriot, Lucy Shapiro, and Zemer Gitai

Work previously published in the Proceedings of the National Academy of Sciences [142]

P a g e | 35

ABSTRACT

The actin homolog MreB contributes to bacterial cell shape. Here, we explore the role of the coexpressed MreC protein in Caulobacter and show that it forms a periplasmic spiral that is out of phase with the cytoplasmic MreB spiral. Both mreB and mreC are essential, and depletion of either protein results in a similar cell shape defect. MreB forms dynamic spirals in MreC-depleted cells, and MreC localizes helically in the presence of the MreB-inhibitor A22, indicating that each protein can form a spiral independently of the other. We show that the peptidoglycan transpeptidase Pbp2 also forms a helical pattern that partially colocalizes with MreC but not MreB. Perturbing either MreB (with A22) or MreC (with depletion) causes GFP-Pbp2 to mislocalize to the division plane, indicating that each is necessary but not sufficient to generate a helical Pbp2 pattern. We show that it is the division process that draws Pbp2 to midcell in the absence of MreB’s regulation, as cells depleted of the tubulin homolog FtsZ maintain a helical Pbp2 localization in the presence of A22. By developing and employing a novel computational method for quantitating shape variance, we find that a FtsZ-depletion can also partially rescue the A22-induced shape deformation. We conclude that MreB and MreC form spatially distinct and independently localized spirals and propose that MreB inhibits division plane localization of Pbp2 while MreC promotes lengthwise localization of Pbp2; together these two mechanism ensure a helical localization of Pbp2 and thereby the maintenance of proper cell morphology in Caulobacter.

36 | P a g e

CHAPTER 2

INTRODUCTION

Prokaryotes exhibit a wide variety of cell shapes (including rods, spheres, spirals, squares, and stars), but the mechanisms by which these shapes are achieved are poorly understood. The extracellular peptidoglycan layer provides structural rigidity for bacterial cells and is of central importance in the establishment and maintenance of cell shape [13, 17]. This layer is a meshwork of disaccharide chains (alternating N-acetylglucosamine and N-acetylmuramic acid sugars) crosslinked by short peptide bridges. Rod-shaped bacteria are believed to possess two peptidoglycan synthesis complexes with distinct activities: one for elongation along the cell length and the other for cell division [13, 26, 173]. It is thought that maintenance of a regular rod shape requires the activities of these two complexes to be carefully balanced [173]. The bacterial cytoskeleton also plays a role in the establishment and maintenance of cell shape [17]. Homologs for the three major types of eukaryotic cytoskeletal elements have been identified in bacteria: the actin homolog is MreB, the tubulin homolog, FtsZ, and the intermediate filament homolog, Crescentin [174]. Of these known cytoskeletal elements, MreB is the only one required to establish an underlying rod-like character. Mutations in mreB confer a spherical-like morphology to the normally rod-like cells of E.coli, B.subtilis, and Caulobacter [38, 39, 69, 71, 116, 117, 141, 175]. Additionally, there is a striking phylogenetic correlation: most species that possess an mreB-like gene have non-spherical shapes [116]. MreB-like proteins appear to influence morphology by regulating the site of peptidoglycan synthesis. Fluorescently-tagged vancomycin labels sites of new peptidoglycan assembly by binding the D-ala-D-ala ends of nascent peptidoglycan subunits (prior to crosslinking by a transpeptidase) [43]. In B.subtilis, insertion of precursors was found to occur both at the division plane and in a helical pattern along the perimeter [43]. A deletion of the mreB-like gene, mbl, disrupted the helical—but not the septal—insertion of peptidoglycan precursors. MreB homologs are known to form helices in E.coli, B.subtilis, and Caulobacter [39, 59, 116, 117, 176-178]. Therefore a reasonable hypothesis is that MreB ensures a helical pattern of cell wall growth during elongation by directly positioning the peptidoglycan precursors along its own spiral scaffold; however,

P a g e | 37

there is not yet any direct evidence for an interaction between the peptidoglycan (or any of its modifying enzymes) and MreB. MreB is thought to act with the coexpressed MreC protein to regulate cell shape [141, 179, 180]. MreC resides primarily external to the cytoplasmic membrane, anchored by a single transmembrane domain near the N-terminus [179]. In most bacteria mreB is in an operon with mreC, and disruptions of mreC in E.coli and B.subtilis confer the same spherical morphology as disruptions in mreB [141, 181, 182]. In B.subtilis, MreC adopts a spiral localization and is required for the helical distribution of peptidoglycan precursors [180]. Here, we explore the in vivo spatial and temporal organization of components of the mre operon (Fig 1A) in order to better understand how the actin cytoskeleton of bacteria contributes to cell shape. We have focused our investigation on Caulobacter crescentus, in which events of the cell cycle can be easily followed and the dynamics and functions of MreB have been relatively well studied [39, 40, 117, 120]. Caulobacter has a unique life cycle: it begins as a motile ―swarmer‖ cell, differentiates into a ―stalked‖ cell (shedding the flagellum and growing an appendage called the ―stalk‖), and then divides asymmetrically to produce one swarmer cell and one stalked cell [183]. We demonstrate that MreC localizes to a spiral that does not overlap with the MreB spiral and does not require MreB activity to form. We also find that Pbp2 (a homolog of the E.coli elongation-specific peptidoglycan transpeptidase whose gene also lies in the Caulobacter mre operon, Fig 1A) forms a helical pattern that partially overlaps with that of MreC but not MreB. In the absence of either spiral, Pbp2 localizes to the division plane in a FtsZ-dependent manner. Finally, we develop a novel computational shape analysis method to quantitatively show that a FtsZ-depletion can partially rescue the shape defect resulting from inactivation of MreB. We propose a model in which MreC and MreB form separate and independently localized spiral structures in Caulobacter that are both required for the proper localization of a transpeptidase and, thereby, the generation of a rod-like shape.

38 | P a g e

CHAPTER 2

RESULTS

MreC is helical in Caulobacter To determine its subcellular localization, a C-terminal fusion of MreC to mRFP1 was chromosomally integrated at the site of the inducible promoter, Pxyl [184]. The resulting strain (LS4272) carries an unlabeled copy of mreC at the endogenous locus and a xylose-inducible mRFP1-tagged copy at the Pxyl locus. MreC-mRFP1 is fully functional (see below), and represents the endogenous protein distribution, as immunofluorescence microscopy reveals a similar pattern [185]. MreC-mRFP1 localized to several puncta, alternating in a zig-zag fashion along the perimeter of the cell (Fig 1B), a distribution reminiscent of known helical proteins such as MreB. Using 3D deconvolution microscopy with an MreC-mCherry fusion (less prone to photobleaching than mRFP1 [186]), it was possible to resolve a continuous helical pattern for MreC (Fig 1C). Given its predicted topology [179], we believe this spiral pattern for MreC lies primarily in the periplasm. MreC did not dramatically alter its localization pattern through the cell cycle, as all cells contained full-length spirals (Fig 1B). Unlike MreB [39, 117], MreC did not form a division-plane ring in predivisional cells; in contrast, we often observed a gap in the MreC pattern at this site (arrows in Fig 1B). MreC was also detected in puncta in the stalk (arrowheads in Fig 1B).

MreC does not colocalize with MreB Caulobacter MreB has been previously reported to form a dynamic spiral that condenses into a ring at the division plane of early predivisonal cells and then expands back into a full, longitudinal spiral prior to division [39, 117]. Since MreC also forms a spiral, we explored whether MreC and MreB colocalize in the cell. A yfp-mreB construct was inserted in single copy on the chromosome at the site of the nitrate-inducible promoter (Pnit) of LS4272 (creating a strain, LS4279, that carries yfp-mreB and mreC- mrfp1 in addition to chromosomal copies of the unlabeled genes). Surprisingly, the patterns for MreB and MreC did not overlap (Fig 1D): when both proteins formed spirals, they interdigitated; when MreB was in a ring, MreC was absent from this site (consistent with the gap observed at the division plane in LS4272,

P a g e | 39

arrows in Fig 1B). Simultaneously labeling MreB with two different fluorescent markers (Pxyl::mcherry-mreB, Pnit::yfp-mreB, LS4282) resulted in considerable colocalization, indicating that the antilocalization seen between MreB and MreC is not an artifact of the imaging technique (Supplemental Fig S1). We also performed 3D deconvolution microscopy on a strain expressing both GFP-MreB and MreC-mCherry (LS4285) and, once again, observed nonoverlapping patterns (Fig 1E). To directly explore the dynamic nature of these two proteins in the same cell, we isolated a synchronized population of swarmer cells and imaged them over the course of a cell cycle (Supplemental Movies 1 and 2). At no time during the cell cycle were MreB and MreC found to significantly colocalize. Though it is possible that we were not able to resolve subtle or transient instances of colocalization, it is clear that the predominant localization sites for these proteins do not overlap.

FIGURE 1

40 | P a g e

CHAPTER 2

mreC is essential and required for proper cell shape Given the surprising result that Caulobacter MreC is localized to a helix distinct from that of MreB, we examined the null phenotype of mreC in Caulobacter. We were only successful in deleting mreC in the presence of a cosmid containing the mre operon, suggesting that mreC is essential. We then created a depletion strain (LS4275) containing a deletion of mreC at the endogenous locus and a plasmid-borne copy of mreC-mrfp1 under the Pxyl inducible promoter. Growth of this strain depended on the presence of xylose in the media, confirming that mreC is essential. Because the only copy of MreC was labeled with mRFP1, it was possible to use fluorescence microscopy on live cells to monitor the presence of MreC during a depletion. When grown in xylose, the strain exhibited a normal morphology (Fig 2B) and grew at rates similar to wild type (data not shown), indicating that the mRFP1-tagged version of MreC is fully functional. When grown without xylose, the viability of the strain declined drastically after 12 hours in rich media (Fig 2A), and the cells became ―lemon-shaped‖ with increased width but pointed poles (Fig 2B). This morphology is identical to cells depleted of MreB [39, 117] or treated with A22, a chemical inhibitor of MreB [40, 187].

MreC and MreB localize independently of one another To investigate the possibility that the formation of the MreC spiral is dependent on the presence of MreB structures, we examined the localization of MreC-mRFP1 after perturbing MreB activity with the drug, A22. Since A22 rapidly and reversibly delocalizes MreB in Caulobacter [40], we believe it enables a more direct assay of MreB’s effect on positioning MreC than a depletion of MreB, which gradually removes the protein over many cell cycles. If MreB were required to establish an MreC spiral, MreC-mRFP1 expressed in the presence of A22 would not be able to localize helically. Therefore, we added A22 to cells of LS4279 (Pxyl::mreC-mrfp1, Pnit::yfp-mreB) to disrupt MreB localization prior to inducing MreC-mRFP1 expression; expression of MreC-mRFP1 was then induced for six hours in the continued presence of A22. While YFP-MreB was delocalized in these cells,

P a g e | 41

MreC-mRFP1 that was synthesized in the presence of A22 still formed a spiral pattern (Fig 2C). This spiral pattern for MreC was also maintained if A22 was added to cells that were already expressing MreC-mRFP1 (Supplemental Fig S2). Since the spiral formation of MreC was found to be independent of MreB, we investigated the possibility that MreB localization depends on MreC. Accordingly, we examined LS4278 (ΔmreC, Pxyl::mreC-mrfp1, Pnit::yfp-mreB) after simultaneously depleting MreC and expressing YFP-MreB. In the absence of MreC, cells lost their proper shape, but helices and rings of YFP-MreB were still observed (Fig 2D). MreB remained dynamic in MreC-depleted cells, as YFP-MreB viewed over the course of a cell cycle still properly transitioned between spiral and ring states (Supplemental Movie 3). The polar localizations of two histidine kinases, PleC and DivJ, which have been shown to be dependent on the presence of MreB [117], were also correctly localized after MreC- depletion (Fig 2E-F). From these results we conclude that MreB remains correctly localized, dynamic, and functional in the absence of MreC.

FIGURE 2

42 | P a g e

CHAPTER 2

Pbp2 forms a spiral-like pattern that partially colocalizes with MreC Another gene in the Caulobacter mre operon is pbp2 (also called mrdA, Fig 1A), a homolog of the E.coli elongation-specific transpeptidase (an enzyme that catalyzes peptidoglycan crosslinking). Like MreC, Pbp2 is predicted to lie primarily in the periplasm but possess a cytoplasmic N-terminal domain. Since mreB and mreC both affect cell shape and lie in the same operon with pbp2, we explored the possibility that MreB and MreC regulate morphology by acting on Pbp2. We fused GFP to the N- terminus of Pbp2 and inserted the construct at the Pnit locus of the chromosome (LS4287, which also possesses unlabeled pbp2 at the endogenous locus). Expression of GFP-Pbp2 revealed a nonuniform distribution of puncta and bands (Fig 3A), reminiscent of the helical localization of MreB and MreC. The GFP-Pbp2 pattern closely resembles the Pbp2 distribution observed using immunofluorescence microscopy [39], indicating that the GFP fusion protein accurately represents Pbp2 localization. Like MreC, GFP- Pbp2 did not change its localization dramatically through the cell cycle, but it did appear to be excluded from the division plane of predivisional cells. Given that both MreB and MreC are helical but antilocalized, we simultaneously labeled Pbp2 and either MreB or MreC. Using 3D deconvolution microscopy, we observed that GFP-Pbp2 often colocalized with MreC-mCherry (LS4335, Pxyl::mreC- mcherry, Pnit::gfp-pbp2, Fig 3B) but avoided the sites of YFP-MreB spirals and rings (in LS4289, Pxyl::mcherry-mreB, Pnit::gfp-pbp2, Fig 3C).

Pbp2 mislocalizes to the division plane in the absence of either MreB or MreC structures We explored the dependence of Pbp2’s helical localization on MreC, using an MreC-depletion, and on MreB, using an A22-treatment. LS4288 (ΔmreC, Pxyl::mreC- mrfp1, Pnit::gfp-pbp2) was grown under conditions to express GFP-Pbp2 while depleting MreC-mRFP1 (M2GN lacking xylose) (Fig 3D). The same strain was treated with A22 in PYE+xylose and then switched into M2GN+xylose to induce GFP-Pbp2 in the presence of A22 for six hours (Fig 3E). Surprisingly, both treatments had the same effect: GFP- Pbp2 mislocalized to a band at the incipient division plane of stalked and predivisional cells (arrows in Fig 3D, E). While some GFP-Pbp2 appeared in puncta in the cylindrical region of the cell, the division plane or midcell region appeared to be the dominant site.

P a g e | 43

FIGURE 3

Interestingly, we believe it is specifically newly synthesized Pbp2 that is mislocalized. GFP-Pbp2 synthesized prior to A22-treatment did not rapidly relocalize (Supplemental Fig S3A). In addition, if cells expressing GFP-Pbp2 received A22 and were then switched into media where GFP-Pbp2 was no longer expressed, the previously synthesized GFP-Pbp2 did not accumulate at midcell (Supplemental Fig S3B). These results suggest that Pbp2 localization is regulated at the level of insertion and that established helical Pbp2 patterns can be stable in the absence of MreB. This finding may explain the observation that A22’s effect on cell shape manifests much later than its effect on MreB localization [40]. Given the phenotypic similarity between an MreC- depletion and an A22-treatment, we believe MreC also regulates sites of Pbp2 insertion; however, we were unable to test this hypothesis due to the length of time required to deplete MreC.

44 | P a g e

CHAPTER 2

FtsZ depletion prevents A22-induced mislocalization of GFP-Pbp2 The tubulin homolog, FtsZ, localizes to the incipient division plane before any other known protein and is required for the recruitment of all the other proteins that are necessary for division [51]. It seemed possible, therefore, that the abnormal accumulation of GFP-Pbp2 at the division plane, observed in the absence of either MreC or MreB spirals, might also be FtsZ-dependent. To explore this hypothesis, the Pnit::gfp-pbp2 construct was introduced into a FtsZ-depletion strain in which the expression of FtsZ is xylose-dependent [41]. Since FtsZ is normally absent from swarmer cells [188], a population of cells lacking FtsZ can be easily obtained by resuspending freshly isolated swarmer cells (of a FtsZ-depletion strain) in media lacking xylose. A22 was added to a synchronized population of LS4290 (ΔftsZ, Pxyl::ftsZ, Pnit::gfp-pbp2); the cells were then grown for six hours in M2GN media containing either xylose (FtsZ+) or glucose (FtsZ-). Remarkably, newly-synthesized GFP-Pbp2 accumulated at the division plane in FtsZ+ cells but was able to localize helically in FtsZ- cells upon A22 treatment (Fig 4). Thus the recruitment of Pbp2 to the division plane in the absence of MreB structures is FtsZ-dependent. Additionally, this result shows that MreB is not required for helical Pbp2 localization when the cells are not dividing.

FIGURE 4

P a g e | 45

Quantitative shape analysis demonstrates that a FtsZ-depletion also prevents A22- induced shape defect If the mislocalization of Pbp2 is part of the mechanism by which A22-treated cells lose their proper shape, we may expect that a FtsZ-depletion, by preventing Pbp2 mislocalization, would also mitigate the cell shape defect induced by A22. However, the shape transformation observed upon A22 treatment is complex. Simple linear measurements (i.e., width or length) would not only be inadequate to capture this non- rigid deformation process but would also be difficult to make unambiguously on the crescent-shape Caulobacter cells. Thus, we sought a way to comprehensively represent cell shape such that quantitative differences between populations could be accurately determined. We developed a method to identify the principal modes of shape variation in our dataset and quantify each cell’s position along those modes [189, 190]. We mathematically represent the shape of each cell by transforming its image into a ―signed distance map‖ (which records the distance from every pixel in the image to the nearest shape edge). The signed distance map for every cell in the dataset is plotted in a high- dimensional space; in this space, addition and scalar multiplication of shapes is well- defined and meaningful (Supplemental Fig S4). We calculate the mean shape and use principal components analysis (PCA) to identify the major modes of variation. Each point in the high-dimensional space is then re-represented in terms of its location along these major axes. In this way, the shape of a given cell can be quantified with only a few highly meaningful parameters that represent the complex shape variation present in the data. We applied this PCA shape analysis to the cells of LS4290 grown under four different conditions: FtsZ+, FtsZ-, FtsZ+A22+, and FtsZ-A22+. For this experiment, swarmer cells of LS4290 were isolated and grown in M2GN with or without xylose (FtsZ+ and FtsZ-, respectively) and with or without A22. Imaging was performed after six hours of growth, a time period that allowed the cells to begin a shape transformation but remain viable. Cells from all four classes were collected into a single dataset for the PCA analysis. We selected the top three modes, which together accounted for >95% of the total variance. Of the three parameters chosen for analysis, the primary mode (roughly

46 | P a g e

CHAPTER 2 describing length and width) encompassed the vast majority of the variation in the total dataset (91.85%) and is gratifyingly consistent with the qualitative observations that have been made of FtsZ-depleted and A22-treated cells (Fig 4). When plotted along this axis, FtsZ-depleted cells fall toward the negative end (corresponding to long and narrow shapes) whereas A22-treated cells fall toward the positive end (corresponding to short and wide shapes) (Fig 4). Amazingly, the FtsZ-A22+ distribution appeared closer to the non-A22 treated FtsZ+ (wild type) and FtsZ- (filamentous) distributions than that of FtsZ+A22+, indicating that the absence of FtsZ impaired the ability of A22 to change cell shape (Fig 4). Interestingly, when plotted along either the second or third axis of variation, it was evident that the A22-treated cells (regardless of FtsZ content) were straighter than non-A22-treated cells, indicating a possible role for MreB in generating curvature (Supplemental Fig S5). To visualize how shape is described by the combination of these three parameters, we used the NCSA Partiview viewer [191] to create a ―shape map‖: in this representation the actual bacterial shapes are plotted at the Euclidian point corresponding to its shape parameters (see Supplemental Fig S6, Movies 4-5). We used all three significant PCA shape parameters to quantitate the differences in shape by first calculating the Euclidian distance separating the medians of the distributions in the 3-dimensional PCA space (see Methods). We then used Monte-Carlo permutation tests to determine which of the three test groups were closest to wild type (FtsZ+). Through this analysis we were able to determine that the FtsZ-A22+ class was significantly closer to wild type (FtsZ+) than was either FtsZ+A22+ (p<0.0001) or FtsZ- (p=0.0294). Thus, we quantitatively conclude that a FtsZ-depletion rescues the A22- induced shape defect. However, the rescue is only partial since the FtsZ-A22+ group remained significantly different from wild type (FtsZ+, p=0.0007). Because the FtsZ-depletion is able to rescue Pbp2 mislocalization, as well as the cell shape defect, of A22-treated cells, we conclude that a helical pattern of Pbp2 is critical to the generation of a rod-like shape and that MreB and MreC contribute to the maintenance of proper shape by regulating the localization of Pbp2.

P a g e | 47

DISCUSSION

It has been previously shown that MreB forms a spiral in live bacterial cells [59, 117, 177, 178], presumably due to its ability to polymerize into filaments [8]. Spiral localizations have also been observed for other proteins that polymerize (such as FtsZ [167, 168] and MinD [59]), proteins that cannot polymerize (such as Pbp2 [39] and SetB [192]), and even non-protein molecules (such as LPS [193] and nascent peptidoglycan [43]). Here, we show that MreC appears as a spiral in the periplasmic compartment of Caulobacter cells. While the spiral localization is not unprecedented, we were surprised to find that the MreC pattern is both nonoverlapping with and independent of MreB structures. It seems reasonable that MreB, much like actin in eukaryotic cells, could form a scaffold that directly positions other molecules, making the spiral localizations of non- polymerizing molecules MreB-dependent. This model cannot apply to MreC, however, as it maintains its spiral pattern in the absence of MreB spirals. The mechanism by which MreC localizes helically remains unclear. Even if MreB is not directly positioning MreC molecules into a helical configuration, however, the antilocalization we observe requires there to be some communication between the two. In addition to showing that spiral patterns can have distinct spatial distributions, our results show that they can exhibit different dynamic behaviors: MreB transitions between a spiral and a ring, whereas MreC remains in a longitudinal helix. Though MreB and MreC form separate helices, both are required to correctly position the cell wall transpeptidase, Pbp2, into a lengthwise spiral pattern. The absence of either MreB or MreC spirals causes Pbp2 to mislocalize to a band at the division plane. This phenotype raises two important questions: first, why does Pbp2 localize to the division-plane in the absence of these regulators, and second, how do MreB and MreC prevent this abnormal accumulation? In attempt to address the first question, we must consider that the localization of transpeptidases (which are homologous to Caulobacter Pbp2) can be driven by the availability of their substrate (uncrosslinked peptidoglycan) [194, 195]. The division plane of a dividing cell contains high concentrations of peptidoglycan precursors, as demonstrated with fluorescent vancomycin staining in B.subtilis [43]. These molecules are required to build the new cell wall at division.

48 | P a g e

CHAPTER 2

Normally, they are substrates for the division-specific transpeptidase (Pbp3 in E.coli). It is possible, however, that the peptidoglycan precursors at the division plane could also be substrates for Pbp2 and thereby recruit Pbp2 into the division plane. The appearance of these peptidoglycan substrates at the division plane is known to require FtsZ [43], so our observation that cells deprived of FtsZ never accumulate Pbp2 at the division plane is consistent with this model. Given that we only see Pbp2 localize to the division plane under abnormal conditions (in the absence of MreB or MreC structures), it is possible that the accumulation of Pbp2 at this site could be detrimental to the cell. In support of this idea, we show that rescuing the mislocalization of Pbp2 (induced with A22) with a FtsZ- depletion also partially rescues cell shape, indicating that an accumulation of Pbp2 at the division plane is likely to contribute to a deformation in shape. In this scenario, the cell would need an active mechanism for allowing the accumulation of peptidoglycan precursors at the division plane without recruiting Pbp2. To answer the second question that is raised by the Pbp2 mislocalization result— how MreB and MreC regulate Pbp2 localization—we must consider where in the cell all three of these proteins are under wild type conditions. MreB is present at the division plane in Caulobacter predivisional cells, but Pbp2 and MreC are excluded from this site. Additionally, Pbp2 partially colocalizes with MreC but not MreB. Given these observations, we propose that MreB inhibits the localization of Pbp2 at the division plane, while MreC promotes helical Pbp2 localization along the cell length. This model is supported by recent biochemical evidence demonstrating an interaction between MreC and Pbp2 [185]. Surprisingly, neither mechanism is sufficient on its own to prevent accumulation of Pbp2 at the division plane: if the MreB-ring is missing, Pbp2 is presumably not pushed out of the division plane, and if the MreC-spiral is absent, Pbp2 is not pulled into the longitudinal helix. Since we observed that only newly synthesized Pbp2 is mislocalized in the absence of MreB structures, we believe its localization must be regulated at the level of insertion. Our results suggest a model (Fig 5) whereby Pbp2 can accumulate either in a longitudinal helix or a band at the division plane, each with distinct consequences for cell shape (rod vs. lemon). As stated above, MreB and MreC both promote a helical Pbp2

P a g e | 49

pattern, though through different mechanisms. FtsZ, by organizing the assembly of the division machinery, ends up neutral in this model: while it triggers the accumulation of peptidoglycan substrates to attract Pbp2, it also triggers the formation of MreB-rings to inhibit Pbp2 accumulation at the division plane [39]. In conclusion, we have shown that the maintenance of proper morphology involves a complicated interplay between multiple dynamic spiral assemblies in both the cytoplasmic and periplasmic compartments of the cell.

FIGURE 5

ACKNOWLEDGEMENTS

We thank the members of the Theriot and Shapiro labs for advice and support; Aaron Straight for assistance with deconvolution microscopy; M.R.K. Alley for development of the Pnit inducible promoter; Yves Brun, Craig Stephens, M.R.K. Alley, and Roger Tsien for strains; and Kristina Godek, Patrick McGrath, Kirstin Milks, Sean Murray, Susanne Rafelski, and Aaron Straight for critical review of the manuscript. This work was funded by Stanford Graduate Fellowship (to ND), DOD National Defense Science and Engineering Fellowship (to ZP) and NIH grant RO1 GM51426-11 (to LS).

50 | P a g e

CHAPTER 2

AUTHOR CONTRIBUTIONS

I performed many of the experiments. Zachary Pincus performed the shape analysis and statistics. Julie Theriot, Lucy Shapiro, and Zemer Gitai provided intellectual input and resources.

MATERIALS AND METHODS

Bacterial growth conditions and strain construction Caulobacter crescentus strains CB15N and derivatives were grown at 30°C in PYE rich media or M2G minimal media supplemented with appropriate combination of antibiotics [196]. To induce from Pnit, cultures were grown for 6-16 hours in M2GN media (substitutes 10mM NaNO3 for NH4Cl in standard M2G). To induce from Pxyl [184], cells were grown in the presence of 0.03% xylose for 2-16 hours. A22 was always used at a concentration of 10 μg/ml. Details of the cloning strategies are provided in Supplemental Text. Each fusion construct was cloned into plasmids that cannot replicate in Caulobacter and then introduced separately into CB15N with conjugation or electroporation [196]. Phage transduction (ΦCR30) was used to move all constructs from CB15N into other backgrounds [196].

Depletions To create the MreC-depletion strain we deleted the endogenous gene in the presence of cosmid containing the mre operon (gift of C. Stephens, Santa Clara University), and then traded the cosmid with a low-copy plasmid bearing Pxyl::mreC- mrfp1 (LS4273, see Supplemental Text). To deplete MreC, LS4275 and derivatives were grown overnight in PYE containing 0.03% xylose, washed with two volumes and grown in media containing 0.2% glucose (keeping cultures in log phase). To deplete FtsZ, YB1585 (gift of Y.Brun, Indiana University [41]) and derivatives were grown overnight in media containing 0.3% xylose, synchronized [197], and resuspended in media containing 0.2% glucose.

P a g e | 51

Microscopy Cells were immobilized on 1% agarose/M2G pads (for widefield imaging) or poly-lysine treated coverslips (for deconvolution). Widefield imaging was performed using an Axioplan 2 microscope (Carl Zeiss), equipped with phase-contrast and epifluorescence optics, coupled to a cooled CCD camera (MicroMAX 512 BFT; Princeton Instruments). Cells were viewed with a 100X/NA1.4 objective. 16-bit images were acquired (and converted to 8-bit) with Metamorph software (Universal Imaging). Deconvolution microscopy was performed using an Olympus IX70 microscope coupled to a Photometrics CoolSnap HQ camera. Cells were viewed with a 100X/NA1.4 objective; image acquisition and deconvolution were conducted with SoftWoRx software (Applied Precision). All images were processed in Adobe Photoshop, converting resolution to 300dpi with a bicubic interpolation and adjusting levels and brightness/contrast.

PCA shape analysis See Supplemental Text for a detailed description. In brief, a total of 1071 bacterial cells were isolated with intensity thresholding, vertically aligned, oriented with respect to curvature, and converted into a signed distance map [189, 190]. The properly-oriented signed distance map for each bacterium was then treated as a point in a 9086-dimensional vector space. Principal components analysis (PCA) was applied to the collection of these points, and the top three principal modes were selected. The difference in median shape (in terms of these three parameters) was calculated pairwise for all treatment groups. To determine statistical significance, we employed a Monte Carlo approximation to Pitman’s permutation test of two populations [198, 199].

52 | P a g e

CHAPTER 2

SUPPLEMENTAL TEXT

Creation of fluorescently-tagged constructs: Standard molecular biology techniques were used for all cloning procedures. Oligonucleotide sequences are available upon request (please email [email protected]).

MreC-mRFP1: NdeI and BglII restriction sites were engineered at the first and last condons of the mreC (CC1544) gene respectively, changing the GTG start codon to ATG and removing the stop codon to allow a C-terminal fusion to mrfp1. BglII and EcoRI sites were engineered to the start and stop codons, respectively, of mrfp1. Each gene was inserted downstream of the Pxyl promoter in pXGFP4-C1.

YFP-MreB and mCherry-MreB: A GFP-MreB fusion has been previously described [117]. The GFP in this vector was replaced with either YFP or mCherry (courtesy of R. Tsien, UCSD [186]) using NdeI and KpnI sites engineered to the first and last codons with PCR. To create a Pnit- inducible fusion, the yfp-mreB construct was then moved from pXGFP4-C1 into pXGFP7-C1 (gift of M.R.K. Alley, Anacor Pharmaceuticals, Palo Alto, CA) using NdeI and HpaI sites (LS4276).

GFP-Pbp2: The GFP-Pbp2 fusion was created by inserting the pbp2 gene (CC1546) downstream of gfp in the pXGFP4-C1 plasmid, using KpnI and XbaI sites engineered at the start and stop codons of pbp2. The gfp-pbp2 section was then moved into pXGFP7- C1 with NdeI and HpaI to create a Pnit-inducible version of this fusion (LS4286).

Polarity markers: The DivJ-GFP (under its endogenous promoter) and PleC-GFP (under constitutive Lac promoter) constructs have been previously described [200].

P a g e | 53

Creation of MreC-depletion strain and assessment of viability: To delete the endogenous mreC, regions of homology flanking the mreC (CC1544) gene (500bp on either side) were cloned into the pNPTS138 integration vector (gift of M.R.K. Alley, Anacor Pharmaceuticals, Palo Alto, CA) and the resulting plasmid (LS4274) was conjugated into CB15N. We then selected for the excision of the deletion construct with 3% sucrose. We were not able to obtain any colonies harboring a deletion of the mreC gene unless a cosmid containing the entire mre operon was introduced into the strain prior to the sucrose selection (gift of C. Stephens, Santa Clara University, Santa Clara, CA). Strains bearing chromosomal deletions of mreC were verified by PCR. Pxyl::mreC-mrfp1 was moved from pXGFP4-C1 into pMR10 with NheI-EcoRI digestion and ligation. The pMR10/Pxyl::mreC-mrfp1 plasmid (LS4273) was introduced into the mreC deletion strain (containing the mre cosmid) by conjugation. We then identified a strain that had maintained the pMR10/Pxyl::mreC-mrfp1 plasmid but lost the mre cosmid with replica plating as being Kan-resistant (marker for pMR10), Tet-sensitive (marker for cosmid), and xylose-dependent (marker for inducible mreC). To assess viability, colony-forming units were counted after the switch from xylose-containing to glucose-containing media: every 2-4 hours, aliquots of -5 -6 approximately 10 and 10 OD660 units were plated onto PYE+0.03% xylose. Colonies were counted after 2 days of growth at 30˚C. CFU/OD was calculated and plotted against time of depletion.

A22 treatments: All A22 treatments were performed with a concentration of 10 μg/ml, which is sufficient to delocalize MreB while allowing growth to continue and cells to remain viable for another 6-8 hours [40]. A22 was diluted 1000X into sterile media from a stock solution of 10 mg/ml in methanol. Cultures were then were washed and diluted with media containing a final concentration of 10 μg/ml A22.

PCA shape analysis: Phase-contrast images of the cells in each treatment condition were automatically intensity-thresholded using ImageJ [201] to yield binary ―mask‖ images of the cells.

54 | P a g e

CHAPTER 2

These mask images were manually edited to remove non-bacterial particles and to split touching bacteria into separate objects. Subsequent image processing steps were automated in custom C++ and Python code, which relied heavily on the Insight Toolkit (a suite of tools and methods designed for medical image segmentation and registration) [202]. Individual image objects (four-connected pixel regions of ―bacteria‖ pixels separated by ―background‖ pixels) were isolated from each image, excluding those that touched an image edge. A total of 1071 bacterial shapes were thus isolated (FtsZ+: 214 cells; FtsZ-: 108 cells; FtsZ+A22+: 544 cells; FtsZ-A22+: 205 cells); all were padded to 77118 pixels. Each cell was aligned along its long axis, as determined by the second moments of its spatial pixel distribution. Each axis-aligned mask image was then converted to a signed distance map by setting the value of each pixel to be the distance from that pixel to the nearest bacterial edge and negating the distances for pixels within the bacterium [203]. Because Caulobacter have a distinctive crescent shape, it was necessary to further align the cell shapes so that they all curved in the same direction. This was accomplished through a simple expectation-maximization procedure. First, the average of all of the signed distance maps was calculated. Each bacterium was then mirrored across its long axis if the mirror image had a lower pixel-wise sum of squared differences with the average image than did the original. A new average was calculated and the entire procedure was repeated until no bacteria were flipped (four iterations in our case), after which they all curved in the same direction. The properly-oriented signed distance maps for all cells were treated as points in a 77118 = 9086-dimensional vector space and principal components analysis (PCA) was applied to these points, after the work of Leventon et al. [189, 190]. In principal, PCA can be thought of as finding the single direction in the original space along which the maximum variation in the data set lies. Given that direction, it then finds an orthogonal direction that accounts for the maximum amount of remaining variance, and so on until a new mutually-orthogonal basis set is constructed. In practice, this calculation can be performed in one pass through the singular value decomposition of the data set, or through an eigen decomposition of the covariance matrix of the re-centered data. We use the latter technique, which is accomplished as follows. First, the average signed distance

P a g e | 55

map is computed and subtracted from each image. Then each image is packed into a single row of a 9086N matrix, where N is the number of data points (1071 in our case). If we call this matrix A, the covariance matrix of this re-centered data is simply AAT (the A matrix multiplied by its own transpose). The eigenvectors of AAT are the new PCA basis set, and the eigenvalues of AAT describe the variance of the data along the direction of the corresponding eigenvector. We focused our attention on the set of principal modes that accounted for at least 95% of the variance in our original data set. In our case this was the top three eigenvectors, which accounted for 91.85%, 2.92%, and 1.23% of the variance, respectively. The remaining modes were discarded to remove noise and reduce the dimensionality of the space. Bacterial shapes were parameterized along these three axes of shape variation by projecting the mean-subtracted signed distance map onto each principal mode in turn. Because eigenvectors have unit norm, this projection can be accomplished by simply taking the dot product between the eigenvector and the mean- subtracted distance map. In this manner, each bacterial shape in our dataset was decomposed into three shape parameters. To visualize what each mode of variance ―means‖ in terms of shape changes, we used the principal modes (eigenvectors) to make synthetic bacterial shapes. Given a set of parameters, it is simple to reverse the above decomposition operation: multiply each principal mode by the parameter, then sum the products and add back the mean. This produces a pseudo-signed distance map, which can be converted to a binary shape mask by thresholding all nonpositive (―inside‖) values to one pixel intensity and all positive values (―outside‖) to another intensity.

Statistical Analysis: A straightforward nonparametric permutation approach was taken to assess whether the cell shapes described by the PCA analysis differed significantly between populations. We assumed the null hypothesis that none of the different treatment conditions had any effect on cell shape. If this were the case, then randomly re-shuffling the data between different groups would not change the median shape for each group (since shape is not sensitive to treatment). The Euclidian distance between the median

56 | P a g e

CHAPTER 2 shapes of two treatment groups was calculated. This distance was calculated in the 3- dimensional PCA space, such that all three major modes of variation would contribute to the statistical analysis. Then the data from these two groups were randomly permuted and broken into two new groups (of size equal to the originals). The distance between the median shapes of these two new groups was calculated. After 10,000 iterations of permutation and distance calculations, we examined the rate at which the distance between medians was as large or larger in a random permutation as it was in the original data. This rate is simply the rate at which such a result would be expected under the null hypothesis that treatment had no effect, and thus can be directly interpreted as a p-value. Because nine such pairwise comparisons were made, it was prudent to use a Bonferroni correction on the significance threshold to limit the false-positive rate. Each pairwise comparison was below the corrected significance threshold of 0.05/9 = 0.0056. The distance between FtsZ+ (wild type) and FtsZ+A22+ was 170.225 with a p- value <0.0001 that a distance this large would occur due to chance. This result demonstrates that A22 treatment induces a significant shape alteration in otherwise wild- type cells (containing FtsZ). Likewise, the difference between the two non-A22-treated populations, FtsZ+ and FtsZ-, was 128.620, p=0.0001, indicating (as expected) that the FtsZ-depleted cells differ significantly from wild type cells. Interestingly, the distance between the two different A22-treated populations, FtsZ+A22+ and FtsZ-A22+, was 235.045, p<0.0001 (also highly significant), indicating that depletion of FtsZ alters the A22-induced shape. A FtsZ-depletion did not completely rescue the A22-induced shape defect, however, as there remains a significant distance between FtsZ-A22+ and FtsZ+ of 74.037, p=0.0007. We directly tested the hypothesis that our proposed ―rescued‖ population, FtsZ- A22+, is closer to wild type than is either FtsZ+A22+ or FtsZ- by a similar permutation approach. We defined a null hypothesis that FtsZ-A22+ condition did not make the cell shape any more like the wild type shape than did a second test condition (either FtsZ+A22+ or FtsZ-). We first computed the distance between the median shapes of the wild type and FtsZ-A22+ and between wild type and FtsZ+A22+ (or FtsZ-); then we calculated the difference between those two distances. To test whether the size of this difference was statistically significant, we randomly reshuffled the data from the FtsZ-

P a g e | 57

A22+ and FtsZ+A22+ (or FtsZ-) groups and again computed the difference in distances between each test group and the wild type group. The rate at which this difference is greater than or equal to the unshuffled difference (in the original dataset) is the p-value for rejecting the null hypothesis. Both tests concluded that FtsZ-A22+ treatment made the cells more like wild type than either the FtsZ+A22+ or FtsZ- treatment (p < 0.001 and p = 0.0294, respectively). Again, since these distances were calculated in the 3-dimensional PCA space, the statistical differences we see include all three major modes of shape variation (encompassing >95% of the total variance). In order to show how each mode of variation contributes to the overall differences in shape, we show the distributions for each class along each axis separately in Fig 4 (mode 1) and Supplemental Fig S5 (modes 2 and 3). Since the primary mode of variation accounts for >91% of the variation within these three modes, it contributes most to the median shape. Additionally, this mode appears to represent the qualitative descriptions that have been made of FtsZ-depleted and A22- treated cells. The second and third modes, however, describe more subtle variations in shape. Both of these modes appear to mainly address curvature (Fig S5). Interestingly, both the A22-treated classes appear straighter than either non-A22-treated group: whereas in the primary mode of variation the median of the FtsZ-A22+ group falls closer to those of the non-A22-treated groups than that of FtsZ+A22+, in the second and third modes both A22-treated groups lie together and separated from the non-A22 treated groups (Fig S5). To examine how each shape is described by a combination of these parameters we provide 2-dimensional and 3-dimensional ―shape-map‖ representations. For these ―shape maps‖, we used the NCSA Partiview viewer [191] to plot the actual shape of the bacterial cell at the Euclidian point corresponding to its values for each of the PCA shape parameters. The 2D shape map is presented Supplemental Fig S6, in which the (x, y) point for each cell corresponds to its parameter values in the first and second principal modes, respectively. From this plot it is evident that along the X-axis (the first principal mode, roughly describing length and width), the FtsZ-A22+ treated population looks more wild type than the FtsZ+A22+ populations; while along the Y-axis (the second principal mode, roughly describing curvature and width), both the A22-treated populations can be distinguished from the non-A22-treated populations. The 3-dimensional shape map is

58 | P a g e

CHAPTER 2 presented in Supplemental movies 4 and 5. As in Fig S6, we display the actual bacterial shapes in space, but for the 3D representation we plot each shape at the (x, y, z) point corresponding to its parameter values in each of the three principal modes. We have also made our dataset available on our website in a format that allows for interactive visualization with Partiview (http://cmgm.stanford.edu/theriot/).

P a g e | 59

SUPPLEMENTAL FIGURES

SUPPLEMENTAL FIGURE 1

SUPPLEMENTAL FIGURE 2

60 | P a g e

CHAPTER 2

SUPPLEMENTAL FIGURE 3

P a g e | 61

SUPPLEMENTAL FIGURE 4

62 | P a g e

P a g e | 63

64 | P a g e

Chapter 3 Modifying morphology with mutations in mreB

Natalie Dye, Zachary Pincus, Isabelle Fisher, Lucy Shapiro, and Julie Theriot

P a g e | 65

ABSTRACT

The mechanism by which bacteria coordinate the spatial pattern of synthesis and degradation of their peptidoglycan cell wall to maintain their cell shape during growth and division is not well understood. In the rod-like bacteria E.coli, B.subtilis, and C.crescentus, this process requires the gene mreB, which encodes a member of the actin superfamily. To investigate the role of MreB in cell shape maintenance, we used the drug A22 to select for spontaneous mutations in mreB in Caulobacter crescentus. We obtained 36 mutant strains, all of which contain single lesions in the mreB gene and repeatedly grow in the presence and absence of A22. We identified mutants with cells that appeared to be longer, straighter, wider, thinner, or more variable than wild type. By quantitatively measuring the shapes of the cells grown in the presence and absence of A22, we show that length, width, curvature, and response to A22 can be partially uncoupled. In addition, a subset of mutants was found to have a unique and dramatic variable cell width phenotype, with relatively wide cell bodies and long, thin extensions of the cell poles. Timelapse microscopy showed that these cells appear to actively elongate at or near the pole and that extensions can grow from the new pole immediately after division. We generated fluorescent fusion proteins to a subset of our collection of A22-resistant MreB mutants. The subcellular localization patterns of these mutants could be described as peaked at midcell (similar to wild type), broadly distributed about midcell and away from the poles, uniformly distributed along the cell length, or peaked at the poles. We found that the polar localization of MreB is highly correlated with the development of relatively pointed poles and variable cell widths. Also, we found that the ability of the cells to regulate the localization of MreB in a cell cycle dependent manner is anti-correlated with cell size: cells that transition MreB from the cell sides to midcell are smaller than those that maintain MreB along the sides or at the poles. These mutant strains provide novel insight into how MreB’s protein structure, cellular dynamics, and activity contribute to its function in bacterial cell shape.

66 | P a g e

CHAPTER 3

INTRODUCTION

The bacterial actin homolog MreB is required for the maintenance of a rod-like shape in many organisms [204]. We have shown that the depletion or depolymerization of MreB in Caulobacter results in the mislocalization of Pbp2, an elongation-specific peptidoglycan transpeptidase, from the cylindrical sides to the division plane (Chapter 2, [142, 143]). Others have shown that MreB interacts with many of the cytosolic and membrane bound enzymes that synthesize peptidoglycan subunits and that the insertion of new peptidoglycan subunits along the cylindrical sides of the cell is dependent on MreB [43, 121, 122, 141, 143, 205-207]. It has also been shown that the localization of MreB is cell-cycle regulated in Caulobacter [39, 117]. Early in the cell cycle, MreB localizes to a helical-like pattern throughout the cell. In the middle of the cell cycle, MreB colocalizes with FtsZ and other divisome components. Late in the cell cycle, MreB is released from the division plane and again fills the cell in a helical like pattern. Taken together, these data suggest that MreB is required for the spatial and temporal control of cell wall growth. The mechanism by which MreB contributes to this process is unclear. It is known that the expression level of MreB is important. Caulobacter cells overexpressing MreB decline in viability, mislocalize polar markers and chromosome loci, and exhibit aberrant MreB localization patterns [117]. It is also known that MreB must exist at least in large part in polymer form, as treatment with the destabilizing drug A22 phenocopies the depletion of MreB [40]. It is not known, however, how the structure, size, or dynamic behavior of these polymers contributes to function. For example, what would be the consequence of over-stabilizing MreB polymers? The role of the nucleotide cycle of MreB in vivo is also largely unknown. In this work, we took a genetic approach to address these outstanding questions. We took advantage of the fact that spontaneous mutations in mreB can be easily isolated in Caulobacter by selecting for resistance to the drug A22, which binds to MreB in vitro with micromolar affinity [40, 208]. While it has been previously noted that the resulting mutant strains can have varying morphological phenotypes, no systematic analysis of cell shape in these mutants has been performed. We reasoned that different mutations in mreB

P a g e | 67

could differentially affect protein activity and that by studying the cell shape phenotype of these strains, we could gain insight into the function of MreB in the maintenance of cell shape. This work was inspired by a study that was done many years ago in which mutations in the gene encoding β–tubulin were isolated in yeast by selecting for resistance to low doses of the destabilizing drug, Benomyl [209]. These authors found that the growth of many of the mutant strains they isolated was still affected by the presence of the drug. In fact, two of their mutants were actually found to grow better in higher concentration of drug or in low temperature (which destabilizes microtubules). From these data the authors proposed that their mutant strains had microtubules that varied in stability and that the stability of these filaments could be tuned with low concentrations of Benomyl or low temperature. In the years following, it has been shown that indeed these mutants form polymers with altered dynamics in vivo and that the presence of Benomyl can further alter microtubule dynamics [210]. Therefore, we thought it possible that A22-resistant Caulobacter could have MreB polymers with varying stabilities and dynamic behaviors in the cell. We isolated 36 unique mutations in 26 amino acids of MreB. By quantitatively measuring the shapes of cells grown in the presence and absence of A22, we show that length, width, curvature, and sensitivity to A22 can be partially uncoupled in this collection of mutants. In addition, we describe a unique and dramatic variable width phenotype, in which certain cells appear to elongate at or near the poles to develop long thin polar extensions from a relatively wide cell body. By quantitatively measuring the distribution of Venus-MreB in a subset of our A22-resistant mutants, we show that the cell cycle regulated dynamic localization of MreB is altered to varying degrees. We show that some of these mutants localize MreB to the poles and that this polar localization is associated with the development of pointed, rather than rounded, cell poles and a largely variable cell width. In two of these mutants, the polar localization and pointed-pole phenotype can be rescued with the addition of A22, implying that these mutants are hyperstabilized. We also show that the ability of cells to regulate the distribution of MreB over the cell cycle anti-correlates with cell size. The results presented in this work demonstrate that mutations in mreB can be used to study the dynamic behavior of MreB and the resulting consequences for cell shape.

68 | P a g e

CHAPTER 3

RESULTS

Isolating A22-resistant Caulobacter To generate a collection of mreB mutant Caulobacter, we grew wild type Caulobacter at 30C on rich media plates containing a modest concentration of A22 (2.5µg/ml) and selected for spontaneous resistance. We sequenced the mreB gene in 87 independently-isolated A22-resistant strains. All strains were found to have point mutations in mreB (Supplementary Table 1). Several strains with the T167A mutation were isolated; this mutation was also the most common A22-resistant mutant in the screen performed by Gitai et al [40]. We also obtained several isolates of D192G (9), N21S (8) C110S (7), and D16G (6). Nevertheless, this selection was not saturated, as we isolated several mutations only once. Many of these singletons have never been previously identified (Supplemental Table 1). Of the 87 strains, we kept only those that possessed a single mutation in mreB and could consistently and repeatedly grow well both in the presence and absence of 2.5 µg /ml A22. We also added a previously isolated strain described in Ref [211]. Our final collection contains 35 unique mutations in 25 amino acids. To observe cell shape, we grew all of the strains simultaneously in a 96-well plate both with and without A22. Once the cells were steadily growing in log phase, we calculated the doubling time for each strain and rapidly imaged all strains (see Methods). In this experiment, we also included wild type Caulobacter and selected ―sister strains‖: two independent isolates of the same mutation (denoted in figures with lowercase subscript a, b, or c). All cells were imaged at approximately the same cell density (OD=0.1-0.25, Supplementary Table 2). We collected images of approximately 300 cells per strain, with and without drug (Supplementary Table 2). The experiment was then repeated to control for day-to-day variation. In total, this dataset contains images of >55,000 cells in 48 strains (35 unique mutations + 12 sisters + WT).

P a g e | 69

FIGURE 1

70 | P a g e

CHAPTER 3

FIGURE 2

P a g e | 71

P a g e | 72

CHAPTER 3

Images of the isolated A22-resistant strains grown in the absence or presence of A22 are shown in Figure 1 and 2, respectively. In these images, subtle variations in can shape can be observed. The most commonly occurring mutation, T167A, appears straight or sigmoidal, consistent with the previous report [40]. We also identified other mutations with this phenotype (marked with yellow type). In addition, we identified many mutant strains that appeared to have a morphology that is relatively similar to wild type (marked with red type), much wider than wild type (dark blue type), or more variable than wild type (marked with white type). Note that different mutations to the same amino acid could have varying effects on cell morphology (for example, compare N21D, N21S, and N21Y). Doubling time does not vary considerably in our mutants (Figure 3). This result is perhaps not surprising, given that we specifically selected only those mutants that could grow well on plates containing A22. In A22, nearly all of the strains have a shorter doubling time (with less variability), indicating that the drug may actually be slightly beneficial for the growth of these strains. Even so, there were a few strains that were observed to have a more variable doubling time (particularly in A22, marked with asterisks in Figure 3). These strains were also observed to have a more variable and dramatic cell morphology.

Measuring the shape phenotype of A22-resistant Caulobacter To quantitatively measure cell shape in our collection of A22-resistant strains, we extracted cell outlines from the phase images and used Principal Components Analysis (PCA) to determine the primary modes of variation in all the cells of our dataset (see Methods, [212]). The first five modes account for >98% of the total variation and appear to be biologically meaningful. These modes are illustrated in Figure 4. The first mode is essentially cell length. It alone accounts for >92% of the total variation, which is perhaps expected, given that cell length will double over the course of the cell cycle and our images were taken from a mixed population of cells in different stages of the cell cycle. Shape modes 2 and 3 correspond to curvature and width, respectively, whereas Shape modes 4 and 5 appear to capture asymmetry within a single cell. High deviations from the mean in Shape Mode 4 indicate asymmetry in curvature and width of the cell near the

P a g e | 73

pole (with one pole wider than the other). Shape Mode 5 appears to reflect variations in width along the long axis of the cell, separating cells with pointed versus rounded poles.

FIGURE 4

74 | P a g e

CHAPTER 3

By definition, these PCA shape modes reflect orthogonal deviations from the mean shape of all cells (regardless of their genotype or cell cycle stage). It is possible, however, that these axes could be correlated to one another when comparing the mean values for each strain in each shape mode. Specifically, of the mutants that have already been studied (T167A [40], G165D [213], and Q26P [211]), two distinct morphologies have been noted: long, thin, and straight (relative to wild type) or normal length, width and curvature. If this trend were to hold true in our dataset—long cells are always thinner and straighter and normal length cells always have normal width and curvature—we would expect a strong correlation between Shape Modes 1, 2, and 3. We did not find this to be the case, however. To investigate this issue, we calculated the mean value in each shape mode for each strain and calculated correlation coefficients between modes (Supplemental Figure 1). We found that the absolute value of Shape Mode 4 (asymmetry) correlates strongly with Shape Mode 1 (length) and to a slightly lesser extent to Shape Mode 5 (polar width). Thus, cells that are very asymmetric tend to be longer, with more pointed poles. Shape Modes 2 (curvature) and 3 (width), however, are only weakly anti- correlated with Shape Mode 1 (length). Thus, cells that are straighter than wild type are not necessarily longer than wild type, and cells that are thinner than wild type are not necessarily longer or straighter than wild type. These data indicate that within our mutant collection, we have partially uncoupled the processes leading to the determination of proper length, curvature, and width. This result suggests that distinct processes underlie the determination of these different shape parameters, and that these MreB mutants are defective in some but not all of these pathways.

Clustering A22-resistant Caulobacter strains by cell shape The definition of the PCA shape modes allows us to measure the shape of every cell in our dataset with a small number of meaningful parameters. With this quantitative metric for the cell shape phenotype, we aimed to classify our mutant strains according to their shapes in the presence and absence of A22. In the hypothetical ―shape space‖ defined by the first four PCA Shape Modes, we calculated the distance between each cell to every other cell and then performed hierarchical clustering on all strains grown in the absence or presence of A22 (see Methods, Figure 5). In this analysis, we included cells

P a g e | 75

from two different cultures of the same strain grown in parallel (denoted with ―1‖ and ―2‖) as a control, and indeed these cells tended to group together, indicating that the clustering method is reliable. Also, cells of sister strains (denoted ―a‖ and ―b‖) clustered near each other. These strains were included to assess the stability of these unmarked mutations. The clustering diagram of cells grown in the absence of A22 is very different from that of cells grown in the presence of drug (compare Figure 5A and B). Interestingly, in the presence of drug, we obtained fewer and tighter clusters than in the absence of drug, indicating that cell shape in this collection of mutant strains is more variable in the absence of drug. From these diagrams, it is apparent that not all strains have the same response to A22. For example, T167A (yellow in Figure 5) clusters with I266S and V170A in the absence of A22 (Figure 5A). In the presence of A22, I266S clusters with E119G and S181L and far from T167A, whereas V170A remains closely clustered to T167A (Figure 5B). In another example, E213G and A325P (blue in Figure 5) are not clustered tightly with any other strain in the absence of A22 (Figure 5A), but in the presence of A22, these strains become tightly clustered with many other strains (red and yellow groups, respectively, in Figure 5B). This analysis of morphology indicates that A22 can still influence morphology in some of these mutant strains. Note that these strains are all ―A22-resistant‖ in that they can survive and grow well in the presence of A22. Nevertheless, the nature of the resistance in each individual strain may vary. To confer A22 resistance, the mutations in MreB must either considerably lower the affinity for A22 or alter the consequence of A22 binding without altering affinity. For example, a mutation that stabilizes MreB polymers could make it more resistant to the destabilizing drug A22, in the same way that certain β-tubulin mutations are thought to both alter microtubule dynamics and confer resistance to the destabilizing drug, Benomyl (See Introduction). We quantified the extent to which each strain responds to A22 by measuring the distance between distributions of cell shapes in the same strain grown with or without A22. To determine whether A22 makes the cells more or less similar to the wild type shape (in the absence of drug), we measured the distance between each strain grown with or without drug to the wild type distribution in the absence of drug. Most cells became closer to wild type in the presence of drug,

76 | P a g e

CHAPTER 3 again demonstrating that A22 can actually have a beneficial effect on these strains. We then calculated the difference between the distance to wild type in the presence of drug and the distance to wild type in the absence of drug. These results are presented in Figure 6A. To combine the shape phenotype with the response to A22, we next clustered the strains according to the sum of the distances in the absence and presence of drug (Figure 6B). The mean Shape Mode values for each strain grown in the presence and absence of drug are presented in Supplemental Figure 2, organized according to the clusters identified in Figure 6B (sum of the shape phenotype and A22 response). This figure summarizes cell shape phenotype of each group of A22-resistant mutants, and demonstrates that indeed length, width, and curvature, as well as the response to A22 can be uncoupled from one another. For example, in the absence of A22, cells in clusters 1, 2, and 3 have similar lengths and widths; however, cells in cluster 1 have much more pointed poles in the absence of A22 than those in clusters 2, and cells in cluster 3 are more curved than cells in clusters 1 and 2.

P a g e | 77

FIGURE 5

78 | P a g e

FIGURE 6

P a g e | 79

A variable width phenotype As mentioned above, A22-resistance has previously been shown to be associated with either a straighter/thinner cell shape, i.e. T167A, or a relatively wild type cell shape, i.e. Q26P. With our analyses, we have extended this conclusion and shown that subtle but quantitative differences exist between strains in these two broad qualitative phenotypic classes. In addition, we have identified novel morphological phenotypes. Specifically, we also found mutants that are wider than wild type (i.e. E119G and G329C), as well as mutants that have pointed poles and a more variable width. Given that the shape of wild type cells is remarkably constant within a population, we were particularly interested in the strains that seemed to have a more variable cell shape, including A325P and to a lesser extent N21D and E213G (Figure 7A). In these strains, it is possible to see large deviations in cell width across the population and even within single cells. Many cells have long, thin extensions near the poles and wide cell centers. In addition to the strains listed above, we also isolated a mutant strain with an extreme version of this variability phenotype (Figure 7A). The nature of the mutation in this strain is also unique, as it is small duplication: R185 and V186 are repeated, so that the sequence VRV at amino acids 184-186 becomes VRVRV. While we had isolated this mutation in the wild type Caulobacter strain in our initial screen, the phenotype was rapidly suppressed and it was removed from our collection. Later, upon sequencing spontaneous A22-resistant mutants of a GFP-MreB merodiploid strain, which contains the endogenous, unlabeled mreB, as well as a Xylose-inducible GFP-MreB, we isolated the same mutation with the same dramatic morphological phenotype (Figure 7A). This mutation is present only in the endogenous unlabeled copy of mreB; the copy of mreB encoded at the xylX locus has a wild type sequence (data not shown). In this strain, the phenotype is relatively stable, with or without the addition of xylose to induce the wild type GFP-MreB (data not shown). It is likely that low levels of the wild type gfp-mreB is expressed and prevents this mutation from being quickly lost from the population. Still, if grown for long periods of time without A22, this strain will also become suppressed. We do not yet know the nature of this suppression, but it will be an interesting future direction.

80 | P a g e

CHAPTER 3

FIGURE 7

P a g e | 81

Long thin projections have also been observed in a strain of Caulobacter containing a mutation in ftsZ, the tubulin homolog required for division [41, 105]. This mutation causes a severe late block in division that leaves two cells attached by a thin bridge resulting from a failed division event. We wondered whether the thin regions of these mreB mutant cells could also be derived from failed division events. To answer this question, we used timelapse microscopy to image the R185V186 duplication strain over multiple rounds of growth and division (Figure 7B). In these movies, it is apparent that these polar extensions are actively elongated throughout the cell cycle and do not derive from failed division events. In addition, we can see that extremely thin cells can be pinched off from the ends of these extensions and go on to grow and divide again. In Figure 7C, we present a summary of our observations of this strain. We observed several thin cells elongate and divide relatively symmetrically to produce two thin daughter cells (Figure 7Bi). We also observed cells with an asymmetric cell width, with one end slightly thinner than the other (Figure 7Bii). In this case, division produced one thin cell and one slightly wider cell. In the next round of division, this wide cell either grows and divides symmetrically into two wide cells (often with polar extensions), or asymmetrically to produce one thin cell and one wide cell (Figure 7Bii). At some point, they become too wide to divide. These cells appear to be a terminal state that can only elongate at the ends and eventually lyse (Figure 7Biii).

82 | P a g e

CHAPTER 3

Determining the cellular localization of MreB mutants To begin to understand how these mutations in mreB generate the different observed cell shape phenotypes, we characterized the subcellular localization and dynamic behavior of the mutant proteins in cells. We selected representative strains from each phenotypic class for this experiment. To report on the localization of MreB, we constructed N-terminal fusions of the mutant proteins to Venus, a YFP-derivative, and inserted them into the chromosome of the mutant strains at the xylX site. These strains therefore contain two copies of mreB: one inducible labeled version and one endogenously expressed unlabeled version; both copies containing identical mutations. We induced the expression of these reporter constructs with a low amount of xylose for one hour and imaged mixed populations. Images of selected cells are shown in Figure 8A. It has been previously shown that the A22-resistance mutations T167A, Q26P, and G165D abolish the cell cycle regulated transition of MreB from a lengthwise helix along the cell sides to a midcell ring at the division plane [40, 211, 213]. The distribution of GFP-fusions to these MreB mutants has been reported to be dispersed in a lengthwise helical-like pattern throughout the cell cycle. We could observe a similar pattern for the mutants V324A, D162G, V170A, D192G, and D189G (Figure 8A, middle). In contrast, the mutants G165A, N21S and C110S appeared to be able to localize MreB to the midcell region in at least some of the cells (Figure 8A, top). Surprisingly, the mutants E213G, D16G, N21D, and A325P actually appeared to localize a significant proportion of MreB to the poles of the cell (Figure 8A, bottom), which is a novel pattern that has not yet been observed for other MreB mutants or wild type MreB.

P a g e | 83

P a g e | 84

CHAPTER 3

P a g e | 85

Analyzing MreB localization in subcellular regions From these data, we reasoned that the distribution of MreB into different subcellular regions—the cylindrical side walls, midcell, and the poles—could be a distinguishing characteristic of these strains and have important consequences for the cell shape phenotype. Accordingly, we next developed metrics that would allow us to quantitatively assess the distribution of MreB along the long axis of the cell. For each cell, we defined a centerline with 50 evenly spaced points to generate a relative coordinate system. We then calculated the average intensity of MreB at each point along this centerline to generate a one-dimensional profile (see Methods). In Figure 8B, we present the average intensity profile of Venus-MreB along the centerline of all cells in a given strain, regardless of cell cycle stage. In this figure, we see that the one-dimensional MreB profiles from the A22-resistant strains appear to lie on a continuum from peaked at midcell, to flat, to polar. Thus, even with this relatively simple metric, we can begin to separate the strains according to their localization phenotype. To determine meaningful variables for describing this intensity profile in each cell, we again used PCA (see Methods), this time to assess the Principal modes of variation in the distribution of MreB along the centerline (rather than in cell shape). We find that the primary mode of variation in the dataset containing all cell intensity profiles appears to capture the difference between peaked, flat, and polar distributions (Figure 9). This first mode accounts for 35% of the total variation. The second primary mode appears to reflect asymmetry in the distribution of MreB from one end of the cell to the other and captures 16% of the total variation. The remaining PCA modes of fluorescence variation capture finer details in the profiles. For simplicity, we used only the first two modes for further analysis. These modes will hereafter be referred to as ―Fluor Mode 1‖ and ―Fluor Mode 2‖ to distinguish them from the PCA Shape Modes described in previous sections. Importantly, if we generate PCA axes using only the mean profiles for each strain, we obtain similar modes of variation (data not shown). In this case, however, almost 90% of the total variation can be explained by the first mode, which is identical to the Fluor Mode 1 generated from all cells (Figure 9), indicating that this mode is capable of distinguishing important variation between strains in the subcellular distribution of MreB. Representative cells from each

86 | P a g e

CHAPTER 3 localization class, their corresponding profiles, and their PCA mode values are shown in Supplemental Figure 3 to illustrate how these PCA axes describe the MreB profiles.

FIGURE 9

P a g e | 87

In our analysis, we used the absolute value of Fluor Mode 2 because we viewed only the magnitude of asymmetry, and not the orientation, to be meaningful. Nevertheless, we acknowledge that there are likely to be meaningful asymmetries in the localization pattern. For example, wild type Caulobacter divide asymmetrically to produce two slightly different sized daughter cells. Thus, we expect the division plane to always be further away from the stalked pole than from the swarmer pole and that peaks of MreB in the midcell region would be closer to one pole than the other. Additionally, we often observed asymmetric localization of MreB to one pole but not the other in the polarly localized MreB mutant strains. Unfortunately, in the phase images, the two poles are morphologically similar (stalks are not always visible), so aligning the cells by shape is not sufficient to distinguish one pole from the other. Therefore, at this time, we lack an unbiased way of systematically orienting the MreB profiles towards one pole or the other. In the future, we hope to overcome this challenge by introducing a fluorescent marker of polar identity that is compatible with simultaneous imaging of Venus-MreB. Several such markers are known in Caulobacter. In addition to the PCA Fluor Modes, we also directly measured the intensity of MreB in defined regions of the cell: the poles (points 1-5 and 45-50), the sides (6-15 and 35 to 45), and the center (16-35). We then calculated ratio of intensities in the center to that at the sides (C:S ratio) to estimate the size of the midcell peak, as well as the ratio of intensities at the poles to that at the sides (P:S ratio) to estimate the extent of polar localization.

Distinguishing A22-resistant mutants by subcellular localization patterns To further evaluate the localization phenotypes in our A22-resistant strains, we plotted the Fluor Mode 1 values against the absolute value of the Fluor Mode 2 values for each cell in each strain. These plots (Figure 8C), are consistent with the mean profiles presented Figure 8B and demonstrate the variability in this profile across all cells in a given strain. In particular, it is notable that those strains in the middle of the continuum, displaying a flat or very broadly peaked profile—D162G, V170A, and D192G—have a variable and asymmetric localization pattern. Given that the localization of wild type MreB is cell-cycle regulated, we reasoned that some of the variability in these Fluor

88 | P a g e

CHAPTER 3

Mode values within a strain (in particular Fluor Mode 1) could be reflecting differences in cell cycle stage. Accordingly, we also plotted Fluor Mode 1 as a function of cell length to roughly sort these cells by cell cycle stage (Figure 8D). In wild type cells, Fluor Mode 1 is low in very short cells. Interestingly, the Fluor Mode 1 values of very short wild type cells can be negative, corresponding to a slightly polar distribution, indicating that perhaps even wild type MreB can localize to the poles in some cells (particularly at the beginning of the cell cycle). Fluor Mode 1 peaks in the middle of the length distribution (near the middle of the cell cycle) and is low in the longest cells (late in the cell cycle). A similar pattern can be observed by plotting the C:S ratio as a function of cell length (Figure 8E). In the A22-resistant strains, the correlation between Fluor Mode 1 (or C:S ratio) and cell length appears to be less tight, indicating that the cell cycle regulation of MreB localization is altered to varying degrees in these A22-resistant strains. Fluor Mode 1 peaks somewhat in G165A, N21S, C110S, and V324A and hardly at all in D162G, V170A, D192G, and D189G. In the polar MreB mutants, there does appear to be a slight correlation between Fluor Mode 1 and cell length, but one that is opposite that of wild type. Fluor Mode 1 is the lowest in the middle of the length distribution, indicating that there could be some sort of cell cycle regulation in these mutants but one in which MreB is depleted from midcell and accumulated at one or both poles. This correlation can also be observed in the plot of P:S ratio as a function of length (Figure 8F).

P a g e | 89

Characterizing the cell cycle regulation of the mutant localization patterns Cell length is only an approximate measure of cell cycle stage. To directly measure the localization of these mutant MreB proteins over time, we performed timelapse imaging in synchronized populations of selected strains. For each cell, we calculated the cell length, the intensity of MreB along the centerline, and the PCA Fluor mode values describing this profile. One representative cell for each strain and the corresponding analyses of the fluorescence distributions are shown in Figures 10-12. For each strain, we then calculated the averages of these parameters across all cells at a given timepoint. Mean profiles and Fluor Modes for each strain are presented in Figure 13. As expected from previous data, the wild type protein was found to localize to a tight band near midcell, though offset slightly, at approximately 30 min after the start of imaging. It remains in a tight band, at a constant relative position in the cell until just before division. Fluor Mode 1 is slightly negative at the beginning of the experiment, peaks around 60 min and then drops again before division. Note that Fluor Mode 2 also peaks around 60 min, reflecting the asymmetric distribution of the MreB peak along the centerline. G165A has a similar pattern of localization through the cell cycle, although it appears that the entire cell cycle is slightly delayed. The average time to division in this strain was found to be slightly longer (see Figure 13C); note, however, that the time resolution is rather low in this experiment, given that we only captured images every 15 min. Fluor mode 1 in G165A, as in WT, is negative at the beginning of the cell cycle, corresponding to polar MreB localization. Fluor mode 1 gradually increases as the cell slowly accumulates a peak of MreB near midcell. This midcell peak never quite becomes as tight or persists for as long as those in wild type cells. Nonetheless, it appears that this strain is able to at least partially regulate the localization of MreB according to the cell cycle.

90 | P a g e

CHAPTER 3

FIGURE 10

P a g e | 91

FIGURE 11

92 | P a g e

CHAPTER 3

In contrast, the localizations of V324A and D189G do not considerably vary during the cell cycle (Figure 11). V324A and D189G are subtly, but consistently, different. While neither is recruited to a tight midcell band, V324A avoids the poles, whereas D189G can localize along the entire cell length. This difference is apparent both in the plots of intensity along the centerline over time Figure 11, middle) and in the plots of the PCA Fluor Modes over time (Figure 11, right). In V324A, Fluor Mode 1 remains high throughout the cell cycle, and never drops below 0, indicating that the intensity at the pole is always low. Flour Modes 1 and 2 vary considerably over time in D189G, due to the transient appearance of bright puncta in various regions of the cell. A325P is also not strongly cell cycle regulated: it remains at the pole throughout (Figure 12). Note that we found examples of both unipolar and bipolar MreB. The asymmetry, precise timing, and molecular requirements of this polar accumulation of MreB will be the subject of future work (See Discussion). Note that in these cells, the accumulation of A325P-MreB is associated with the thinnest parts of the cells (Figure 12A). Upon division, the new pole seems to quickly acquire a new polar spot of MreB (t=150min, asterisk). This polar spot then remains associated with the spot throughout the next cell cycle, as the pole gradually grows into a thin polar extension (arrow).

P a g e | 93

FIGURE 12

94 | P a g e

CHAPTER 3

In Figure 13 A-B, we present the average intensity profiles and Fluor Mode 1 values of these MreB mutants over time, generated by averaging the data from a population of cells at a given timepoint (N=42-104). These data are consistent with the patterns observed in the representative individual cells (Figures 10-12). We also determined from these timelapses that the rate of elongation is not significantly different across these five strains (Figure 13C). The rate of division in the mutants was also found to be similar to wild type, although slightly longer in G165A and D189G. These data are consistent with the observation that the bulk doubling time of the cultures, as measured by change in optical density, does not change considerably (See Figure 3). These results indicate that the difference between these strains is the way in which they grow and change shape, rather than the speed of growth. Lastly, we quantified the extent to which MreB localization changes over the course of the cell cycle. For each strain we subtracted the Mean Fluor Mode 1 value at time 0 from the Mean Fluor Mode 1 value at the midpoint of the cell cycle. This difference will hereafter be referred to as the index of cell cycle regulation (CCR-Index). These data are presented in Figure 13E. This metric indicates that G165A-MreB is regulated as well as wild type. Note that this result is slightly misleading, as the timing of accumulation of MreB at the midplane is delayed in this mutant. Nonetheless, this metric does distinguish G165A from V324A and D189G quite well. The MreB distribution changes only slightly in these strains, much less than in G165A. A325P also changes its distribution somewhat over the cell cycle. As predicted from the static data of the mixed population of cells, Fluor Mode 1 actually decreases in the middle of the cell cycle in this strain, and therefore the CCR-Index is negative. This decrease is likely due to the transition from monopolar to bipolar. The CCR-Index therefore provides an estimate of both the extent to which MreB changes over the cell cycle, as well as the direction of that change (towards or away from midcell).

P a g e | 95

FIGURE 13

96 | P a g e

CHAPTER 3

Correlating MreB localization with cell shape We have shown that A22-resistance mutations in mreB can have distinct cell shape phenotypes. In our collection of mutant strains, length, width, curvature, asymmetry, and polar width can be partially uncoupled. We have also shown that the subcellular localization of MreB and its dynamic distribution over the course of the cell cycle are altered to varying degrees in these mutant strains. The next obvious challenge was to determine how these different measures of cell shape are correlated with the subcellular localization and dynamic cell-cycle regulated redistribution of MreB. Toward this end, we chose to represent the fluorescence localization pattern for each strain in two separate, albeit related, parameters. The first is the mean Fluor Mode 1 value for the mixed population. This value distinguishes between distributions that are flat, peaked and polar, but partially ignores any cell-cycle-dependent changes in this distribution. The second parameter is an estimate of the CCR-Index. Rather than performing timelapse experiments with all of our strains, we considered cells within the 10th percentile of total cell lengths in that strain to be in the earliest stage of the cell cycle, and those in the 45th-55th percentile of cell lengths to be in the middle stage. We then calculated the difference between the means of those two sub-populations. This value should distinguish strains by the extent and direction to which MreB localization changes as a function of cell length.

P a g e | 97

FIGURE 14

98 | P a g e

CHAPTER 3

To determine how cell shape correlates with MreB localization, we calculated the correlation coefficients for the mean value for each strain in cell area, length, width, Shape Mode 2 (curvature), Shape Mode 4 (asymmetry), Shape Mode 5 (pole width), Fluor Mode 1, and CCR-Index. The data are presented in Figure 14. Consistent with our analysis of the entire population of A22-resistant mutants grown in the absence of A22 (Supplemental Figure 1), Shape Modes 4 and 5 are strongly correlated with each other, as well as with area and length in this subpopulation of mutants. Interestingly, we observed a strong anti-correlation between Fluor Mode 1 and Shape Mode 5 (Box in Figure 14A- B). Thus, cells with polar MreB develop pointed poles and have more variable cell widths. We also observed a strong anti-correlation between the extent and direction of cell cycle regulation (CCR-Index) and cell area (Figure 14A Box, C). Fluor Mode 1 and the CCR-Index are strongly correlated, indicating that strains with similar average MreB distributions change similarly over the course of the cell cycle. Average cell width and curvature were found to be completely uncorrelated with either metric of distribution of MreB. From these data, we conclude that the dynamic subcellular localization of MreB may influence cell size and the variability in cell width (particularly near the poles) but not curvature or average cell width. Cells with a slightly polar distribution of MreB direct even more MreB to the poles as the cell cycle progresses; these cells are larger, more asymmetric, and have pointed poles. In contrast, cells that at least partially accumulate MreB at midcell in the middle of the cell cycle are smaller, more symmetric, and have rounded poles. Average cell width and curvature do not significantly correlate with the subcellular localization and dynamic behavior of MreB. Because we can uncouple these phenotypes from the subcellular localization of MreB, we think it is plausible that distinct processes exist for the determination of proper width, length and curvature and that some of the A22-resistance mutations can affect binding sites for partner proteins in these processes. In order to understand the molecular mechanism by which the mutations in MreB contribute to the cell shape phenotype, we must investigate the interactions between MreB and its interacting partner proteins.

P a g e | 99

FIGURE 15

100 | P a g e

CHAPTER 3

Correlating Crescentin localization with Curvature In Caulobacter, curvature requires the intermediate-filament protein Crescentin (encoded by creS, [123]). MreB and Crescentin can be co-immunoprecipitated from cell lysates, indicating that the two proteins can directly associate with one another [213]. In addition, the treatment of Caulobacter with A22 has been shown to disrupt Crescentin localization, causing the Crescentin to detach from the cell surface and recoil into a loose helix [213]. In Figure 15, we present the localization of GFP-Crescentin in selected A22- resistant strains. As shown previously, GFP-Crescentin was found to follow the inner curvature of wild type cells. In the A22-resistant mutants, GFP-Crescentin localization is disrupted to varying degrees. We found that the localization of Crescentin correlates well with the average Shape Mode 2 values for each strain. D16G is nearly identical to wild type with respect to Crescentin localization and Shape Mode 2 (0.26 vs. 0.25). V324A has a slightly lower value for Shape Mode 2 (0.1). The localization of Crescentin in this strain is mostly normal, although some irregularities were noticed, including cells with Crescentin that was unevenly distributed throughout the cell (asterisks in Figure 15). In D189G, these irregularities became more common, with Crescentin often appearing linear or diagonal with respect to the long axis of the cell. In the strain E119G, which has a negative average Shape Mode 2 value (straighter than wild type), many of the cells have aberrant Crescentin localization. The localization of Crescentin in this mutant resembles that in A22-treated cells, appearing helical and ―detached‖ from the cell membrane. From these data, we conclude that the interaction of MreB and Crescentin is perturbed in some but not all of the A22-resistant strains and that the curvature phenotype can be correlated with proper localization of the intermediate filament homolog Crescentin. Note that although D189G and E119G both have negative Shape Mode 2 values, they differ very obviously in cell width. Likewise, D16G and V324A differ only slightly in their curvature but much more considerably in width and length. These data highlight, once again, that the process determining cell curvature is separate from that determining other parameters of cell shape.

P a g e | 101

FIGURE 16

Localizing the peptidoglycan transpeptidase Pbp2 in selected mutant strains In the previous chapter, we showed that MreB collaborates with MreC and Pbp2 in the determination of proper cell width. In this chapter, we have shown that MreB could be localized to the poles and that this localization pattern is associated with a variable cell width. As such, we were particularly interested to examine the localization of MreC and Pbp2 in these mutants. In Figure 16, we present the colocalization of MreB with Pbp2 in two of the polarly localized MreB mutant strains. We find that Pbp2 is not recruited to the poles and does not considerably colocalize with MreB (consistent with the results presented in Chapter 2). We obtained similar results with MreC (data not shown). Thus, even though it appears that these polar regions have a high concentration of MreB and appear to be actively elongating (see Figure 7B), they do not recruit either MreC and Pbp2.

102 | P a g e

CHAPTER 3

Changing the distribution of mutant MreB with A22 In Figures 1, 2, 5, and 6 we showed that the response to A22, as measured by morphology, varies by strain: while some strains maintain a constant shape in A22, others considerably change their morphology in the presence of drug. Presumably those strains that do not change shape in A22 have a mutation in mreB that prevents binding of the protein to A22. In the remaining strains, we reasoned that A22 alters either the activity or the dynamic subcellular localization of MreB (or both). To characterize the effect that A22 has on the localization of MreB, we grew our Venus-MreB merodiploid strains for an extended period of time in the presence of A22 and then induced the expression of the fluorescent variants for one hour prior to imaging. Representative images and measurements of the intensity of MreB along the centerline are shown in Figure 17. For technical reasons, D192G was omitted from this analysis and is not shown. In contrast to wild type, all of the mutant strains that were analyzed maintained their punctate localization at the membrane in the presence of A22. By comparing Figure 17 to Figure 8, we can see that the localization of MreB changes in some but not all of the strains in the presence of A22. For example, E213G and A325P are largely polar in the absence of A22 (Figure 8) but are much more uniform or slightly peaked at midcell in the presence of A22 (Figure 17). In contrast, the localization of D16G and N21D MreB mutants do not change as much in the presence of A22. Thus, even within the same localization class, the response to A22 can vary. The plot of Fluor Mode 1 versus Fluor Mode 2 (Figure 17C) indicates that the variance in the distribution of MreB can change in the presence of A22. In addition, the correlation between Fluor Mode 1 and cell length appears to be altered in varying ways in these strains (Figure 17D). Figure 18 summarizes our localization data for these strains in the presence and absence of A22. In this figure, we plot the mean Fluor Mode 1 value for each strain, as well as the estimated CCR Index, in the presence and absence of drug.

P a g e | 103

104 | P a g e

CHAPTER 3

P a g e | 105

To determine whether the A22-induced changes in MreB distributions that we observed could be correlated with the A22-induced changes in cell shape, we again calculated correlation coefficients between the PCA shape modes, Fluor Mode 1 and the CCR-Index, this time using only the values for cells grown in the presence of A22. As shown in Figure 19A, we find that in A22, Fluor Mode 1 and the CCR-Index are no longer significantly correlated with any of the measurements of cell shape. Given that we observed such a strong correlation between area, Shape modes 4 and 5, and both metrics describing the localization of MreB in the absence of A22, we conclude that the observed A22-induced changes in subcellular localization are insufficient to completely explain all of the observed differences in morphology in these strains. Instead, it seems likely that A22 can affect either localization or activity of these mutants. It is also possible that some of these mutants do not bind A22 at all, and therefore their shape and localization of MreB should remain unchanged. To take a closer look at the behavior of individual strains in their response to drug, we plotted Shape Mode 5 as a function of Fluor Mode 1 (Figure 19B) and cell area as a function of CCR-Index (Figure 19C) for each cell with and without A22. From the data presented in Figure 19B, it is clear that some of the strains— namely G165A, C110S, V324A, and D16G—do not significantly change either their pole width or their subcellular localization of MreB in A22. Note that these strains were shown in Figure 6A to have very similar cell shapes in the presence and absence of A22. From these data, we think it is likely that these mutant proteins do not bind A22. In contrast, the strains A325P, E213G, and D189G display large decreases in Shape Mode 5 and increases in Fluor Mode 1 in response to A22. Thus, these strains shift their localization of MreB towards midcell and away from the poles and develop rounder poles and more symmetric cell widths. This result is consistent with the correlation between Shape Mode 5 and Fluor Mode 1 in the absence of A22 (Figure 14). The strains N21D and V170A, however, have the opposite response to A22: their MreB distribution becomes more polar, and yet they still develop rounder poles (both Fluor Mode 1 and Shape Mode 5 decrease). D162G also becomes more polar and yet there is practically no change in Shape Mode 5. Therefore, A22 can decouple Shape Mode 5 from MreB localization.

106 | P a g e

CHAPTER 3

In Figure 19C, we show that cell area is also no longer correlated with the cell cycle regulation of MreB localization in cells grown in A22. We found that cells are systematically smaller in the presence of A22 than in its absence. Even those strains that changed shape the least in the presence of drug were found to be smaller in A22. The reason for this systematic drop in size is unclear (see Discussion).

FIGURE 18

P a g e | 107

FIGURE 19

108 | P a g e

CHAPTER 3

DISCUSSION

A22 resistant Caulobacter strains can have distinct cell shapes and MreB localization patterns In this work, we quantitatively compared the cell shape phenotype and the cell cycle regulated MreB localization pattern in a collection of Caulobacter strains bearing single point mutations in mreB. We determined that greater than 98% of the variation in cell shape can be described by five orthogonal PCA axes. These axes roughly correspond to length, width, curvature, cellular asymmetry, and width of the cell near the poles. Among strains, length is only moderately correlated with width and curvature. Asymmetry is strongly correlated with length, particularly in the absence of A22. The distribution of MreB was also found to vary in these mutant strains. The average localization of MreB in these mutants was found to be either peaked at midcell, broadly distributed in the middle of the cell and away from the poles, uniformly distributed, or polar. The polar localization of MreB is strongly correlated with pole width. The ability of MreB to change its localization over the course of the cell cycle is anti-correlated with cell size: those that properly redistribute MreB near midcell at the middle of the cell cycle are smaller than those that have a uniform distribution of MreB or accumulate more MreB at the poles as the cell cycle progresses. Average cell width and curvature do not significantly correlate with the average subcellular localization of MreB in the cell or the dynamic behavior of MreB over the cell cycle. We propose that the mutations that confer resistance to A22 can also perturb other aspects of MreB function and dynamics. These changes could result from the mutation altering the kinetics of the nucleotide cycle (binding, hydrolysis and release of ATP) or the binding of MreB to other interacting proteins in the cell. For simplicity, we assumed that the mutations we isolated do not alter the concentration of MreB in the cell, although this assumption has not yet been confirmed experimentally. Given that mreB is essential in Caulobacter and that all of these mutant strains grow with comparable doubling times and an approximately rod-like shape, it is reasonable to assume that none of these mutations completely destroy the ability of MreB to fold properly. Also, the mutants that were fluorescently labeled all localize to a punctate pattern on the membrane, which is

P a g e | 109

perhaps indicative of an assembled state. However, it is likely that some of the mutant proteins have subtle conformational changes in the structure that have consequences for protein activity.

110 | P a g e

CHAPTER 3

FIGURE 20

P a g e | 111

Mutant residues cluster around the ATP-binding site In Figure 20, we highlight some of the mutant residues we studied in this work on the solved T.maritima crystal structure of MreB1 [8]. Most of the mutations were found in residues that lie in the center of the folded protein, near the nucleotide binding pocket. Not all of the residues are close enough to be in contact with the nucleotide, however. Note that no mutations were identified in residues at the proposed longitudinal interface between subunits. We also did not isolate any mutations in the proposed binding site for RodZ (bottom-front surface of subunit II, [214]). In Figure 20A, we highlight the residues that change their cell shapes the least in the presence of A22: D16, G165, I123, V324, C110, and T167. These residues all cluster tightly together at the bottom of the nucleotide binding pocket, in closest proximity to the γ–phosphate of the ATP. Given that these residues, when mutated, could render the cells unable to respond to drug, we propose that these residues are directly important for the binding of MreB to A22. This site is consistent with recently published results of molecular docking experiments performed with A22 and T.maritima MreB [208]. Mutations in the remaining residues are likely to confer resistance to A22 indirectly. It has been recently suggested that A22 can be toxic to cells in ways that do not require MreB [215]. If it were true that the cell shape phenotype of these strains in the presence of A22 were due to the ability of A22 to inhibit other cellular targets, we would expect to see that all mutant strains behave similarly. Instead, we observed that cell shape and the distribution of MreB vary depending on the type of mutation in mreB. We did note, however, that cell size (as estimated by two-dimensional area) was systematically smaller in the presence than in the absence of A22. It is known that many bacterial species regulate cell size and shape according to growth phase, growth media, the presence of predators, and cell density (reviewed in [216] and unpublished observations), though the molecular mechanisms are not well understood. In our experiments, we made every effort to keep growth conditions (temperature and media), cell density, and growth phase constant. It is possible that the presence of A22 systematically altered the metabolism of all of these strains so that they all became smaller. In Figure 20B-E, we highlight residues that when mutated produce cells that have the largest deviations from wild type in Shape Modes 1, 2, 3, and 5 when grown in the

112 | P a g e

CHAPTER 3 absence of A22. Since the mean of the absolute value of Shape Mode 4 correlates strongly with mean cell length for each strain, we did not include this mode in this figure. The residues associated with cell length seem to be clustered loosely around the entire nucleotide-binding cleft. Residues associated with polar width (Shape Mode 5, Figure 20E) also did not seem to cluster in one particular part of the crystal structure. In contrast, the residues associated with curvature and width seemed to be more closely clustered at the ―back‖ and ―front‖ of the molecule, respectively (compare top and side views in Figure 20C and D). These sites could correspond to binding interfaces for Crescentin or other proteins involved in the determination of proper cell width. Alternatively these residues could be critical for determining the correct conformation of MreB for it to act in the processes determining curvature and width. Note that the two residues in Figure 20C at the front of the molecule, N21 and A325, are included on this diagram because the mutants N21D and A325P are quite straight. These mutants are also extremely polar, however, and other mutations to these same residues (N21S and A325T) produced cells with normal curvature.

The accumulation of MreB at the poles In this work, we identified four point mutations that cause MreB to localize to the poles. At this time, we do not know the mechanism of this polar recruitment. It is possible that this polar recruitment is due to protein aggregation. It is known that in Caulobacter some proteins are directed to the pole for degradation and that many proteases are localized at the pole [217, 218]. However, we favor an alternative hypothesis that these mutants have become trapped in one part of the wild type cell- cycle-dependent redistribution of MreB. We observed that wild type MreB could localize to the poles, especially very early in the cell cycle (See Figures 10A and 13B). In addition, it has been shown that MreB can co-immunoprecipitate with specific sequences of DNA near the origin of replication, and that A22 perturbs the segregation of this region in cells grown in minimal but not rich media [219]. This region of the chromosome is localized to the poles in Caulobacter and is anchored to the cell surface by the protein PopZ [220-222]. Therefore, at least a small population of MreB should exist at the pole in close proximity with this region of the chromosome, potentially to

P a g e | 113

participate in the chromosome segregation. The mechanism of chromosome segregation in Caulobacter is still largely unknown. Both MreB and ParA, a member of a family of ATPases that are thought to have cytoskeletal function (see Chapter 1), have been implicated in this process, and it is possible that partially redundant mechanisms exist. If the polar accumulation of MreB is dependent on its association with the chromosome, we would predict that this localization will follow that of ParB, which binds near the origin, and change in a popZ deletion strain, which can no longer properly anchor the origin-proximal region to the pole. Alternatively, it is possible that MreB accumulates at the poles to stimulate polar morphogenesis. It has been shown that new cell wall is inserted in a tight region at the base of the stalk, while the rest of that pole stays inert [211]. MreB has been implicated genetically in stalk biogenesis and growth, so it is possible that it is required at early points in the cell cycle to recruit factors that synthesize new cell wall at the stalk [117, 120]. We did not quantitatively characterize stalk growth in any of our mutants, although we did observe many cells with apparently normal stalks in the polar MreB mutant strains. In future work, we hope to look at the localization and segregation of the chromosome, as well as the localization of asymmetric polar markers, in our polar MreB mutants.

114 | P a g e

CHAPTER 3

FIGURE 21

P a g e | 115

The dynamic subcellular localization of MreB In addition to the strains with polar MreB, we isolated strains with MreB peaked at midcell, uniformly distributed, or broadly distributed about midcell and away from the poles. We propose that these localization patterns reflect stages of the wild type cell- cycle-regulated dynamic behavior of MreB. In Figure 21, we propose a preliminary model for the localization of MreB over the cell cycle, which extends that of previous work by including the polar localization phase. At the beginning of the cell cycle MreB is drawn to one pole, perhaps due to its association with the chromosome or other polarly localized proteins. At some point early in the cell cycle, a switch release MreB from the poles. MreB then becomes dispersed along the sidewalls and away from the poles, though remaining in a patchy helical pattern. A second switch happens after the assembly of FtsZ near midcell, which draws MreB to the division plane. It has been shown that MreB assembles at midcell in a FtsZ-dependent manner [39]; however, it is released from midcell relatively early, before many of the other late-recruiting components of the divisome even arrive at midcell (E.Goley, personal communication). Thus, a third switch releases MreB from the division plane. In this model, the polar localized A325P remains in the first stage, D189G remains in the transition between 1 and 2, V324A is trapped in the second stage and G165A is capable of going through all stages (though with slower kinetics than wild type).

A model for the role of nucleotide in regulating MreB dynamics Many members of the actin superfamily, even those that do not form polymers, are regulated by ATP-dependent conformational changes. Also, it is generally true that the dynamic behavior of many cytoskeletal proteins is regulated by nucleotide. Accordingly, it seems likely that the switches that occur during the cell cycle involve the binding and hydrolysis of ATP by MreB. To incorporate ATP hydrolysis into our model, we suggest that ATP-bound MreB accumulates at the poles early in the cell cycle and that the hydrolysis of ATP releases MreB from that site. At the next transition, we propose that ATP-bound MreB accumulates at the division plane, and that again ATP-hydrolysis releases MreB from that site. This model requires the presence of other cellular factors that are independently localized to distinct subcellular regions, namely the poles and the

116 | P a g e

CHAPTER 3 division plane. These proteins should be able to specifically bind ATP-bound MreB and would likely affect the kinetics of nucleotide exchange and/or hydrolysis. New evidence suggests that MreB binds to the transmembrane protein RodZ along the sides of the cell and that this interaction is required for proper cell shape [223-225]. In vitro experiments with T.maritima MreB suggest that this interaction is independent of the nucleotide bound to MreB [214]. The factors that directly control the polar and midcell accumulation of MreB are unknown, however, and there have not yet been any proteins identified that can bind to MreB in a nucleotide-dependent manner or regulate the kinetics of the nucleotide cycle of MreB. This model also predicts that the mutants we isolated should vary considerably in their ability to bind and hydrolyze ATP and/or bind to specific interacting partners in a nucleotide-dependent manner. Indeed many of the residues we isolated are in conserved motifs and are known to be important for the interaction between actin and nucleotide (Supplemental Figure 4, [8, 106, 226]). In a subset of sensitive mutants, A22 may be able to further alter this dynamic cycle. As an example, we would predict that A325P is unable to either bind or hydrolyze ATP or that it is in an otherwise hyper-stable polymeric state. The strain bearing this mutation exhibits a remarkable change in cell shape in A22. With analogy to the Benomyl-dependent β–tubulin mutants, it seems plausible that A22 is able to destabilize this hyperstable mutant and slightly rescue cell shape (with respect to variable cell width) by making the dynamics of this mutant closer to that of wild type. The mutant strain N21D, however, responds to A22 slightly differently, in that the localization is unchanged but the polar width decreases. In this mutant, A22 may affect the activity of MreB or its ability to bind other interacting proteins without affecting its localization. It is possible that nucleotide also regulates the organization of MreB puncta along the side. In this work, we only looked at the distribution of MreB within relatively large subcellular regions, and therefore we cannot distinguish between mutants solely on their organization along the sides. It would be possible in future work to use more sophisticated image analysis methods to analyze the precise steady state organization of MreB in those mutants that have a flat or broadly peaked profile. Since Caulobacter are so small and MreB is only localized to the sides for a relatively short time in the cell

P a g e | 117

cycle in wild type cells, it might be more appropriate to do this type of analysis in a larger organism (i.e. E.coli). Previous work in E.coli and B.subtilis has shown that D158 (D162 in Caulobacter) can affect both the dynamics (as assayed by photobleaching) and structure of MreB in vivo [141, 227]. Alternatively, it should be possible to measure changes in the dynamic behavior of single molecules of these mutant variants [228].

Polar MreB=polar cell wall synthesis? We observed that the polar localization of MreB is associated with the development of pointed poles and variable cell width. We did not directly observe the localization of the R185V186 duplication mutant, but given the tight correlation between Shape Mode 5 and Fluor Mode 1, we think it is likely that this particular mutant is also localized to the poles. In support of that idea, we found that wild type GFP-MreB when expressed in the background of the unlabeled R185/V186 duplication mutation localizes to the poles and the thin polar extensions of the cell (data not shown). It is possible that in these mutants, cell growth is redirected from the lateral sidewalls to the poles. There are some bacterial species that normally grow by adding new cell wall material to the polar regions, while keeping the lateral side walls inert [43]. Caulobacter and E.coli, however, are thought to grow normally by synthesizing new cell wall all along the sides, keeping the poles of the cell inert [42, 211, 229]. If we have indeed isolated a mutant that switches the mode of growth from lateral to polar in Caulobacter, this would be a very intriguing result. We have not directly looked at cell wall synthesis in this mutant, however. Currently there are no reliable methods for directly visualizing the growth of the wall in live Caulobacter cells, although there are ways of doing this in fixed cells. We did, however, localize MreC and Pbp2 in a couple of the polar mutants. These proteins are known to be involved in the regulation of cell width (See Chapter 2). Pbp2 is the elongation-specific transpeptidase that catalyzes the crosslinking of peptide side chains in the peptidoglycan (See Chapter 1); thus, it directly affects the molecular structure of the cell wall. The precise function of MreC is not known, although it is thought to regulate the localization of Pbp2 (Chapter 2, [141-143, 180, 205, 230, 231]). Neither MreC nor Pbp2 were found adapt a polar localization in strains expressing

118 | P a g e

CHAPTER 3

A325P or D16G MreB (Figure 16 and data not shown). Given that MreC and Pbp2 do not normally colocalize considerably with MreB in wild type cells (Chapter 2), this result is perhaps not surprising and provides further evidence to indicate that the localization of MreC and Pbp2 do not absolutely require a direct interaction with MreB and vice versa. Of course, the bulk localization pattern of Pbp2 does not necessarily reflect the sites where Pbp2 is active. It is possible that only a fraction of the molecules that we see in the fluorescence images are active at any time, particularly since the concentration of endogenously expressed Pbp2 in the cell is fairly low (~50/cell in E.coli grown in rich media, [48]). There could be transient activity at or near the poles. Looking directly at the sites where new wall is inserted in these mutants will hopefully help to answer this question. The localization of Pbp2 is often used to approximate the sites of new cell wall growth mainly because it is thought that the many different enzymes involved in the synthesis cell wall, including at minimum transglycosylases, transpeptidases and hydrolases, form a holoenzyme complex in the cell (also discussed in Chapters 1 and 6). Indeed there is some evidence to suggest that these enzymes interact with one another, including bacterial two-hybrid and affinity chromatography data [13, 25, 26, 143, 205, 207], but there is not definitive proof that these enzymes exist as an obligate complex in cells. The existence of a holocomplex was proposed as a way to tightly correlate sites of synthesis and destruction of the cell wall [13, 25]. It has been proposed that bacteria must synthesize new cell wall before destroying the old (―make before break‖, [13, 14]), to avoid compromising the structure that withstands the turgor pressure. However, recent evidence from computational modeling of the Gram negative bacterial cell wall indicates that small perturbations in the cell wall (i.e. destruction of peptide crosslinks) do not in fact lead to catastrophic loss of the whole structure [232] . Therefore, it may not be required that all of the enzymes colocalize or that their activity be absolutely spatially coordinated. In fact, different organisms may have different mechanisms.

How do changes in MreB localization affect cell shape? While the precise molecular mechanism by which MreB contributes to cell wall synthesis and the determination of proper cell shape remains unclear, the mutant strains

P a g e | 119

that we have characterized in this work provide additional tools with which to study this process and may have already provided some insight. In particular, we found that extent to which MreB is able to undergo the cell cycle regulated changes in localization pattern correlates most strongly with cell size (as measured by 2D area). This result suggests that the dynamic localization pattern of MreB in Caulobacter regulates the transition between the elongation and division phases of growth: recruiting MreB to the division plane in wild type cells slows elongation along the sides and redirects synthesis toward the division plane. Whether or not the presence of MreB at the division plane is functionally important for division is not known. If so, this function is not essential, however, given that we could isolate many mutants that no longer completely localize to the division plane. Instead this activity seems to control the size of the cells, which may have important consequences for fitness in different environmental conditions. Note that cell length is less well correlated with our measure of cell cycle regulation of MreB, and cell width is not significantly correlated with the localization of MreB. The precise dimensions of the cell may be controlled by a separate activity that we have not measured here, for example the ability of MreB to communicate and act on the enzymes involved in cell wall synthesis. We and others (i.e. [122, 229]) believe that MreB ensures that cell wall synthesis occurs evenly spaced along the perimeter, through a mechanism that is yet undefined. The accumulation of MreB at the poles is associated with cells with pointed poles and asymmetric cell width. This phenotype could result from MreB directly stimulating growth at the poles or directly inhibiting growth and stimulating growth in the space immediately next to the clusters of MreB. In either case, since MreB is depleted from the middle of the cell in the polarly localized MreB strains, cell growth is not properly dispersed and the cell width increases. Future work, including the incorporation of computational models for the growth of the peptidoglycan, will focus on extending and discriminating between these models.

Future directions The mutants described in this work provide us with new tools with which to study the mechanism of MreB in shape determination and the role that the nucleotide cycle

120 | P a g e

CHAPTER 3 plays in regulating MreB. To really connect the nucleotide cycle to MreB localization and function, we must characterize the activity of these mutants in vitro. In the next chapter, we present the work that we have done to characterize the in vitro behavior of purified wild type Caulobacter MreB, which will provide a basis for the future characterization of the biochemical activity of purified mutant variants. Our model for the regulation of MreB predicts that there are unidentified nucleotide-dependent binding partners of MreB. In Chapter 5, we will present our attempt to biochemically identify proteins that affect the assembly state of MreB in vitro. In Chapter 6, we present specific models for the structure of MreB in vivo and the specific mechanism by which it may influence cell wall growth.

AUTHOR CONTRIBUTIONS

I performed most of the experiments and image analysis. Zachary Pincus performed the clustering analysis and implemented the image swath function into Celltool to calculate the MreB distributions. Isabelle Fisher determined the optimal conditions for isolating A22-resistant Caulobacter, isolated the R185V186 duplication mutant, and performed the timelapse microscopy on this strain. Julie Theriot generated the faces shown in Supplementary Figure 5 and provided significant intellectual input. Lucy Shapiro also contributed significant intellectual input.

ACKNOWLEDGEMENTS

Mike Fero provided generous support for the imaging of the A22-resistant strain, training me on KAMS and the spectrophotometer and providing the high-throughput equipment. We are also grateful to Erin Barnhart and KC Huang for suggesting we use PCA to analyze the MreB distributions and for performing the initial implementation of this idea. Lastly, we received additional A22-resistance strains from Christine Jacobs- Wagner and Zemer Gitai.

P a g e | 121

MATERIALS AND METHODS

Bacterial growth conditions In this work, we grew Caulobacter only in rich media (PYE) at 28-30C, except for the Pbp2 localization presented in Figure 16. In this case, the strains were grown in M2G supplemented with xylose (0.03%) for 2 hours prior to imaging. We observed similar localization patterns in PYE, but the signal was much dimmer. Optical density was measured at 660nm. A22 was synthesized according to [40] and used at a final concentration of 2µg/ml in liquid or 2.5µg/ml on plates. In our experience, the effect of A22 declines considerably with the age of the stock solution. Thus, we prepared fresh stock solutions before each experiment (never longer than two days prior). This practice helped ensure more day-to-day reproducibility. In addition, we added A22 to liquid media only immediately prior to inoculation. To synchronize the cells, 10ml cultures in mid-log phase were spun at 10,000xg for 2 min in six epindorf tubes. Cells were pooled into one tube and washed with ice cold 1X M2 salts. Washed cells were resuspendend in 1X M2 salts and mixed with an equal volume of Percoll (cold). Cells were then spun at 10,000xg for 20 min at 4C. Stalk and predivisional cells were removed and the bottom band of swarmer cells were isolated. Swarmers were washed 2X with 1X M2 salts and then resuspended in PYE.

Selection of A22-resistant mutants To select for A22 resistance, wild type CB15N Caulobacter was grown overnight in PYE at 30C to an OD660nm ~1.0. Solid PYE plates 2.5µg/ml A22 were prepared by adding the appropriate amount of stock solution (in DMSO) to the molten media agar solution prior to pouring and solidifying. Overnight cultures were plated (200µl) on A22- containing PYE plates and grown at 30C for three days. At this point, ~1-50 colonies (of varying sizes) were visible on each plate. One colony per plate was restreaked onto a fresh plate containing A22 and grown for another three days. This step was actually very important, as mutant strains picked from the original isolation plate did not grow well overnight in liquid (unpublished observations). Each strain was grown overnight in a 2ml

122 | P a g e

CHAPTER 3 deep 96-well plate (USA Scientific) in 2ml PYE containing A22. Cells were then frozen in 10% DMSO and stored at -80C. The mutant Q26P was not isolated in our selection; it was obtained from the lab of Christine Jacobs Wagner at Yale [211]. The R186/V186 mutant was spontaneously isolated from streaking LS3814 (Pxyl:gfp-mreB; [117]) on an A22-containing plate. To identify mutations in mreB, we PCR amplified the gene (with flanking regions on both sides) using a small amount of liquid overnight culture. Products were purified with a 96-well PCR cleanup kit (Zymo Research) and sent to be sequenced by Sequetech (Santa Clara, CA).

Imaging the shapes of A22-resistant Caulobacter Many of the mutant strains we isolated become very filamentous and pleiomorphic at high density (unpublished observations). In order to get an accurate measure of the cell shapes of these strains in steady state log phase (rather than the shapes they adapt as they are transitioning between different phases), we tried to prevent the strains from ever becoming too dense. All of the selected strains (See Supplemental Table 2) were grown from a frozen stock or colony inoculation in PYE at 30C with or without A22 in a 2ml 96-well plate (USA Scientific) for only 12 hours, to an OD of ~0.3- 0.4. Then, these cultures were diluted 10-20-fold into prewarmed PYE media (with or without A22) in a Costar flat-bottom 96-well non-treated culture plate (Corning) covered with sterile microporous sealing film (USA Scientific). Growth was resumed on a shaker plate at 28C. Density was monitored every 20-30 min with a 96-well plate reader spectrophotometer. Doubling time was calculated using density readings taken over a 2 hour period (beginning ~1hr after dilution into the Costar plate). Half of the plate was diluted at a delay of 1-1.5 hours, to account for the time it would take to image the first set of cells. For wild type, which will not grow overnight in A22, we diluted the overnight culture grown in the absence of A22 into media containing A22 (grown for ~4- 6 hours in drug). To image the cells we used the medium-throughput techniques and materials developed in Ref [233], including a custom-built apparatus for making 48 agarose pads on a glass slide spaced at intervals compatible with a 96-well plate. Cells were imaged on

P a g e | 123

pads of 1% agarose in M2G or water. A custom-built 48-well metal stamp was sterilized and then used to transfer half of the cells from the 96-well plate to these 48 agarose pads. A large piece of cover glass was placed on top, and the slide was sealed with hot glue.

Fluorescent strains Fluorescent Venus fusions to MreB mutants were constructed by cloning the mreB gene into pXVENN-2 [234]. The gene was amplified from liquid overnight cultures of the mutant strain using primers to introduce NheI and BglII restriction sites. These enzymes were used to move mreB into pXVENN-2, and the resulting plasmid (also encodes resistance to Kanamycin) was electroporated into CB15N Caulobacter. This plasmid integrates into the genome at the xylose-regulated promoter [234]. Phage transduction (CR30) was used to move the fluorescent construct from the wild type background to the A22-resistant mreB background, selecting for Kanamycin resistance. To image the fluorescent strains, we grew them overnight in PYE+Kan+0.2% glucose (to repress Venus-MreB expression). Cells were then washed in equivalent media lacking glucose and grown for another ~2 hours. At an OD of ~0.1-0.2, xylose was added to 0.03%. Cells were allowed to grow for one hour in induction conditions and then imaged on 1% agarose/M2G. To increase the expression of the reporter slightly for timelapse imaging, we increased the concentration of xylose to 0.3% and grew the cells on 1% agarose/PYE. Fluorescent mCherry fusions to MreC and Pbp2 were cloned under the control of a Vanillate-inducible promoter in pVCHYC-1(C-terminal fusion) and pVCHYN-1 (N- terminal fusion), respectively [234]. The mreC gene was moved from the pXyl:mreC- mRFP1 plasmid (see Chapter 2) with NdeI and BglII enzymes. The pbp2 gene was moved from the pXyl:gfp-pbp2 plasmid (see Chapter 2) with KpnI and XbaI enzymes. The resulting plasmids were electroporated into wild type Caulobacter. Phage transduction was used to move these constructs from the wild type background into A22- resistant Venus-MreB strains, selecting for Spectinomycin resistance. The final strains mutant R have Pxyl:venus-mreB (Kan ), Pvan:mcherry-pbp2 (or Pvan:mreC-mCherry, both R mutant Spec ), PmreB:mreB . These strains were grown overnight in PYE+Kan+Spec+0.2% glucose (to repress the expression of the fluorescent reporters). Cells were washed with

124 | P a g e

CHAPTER 3

M2G Kan+Spec and grown for ~7-8 hours to an OD~0.2. Then, xylose and vanillic acid were added to final concentrations of 0.03 % and 0.5mM, respectively. Cells were imaged after two hours of induction. GFP-Crescentin expressed from its endogenous promoter (as well as unlabeled Crescentin) was transduced into selected mutant strains, selecting for Gentamycin (construct characterized in [123]).

Microscopy To measure cell shape in the unlabeled A22-resistant strains, we used a DM6000B automated microscope equipped with a 100 × 1.46 NA HCX Plan APO oil immersion objective (Leica) and a C9100 EM cooled CCD camera (Hamamatsu). Automated image acquisition was performed by KAMS-acquire software [233]. The 48- well microscope slide was scanned across 8x6 rows, imaging four fields in each well before moving to the next. Manual refocusing was performed as needed. To image the fluorescent strains, we used an upright fluorescence microscope (Zeiss, Thornwood, NY) equipped with a plan-apo 100x phase 3 objective lens, conventional epi-fluorescence filter set and a 1024 x 1024 pixel back illuminated EMCCD camera (Andor, South Windsor, CT). For the figures, the brightness and contrast was adjusted and overlay images were produced with Adobe Photoshop.

Analysis of cell shape Images of bacteria were segmented in Matlab (v. 2008a) with binary thresholding, followed by a marker-based watershed algorithm. We chose parameters that seemed reasonable by eye to separate touching cells and recently divided cells. The parameters were kept constant for all images taken under the same set of conditions. For the analysis of cell shape, we used CellTool, an open access software tool for measuring and analyzing cell shape [212]. Contours were extracted from the binary segmented images of bacteria and resampled so that each cell was given a contour of 100 evenly spaced points. The shape modes depicted in Figure 4 were the primary modes of variation in the dataset of all cells of all strains (numbers and strains listed in

P a g e | 125

Supplementary Table 2). To assess the correlation in the modes between strains, we calculated the mean shape mode of all cells in each strain and calculated Pearson’s correlation coefficients in the matrix of mean values for each shape mode in each strain.

Analysis of MreB distributions in mixed populations To measure the distribution of Venus-MreB along the centerline, we first segmented the phase images of the bacteria in Matlab and extracted resampled contours in Celltool as described above. Then, we defined a centerline as the line of 50 evenly- spaced points connecting the two most distant points in the contour (Celltool find_centerlines, endpoint method=distance), and matched each contour to its corresponding fluorescence image (Celltool extract_images). The intensity of MreB at each point along the centerline was measured (Celltool ―image-swath‖ option of measure_contours command, using ―length_profile‖ measurement mode). A depth value of 50 (how far away from the centerline to average) was chosen to average the intensity from the entire width of the cell at each point. We then normalized the total intensity values by forcing each cell to have the same average intensity, so that the only difference between the profiles is the distribution of this intensity along the centerline. Given that the fluorescent reporter is under the same inducible promoter in all of these strains, we feel that this is a reasonable assumption. For each contour we also measured cell area, length of the centerline, and average width of the contour. The PCA Fluor modes, the modes of variation in the fluorescence profiles, were generated in Matlab. We first generated a matrix of all normalized MreB profiles (all cells of all strains) and then calculated the mean profile by averaging all values at each point. This mean was then subtracted from each profile, and the covariance of the resulting values (the deviations from the mean) was measured. The PCA Fluor modes are the eigenvectors of this covariance matrix. The values for each mode were normalized by subtracting the mean and dividing by the standard deviation of all values in that mode (to generate a Z-score). For the ratio metrics, we divided the cell into five regions (points along the centerline): pole 1 (0-5), side 1 (6-15), center (16-35), side 2 (36-45), and pole 2 (46-50). For each region, we calculated the sum of MreB intensity at each point along the

126 | P a g e

CHAPTER 3 centerline (as above) in the region. The C:S (center:sides) ratio was taken as the center intensity divided by the sum of that in side 1 and side 2. The P:S (pole:sides) ratio was taken as the sum of intensity at pole 1 and pole 2 divided by the sum of that in side 1 and side 2. Pearson correlation coefficients were calculated for the matrix of true length and width (not shape modes), Shape Mode 2, Shape Mode 4, Shape Mode 5, Mean Fluor Mode 1, and the CCR Index for all the labeled strains (+/- A22 done separately). Shape parameters were taken from the large dataset of unlabeled strains.

Analysis of Venus-MreB timelapses Images were segmented as described previously. While this method was found to be far better than thresholding alone for separating newly divided cells, it is still not terribly accurate, as it is difficult to accurately determine the exact point of division with phase microscopy. Since we used similar parameters for the image acquisition and segmentation of all timelapses, we believe that at least this error will be systematic and not specific to certain strains. Cells that were obviously not segmented correctly, however, were discarded. The last frame before the two cells were separated at division was considered to be the ―division point‖, and division was estimated to occur halfway between the division point and the next frame. The division rates presented in Figure 13 are the mean and standard deviations in the division time. The rates of elongation were determined by fitting a simple exponential to the rate of increase in length with time. More accurate methods for determining the timing of cell separation and precise cell boundaries can and should be used in future work to detect subtle changes in the growth parameters of these mutants. In Figures 10-13, we reported the intensity of MreB along the centerline at each timepoint as a Z-score. To calculate this value, we calculated the mean and standard deviation intensity at each point for the entire timelapse. For each centerline point at each timepoint, we subtracted the mean and divided by the standard deviation to get a Z-score. To report this intensity as a function of cell length, rather than relative coordinate, we determined the length of each centerline point in each cell. Since each cell has 50 points, the length per point is equal to the length of the cell divided by 50. Normalized PCA

P a g e | 127

Fluor mode values were calculated using the axes defined by all cells (see previous section). The CCR Index is equal to the difference between Fluor Mode 1 at halfway through the cell cycle to that at the start of the experiment.

Clustering strains by shape The distribution of cell shapes from all conditions was summarized as a point distribution in five dimensions by PCA, as described in Pincus and Theriot 2008. The distribution was rescaled to have unit variances along each axis, and optionally, the first axis, which predominantly measured phase in the cell cycle, was dropped. To compare strains and treatment conditions, as opposed to individual cells, however, it is necessary to define a distance metric between sets of points in that five-dimensional space. To this end, we employed the "Earth Mover's Distance" [235], which poses the distance between two point-sets as a transportation problem: given the spatial locations of a set of "producers" and a set of "consumers" and the amount of production or consumption at each position, find a "transport plan" to get the product to the consumer with as little spatial movement as possible. Such problems can be solved in closed form with the "transportation simplex method" [236]. In the case that there are an equal number of cells in each treatment condition, the problem reduces to how to superimpose one set of cells on the other with as little movement as possible; when there are unequal numbers of cells, the total amount of "production" and consumption" is set to unity and spread equally among all of the cell positions, which will result in an assignment that optimally transform one set of points on to the other, allowing for splitting and merging of points to account for the unequal numbers. The total amount of transportation necessary to effect this transformation is then used as the "distance" between any two point-sets (i.e. sets of cells). Pair-wise distances were computed between all strains in the +A22 and -A22 treatment conditions separately, both with and without the first PCA axis. In addition, the sum of the +A22 and -A22 distances was used as an overall distance between any given two strains. After computing these distances, average-linkage hierarchical clustering was applied to the distance matrix in order to visualize the relationships between the strains with and without A22.

128 | P a g e

CHAPTER 3

Highlighting residues on the crystal structure To generate Figure 20, we used UCSF-Chimera to align the protein sequences of T.maritima and Caulobacter MreB and manipulate the crystal structure of TM MreB1 complexed with AMPPNP (PDB: 1JCG, [8]). Note that some of the residues that we identified are not conserved in T.maritima (for example, C110 and S181).

Using Chernoff faces to display the cell shape phenotypes of A22-resistant Caulobacter In Supplemental Figure 5, we display the measured cell shape data for each A22- resistant Caulobacter strain as a series of Chernoff faces [237], as described in the figure legend.

P a g e | 129

SUPPLEMENTARY FIGURE 1

130 | P a g e

CHAPTER 3

SUPPLEMENTAL FIGURE 2

P a g e | 131

SUPPLEMENTAL FIGURE 3

132 | P a g e

CHAPTER 3

SUPPLEMENTAL FIGURE 4

P a g e | 133

SUPPLEMENTAL FIGURE 5

134 | P a g e

CHAPTER 3

P a g e | 135

136 | P a g e

Chapter 4 In vitro assembly of Caulobacter MreB

Natalie Dye, Lucy Shapiro, and Julie Theriot

P a g e | 137

ABSTRACT

In this chapter, we demonstrate that recombinant Caulobacter MreB purified from E.coli forms a polymer in vitro under a wide variety of solution conditions, including in the absence of nucleotide. Light scattering and fluorescence microscopy techniques were used to determine that the MreB polymers are relatively small, and do not form long filaments like the homolog of MreB in T.maritima. While we were unable to obtain monomeric MreB, we observed changes in the size of the polymer by varying the concentration of salt and temperature. Using electron microscopy, we show that the purified polymer is not necessarily filamentous. With standard glow-discharged grids, the protein polymers appeared globular and rather amorphous. With grids coated with poly- L-lysine, however, MreB polymers were found to be filamentous, forming straight filaments up to one micron long and containing 1-4 protofilaments. We interpret these data to mean either that the purified sample contains a mixture of amorphous and filamentous polymeric forms or that interactions at the surface are important for MreB assembly. Lastly, we show that MreB can bind directly and nonspecifically to DNA. By comparing our work to the existing published data on T.maritima MreB and B.subtilis MreB, we show homologs of MreB from different bacterial species can have very different biochemical behaviors. The work presented here will provide a framework for future in vitro work on the role of the nucleotide in the assembly of MreB and its interaction with other proteins.

138 | P a g e

CHAPTER 4

INTRODUCTION

In vitro biochemical assays have been, and continue to be, extremely useful for the study of cytoskeletal proteins. By studying purified actin and microtubules, much has been learned about the inherent dynamics of their polymerization, the role that nucleotide binding and hydrolysis play in this process, and the function of numerous interacting partners in altering polymerization dynamics and filament structure. Many complex cytoskeletal-based processes can even be reconstituted in vitro using purified components, including the actin-based movement of Listeria monocytogenes [238-240] and ParM-based movement of DNA-coated particles [144], allowing us to quantitatively study the mechanistic details of these processes. Thus, with the discovery of MreB as a bacterial actin homolog, it has become important to develop methods for the purification and in vitro polymerization of MreB. Early work towards this end was performed by the research group that solved the crystal structure of Thermatoga maritima MreB1, hereafter refered to as TM MreB1 [8]. These authors cloned one of the two homologs of MreB encoded in the T.maritima genome and purified it by affinity chromatography after recombinant expression in E.coli. The other homolog (mreB2) has not been characterized. TM MreB1 was found to be soluble (at centrifugation speeds >100,000xg) in the absence of ATP and insoluble upon the addition of millimolar levels of ATP or GTP. With electron microscopy, this insoluble fraction was found to form long, highly bundled, filaments (several microns in length) that are composed of a varying number of straight protofilaments. These filaments have a longitudinal repeat of 5.1nm and a lateral spacing of 3.9nm, which is consistent with the dimensions of MreB monomers and similar to the spacing of monomers in actin filaments. Subsequent work has shown that the polymerization of TM MreB1 requires nucleotide (ATP or GTP), is favored by increasing temperature, low pH, increasing concentration of MreB, and millimolar concentrations of divalent cation, and inhibited by high concentrations of monovalent salts [241-243]. While it requires nucleotide to polymerize, the critical concentrations for polymerization in ADP and ATP differ by no more than 3-fold, indicating that TM MreB1 is not likely to exhibit dynamic instability or rapid treadmilling [241]. Light scattering and fluorescence- based experiments have been used to show that TM MreB1 forms large, stiff, highly

P a g e | 139

bundled polymers [208, 241], but fine scale analyses of the ultrastructure of MreB under varying solution conditions have not been performed. While the data on TM MreB1 provide some insight into the inherent polymerization properties of MreB, it is difficult to directly correlate this behavior in vitro with that in vivo, since T.maritima is an anaerobic extremophile that grows at high temperature (65-80C) and is not amenable to genetic manipulation or cell biology experiments. We do not know the in vivo function, localization, or dynamics of T.maritima MreB1 or its paralog MreB2. While T.maritima is rod-shaped, it has an unusually prominent and extended cell envelope (called the ―toga‖) that lies separated away from the cell cytoplasm; thus, it is morphologically quite distinct from the standard bacterial model organisms. From work that has been done in these model bacterial organisms, it is clear that different MreB homologs may have slightly different functions. MreB is thought to have a role in chromosome segregation in Caulobacter but not in other organisms [40, 204, 244-247]. In B.subtilis, there are actually three homologs of MreB, and each has a slightly different morphological phenotype when deleted [116, 121]. It is also known that the subcellular localization pattern and dynamics of MreB localization can vary by species [204]. For example, Caulobacter MreB undergoes a cell- cycle-regulated transition between a length-wise helix and a midcell ring [39, 117], whereas B.subtilis MreB remains in a lengthwise helix throughout [116, 177]. Thus, there is reason to believe that the in vivo behavior of MreB homologs can vary considerably by organism. It is not known whether these differences are due to variations in the intracellular environments, the presence of various interacting proteins, or perhaps variations in the inherent polymerization properties of the MreB homologs. While the primary sequence of actin is conserved extremely well throughout eukaryotes, the sequence of MreB is more variable. The amount of identity between TM MreB1 and that of B.subtilis MreB, E.coli MreB, and Caulobacter MreB is ~50-60%. In species that encode multiple MreB paralogs (B.subtilis, T.maritima and others), often these paralogs are no more similar to each other than to homologs encoded in evolutionarily distant organisms, indicating that different types of MreB may exist even within the same organism. How much the variation in protein sequence contributes to variation in polymerization properties and biochemical behavior is not known. It is conceivable that

140 | P a g e

CHAPTER 4 the polymerization of the homologs of MreB from these bacterial species could be quite different from that of TM MreB1. As mentioned in previous chapters, Caulobacter is a useful model organism in which to study the bacterial cytoskeleton. It contains components of all the known classes of bacterial cytoskeletal proteins. It contains only one mreB homolog, which makes the analysis of function and localization simpler than those organisms that have multiple paralogs. It is synchronizable, which makes analysis of cell cycle regulated transitions easy to study. It is also relatively thin. While this property makes light microscopy more challenging, it also makes cryoelectron tomography (cryoET) feasible [248, 249]. Cytoskeletal filaments are not typically observed in whole bacterial cells using traditional electron microscopy techniques. The only way in which the ultrastructure of bacterial cytoskeletal filaments has been observed in vivo has been with cryoET [105, 137, 155]. The first detection of FtsZ filaments in whole cells was achieved with cryoET in Caulobacter at least in part because the cells are thin (relative to other bacterial model organisms) and thus easier to image [105]. While MreB has not yet been visualized with cryoET, its structure, dynamics, and localization patterns have been investigated in other ways in Caulobacter. Using super- resolution light microscopy techniques and single molecule fluorescence imaging, W.E. Moerner’s group has shown that the MreB structure that appears roughly helical with standard epifluorescence microscopy is actually fragmented and composed of relatively short filaments or clusters of MreB [228, 250]. This organization does not seem to be consistent with the structure of the long, stiff TM MreB1 filaments that have been observed in vitro, either because the Caulobacter protein inherently forms polymers with a different structure or because the cellular environment affects the structure of MreB polymers. A detailed characterization of the polymerization of Caulobacter MreB would help answer this outstanding question. In addition, we described in the previous chapter a series of mutations that can drastically alter the subcellular localization and dynamics of MreB, as well as the shape of the whole cell. Since these mutants were isolated by their ability to resist the destabilizing drug, A22, which binds to the nucleotide-binding pocket of MreB, it seems possible that these mutants could behave very different biochemically, either in their ability to polymerize or bind and hydrolyze nucleotide. With an in vitro

P a g e | 141

assay for Caulobacter MreB, we could begin to link inherent biochemical behavior with cellular localization, structure, and activity. In this chapter, we describe the purification of Caulobacter MreB and the preliminary characterization of its polymerization in vitro. We find that Caulobacter MreB exists in a polymer form in a wide variety of solution conditions, even in the absence of nucleotide. Unlike what has been described for TM MreB1, we were not able to isolate the monomeric form of MreB, though we were able to observe differences in polymer size based on solution conditions. Also unlike TM MreB1, the polymers we observed were found to be small and not necessarily filamentous. Using negative stain electron microscopy to visualize the polymers, we found that MreB does not appear filamentous when prepared and stained with standard methods. Instead, it appears to form globules of a consistent size that are either roughly spherical (~20nm in diameter) or slightly extended (~14nm wide by 40-60nm long). Filaments could be observed, however, when poly-L-lysine was first coated onto the EM grids. These filaments are quite distinct from those of the TM MreB1 filaments, being no longer than one micron in length and ranging in width from 1-4 protofilaments. Fluoresence microscopy and light scattering experiments were consistent with the results of the electron microscopy and high speed centrifugation. Our data suggest that there are indeed significant differences in the biochemical behavior of MreB homologs from different species. The methods developed here may be extended to assess the biochemical activity of mreB mutants and the role of potential interacting partners.

142 | P a g e

CHAPTER 4

RESULTS

Attempt to purify native MreB from Caulobacter Prior to the beginning of this work, Zemer Gitai in the Shapiro lab had purified recombinant Caulobacter MreB from E.coli and used the purified protein to generate an antibody [117]. At the time, no biochemical characterization was performed. When we set out to analyze the polymerization properties of MreB in vitro, we initially thought it could be possible to purify endogenously expressed MreB directly from Caulobacter. Traditional approaches for obtaining purified actin and tubulin rely on the ability to cycle these proteins between polymeric and monomeric states, using centrifugation to separate the two forms away from other components of the cell. Similar methods have been used more recently to purify recombinant ParM and TM MreB1 from E.coli cell lysates [111, 241]. We then developed a protocol that relied on a series of high speed centrifugation steps (300,000xg) to fractionate MreB away from the other cellular components and used a Western blot as an assay to detect the presence of MreB (Figure 1A). Based on how actin and ParM behave in extract, we made the assumption that we could control polymerization of MreB in extract by altering the concentration of salt and ATP. Using high speed centrifugation to separate filament from monomer, we presumed that we should be able to purify MreB from other components by a series of polymerization cycles. We lysed Caulobacter cells in a moderate ionic strength buffer resembling that used in the native purification of ParM (TKM buffer, see Methods) and found that about half of the MreB can be found in the low speed supernatant by Western blot (Figure 1B). That supernatant was then spun at high speed to pellet polymeric MreB (and other large structures), and indeed about 90% of the MreB was found in the pellet fraction. This pellet was resuspended in a buffer lacking salt (G2 buffer, See Methods), sonicated briefly, and dialyzed in this buffer overnight. The next day about 50% was present in the high speed supernatant, indicating that at least some of the polymers were able to depolymerize overnight in this buffer. Upon the addition of salts and ATP to resemble the ―polymerization buffer‖ used by van den ent et al (F1 buffer, see Methods, [8]), a significant portion of the remaining MreB was found in the pellet fraction (Figure 1B).

P a g e | 143

Thus, we were able to complete a full cycle from high speed pellet, to supernatant and back to pellet. With this cycling, we were able to achieve considerable fractionation of the cell lysate (Figure 1B). Unfortunately, we were never able to cycle this pellet fraction back into the supernatant fraction for further chromatography (data not shown). Additionally, through these cycling steps we lost a fair amount of total protein, and we were never able to achieve very high yields of MreB to perform further purification and biochemical experiments. Thus, we decided to return to the recombinant expression approach and purify a His-tagged version of Caulobacter MreB in E.coli.

FIGURE 1

144 | P a g e

CHAPTER 4

Recombinant expression and native purification of Caulobacter MreB Caulobacter His-MreB could be expressed to high levels in E.coli, though its expression causes the E.coli cells to change shape and lyse (Figure 12A and data not shown) so its expression time was minimized. To purify MreB, clarified cell lysate was run through a column of Nickel-NTA resin, and further enrichment was achieved with ion exchange and gel filtration chromatography (see Methods, Figure 2A). Hereafter, this purification strategy will be referred to as the ―native‖ preparation. Upon lysis, MreB is in the soluble fraction after a centrifugation step of 60,000rpm for one hour (data not shown), indicating that at least most of the MreB is cytosolic and does not exist as aggregates or large filamentous polymers. Upon elution from the Nickel column, however, the conformation or oligomeric state of MreB changes: it will no longer rebind to the Nickel (upon dilution of the imidazole, data not shown); the His tag cannot be cleaved by Thrombin, though it can be cleaved when MreB is still bound to the Nickel resin (data not shown, see Methods); a large percentage of the protein will now pellet with high speed centrifugation (Figure 2B); and it elutes in the void volume of Superose 6 and Superdex S200 gel filtration columns, even under fairly extreme conditions (1M NaCl, pH 10, data not shown). Note that a prominent contaminant of approximately 45kDa copurifies and copellets with MreB (asterisk in Figure 2A-B). This band was identified with mass spectrometry to be EF-Tu, an essential factor for protein translation. We were not able to separate EF-Tu from MreB with high concentrations of NaCl, addition of GTP or GDP, divalent cation, or by changing pH; however much of the EF- Tu could be removed from the initial extract with a longer centrifugation step prior to the Nickel column (data not shown). Interestingly, EF-Tu and MreB have very recently been shown to interact in vitro and in vivo [251]. In addition, elongation factor 1α can bind and bundle actin filaments [252-258]. Thus, the interaction between Caulobacter MreB and E.coli EF-Tu may have some physiological and biochemical relevance, though we did not pursue this question further. The ability to pellet with high speed centrifugation and elute in the void volume of a gel filtration column indicates that purified MreB exists as a very large complex, presumably filamentous polymers like those of T.maritima MreB1. Unlike TM MreB1, however, Caulobacter MreB does not require the addition of ATP to form these

P a g e | 145

polymers. ATP is typically not added to the buffers during purification, and its addition does not change the results. It is not uncommon for nucleotide-binding proteins to retain nucleotide from the cellular extract, however; thus it seemed possible that the purified MreB was already bound to nucleotide. Indeed, we found that the ratio of absorbance at 260nm to that at 280nm is approximately 1.0 after purification, indicating that there is likely to be nucleotide retained by MreB through multiple purification steps (including gel filtration and many dialysis steps).

146 | P a g e

CHAPTER 4

FIGURE 2

P a g e | 147

Denaturing preparation of Caulobacter MreB To obtain a preparation of nucleotide-free MreB and remove all traces of contaminating EF-Tu, we chose to purify MreB under denaturing conditions. Actin cannot be refolded upon denaturation, so it was not obvious that this procedure would produce active MreB. Nonetheless, prokaryotic tubulin homolog FtsZ can refold without chaperones while tubulin cannot [259]. While FtsZ/tubulin and MreB/actin are different protein families entirely, the fact that FtsZ can refold in the absence of chaperones indicates that it is possible for some prokaryotic cytoskeletal proteins to fold autonomously. Purification of MreB was achieved using the Nickel resin in 8M urea and gel filtration in 6M urea. The urea was then slowly removed with dialysis to allow MreB to refold, either with ATP, ADP, or no added nucleotide. This purification protocol will hereafter be referred to as the ―denaturing‖ preparation (see Methods). The 260/280nm absorbance ratio was 0.65 after refolding in the absence of nucleotide, indicating that indeed the denaturing preparation was successful in removing residual nucleotide. Once equilibrated in concentrations of urea of 2M or less, MreB was again found to pellet during high speed centrifugation (Figure 2C), indicating that it can also form large polymers upon refolding even in the absence of nucleotide. MreB did not pellet with a low speed centrifugation step (data not shown), so these polymers are likely to be relatively small and not large aggregates or bundles.

High speed centrifugation of purified MreB In order to study the properties of polymerization of any protein, it is valuable to be able to isolate the monomeric form of the protein and control its transition to polymer. Since MreB already appeared to be polymer upon purification, we next used the high speed centrifugation experiments to rapidly screen through multiple different solution conditions, looking for monomeric MreB (Figure 2D-H). These experiments were performed with both the natively purified and refolded preparations of MreB with nearly identical results (although only those of the refolded are shown in Figure 2). MreB was found almost entirely in the pellet fraction after equilibration in 100mM NaCl (or KCl) and millimolar concentrations of Magnesium, Calcium, or EDTA, with or without ATP,

148 | P a g e

CHAPTER 4

ADP, GTP, or GDP (Figure 2D-E and data not shown). Similar results were obtained with non-hydrolyzable ATP analogs, as well as with 1M NaCl, 0.6M Potassium Iodide, and brief pulses of sonication (data not shown). In 100mM NaCl, pH also did not appear to significantly alter the monomer/polymer equilibrium (Figure 2F). In basic pH, more MreB was present in the supernatant fraction, although even at pH 10, MreB was found to elute in the void volume of a gel filtration column (data not shown); thus this soluble form of MreB does not appear to be monomeric. The absence of salt was the condition that seemed most promising and was the only condition in which a significant difference was observed between natively purified and refolded protein (Figure 2G-H). In a minimal buffer lacking salt (G2A, see Methods), natively purified MreB is nearly completely in the soluble fraction, whereas refolded MreB continues to pellet well. While this result was quite promising, the protein continued to flow through a gel filtration column in the void volume even in this buffer (data not shown), indicating that this soluble fraction is also not completely monomeric but likely to be smaller polymers. Omitting the ATP from this no-salt buffer tended to make the natively purified MreB precipitate, but the refolded protein still pelleted (Figure 2H). The difference between refolded nucleotide-free MreB and natively purified MreB, which is likely still bound to nucleotide, indicates that nucleotide could be involved in regulating the size of MreB polymers. We thought it unlikely that the small His-tag was interfering with polymerization or inducing aggregation, but this was still a troubling possibility. Thus, we used Thrombin to cleave at a site in between MreB and the His-tag (see Methods). In a low salt buffer in the absence of ATP, cleaved MreB (refolded or native) continued to elute in the void volume of a gel filtration column (data not shown), indicating that the His-tag is not causing the aggregation of MreB in the absence of ATP. There was also the slight possibility that the recombinant overexpression in E.coli somehow altered the structure of MreB to cause irreversible aggregation. To test this possibility, we expressed His-MreB in Caulobacter using the Xylose-inducible promoter. This promoter, in single copy on the chromosome, was not capable of inducing the high levels of expression that were observed in E.coli. Nonetheless, with large amounts of Caulobacter culture, biochemical amounts of His-MreB could be purified. Again, in a low salt buffer in the absence of ATP, MreB purified from Caulobacter eluted in the void volume of a gel filtration

P a g e | 149

column (data not shown), indicating that the ability of MreB to form large polymers in the absence of ATP was not an artifact of its overexpression in E.coli.

FIGURE 3

150 | P a g e

CHAPTER 4

Electron microscopy of MreB polymers: glow discharged grids Even though the natively purified and refolded MreB preparations were found to behave so similarly in the pelleting and gel filtration experiments, we reasoned that there could be subtle differences in the structures of the polymers formed that could not be detected with these bulk assays. Additionally, the results described above indicate that the natively purified protein is likely to form different types of polymers or polymers of different sizes under varying salt concentrations. To examine the ultrastructure of these polymers, we used electron microscopy, focusing on the differences between native and refolded MreB in the presence or absence of salt (Figures 3-9). With standard methods of sample preparation for negative stain TEM, both natively purified and refolded MreB appeared as globular, fairly amorphous, aggregates in the electron microscope (Figure 3A-D). We observed the protein under all of the conditions used for the pelleting reactions and obtained similar results. In the absence of salt, these aggregates were still visible, though less frequent, and no other unique structures were observed (data not shown). These aggregates had a very consistent size and were either round (Figure 3A), with a diameter of ~20nm, or elongated and rod-like, with a length of 40-50nm and a width of 10-16nm (Figure 3B). The amount of these globular aggregates that we observed appeared to be proportional to the concentration of MreB that was added to the grid, and no such aggregates were visible when only buffer was added to the grid (data not shown), arguing that these globular aggregates are indeed MreB. This result was quite surprising because these globular and relatively amorphous complexes do not at all resemble the long bundles of filamentous polymers seen with TM MreB1. Occasionally large filamentous bundles that resemble those seen with TM MreB1 were observed (Figure 3D); however, these bundles were relatively rare, whereas the globular aggregates were abundant. Additionally, only a small percentage of MreB was found in the pellet fraction of a low speed centrifugation that should pellet large aggregates of this size (data not shown), indicating that it is unlikely that the bulk of MreB is found in these large filamentous bundles. It is not clear whether these large bundles are MreB at all. Because this was such a surprising result, we made every effort to ensure that the

P a g e | 151

purification scheme itself was not yielding inactive, aggregated protein. Several modifications were made to the native purification protocol, including the addition of detergents, A22, and ATP, but no differences were observed in the appearance of the polymers. The protein continued to appear amorphous over several experiments with one exception. On one day, we were able to find structures that appeared to have an ordered, filamentous organization (Figure 3E-F). These structures appeared to be either straight and tube-like or flat and curved. Remarkably, the tube-like structures appeared to have the same dimensions as the elongated rod-like polymers of MreB: ~60-70nm long and 14nm wide. The curved filaments appeared to be composed of 2-4 protofilaments, each about 4nm wide. Unfortunately, these structures were never seen more than once, even when the protein was purified in the exact same manner and the EM grids were prepared with the same procedure. In addition these structures were only localized to a specific area of the grid; only globular, amorphous structures were visible in other areas of the same grid. Nonetheless, the similarity in size and dimension between these one-time appearing structures and the regularly-appearing MreB aggregates made us wonder whether the two species were actually the same and that we were not able to image the MreB polymers reliably, either because the process of preparing the EM sample was destructive or that the sample was not properly immobilized and adhered to the EM grid prior to and during the staining procedure. Thus, we experimented with other stains that are not as acidic as Uranyl Acetate (Ammonium Molybdate and Phosphotungstic Acid) and other methods of altering the charge of the surface of the Carbon-coated EM grid (omitting the glow discharge; glow discharging in the presence of Argon, water or Amylamine; and pretreating the grid with Poly-L-lysine). We obtained similar results with all procedures except for the Poly-L-lysine (PLL) treatment. Pretreating the EM grid with PLL produced dramatically different results and was therefore used for much of the remaining experiments to compare native and refolded MreB preparations in the presence and absence of salt.

152 | P a g e

CHAPTER 4

FIGURE 4

P a g e | 153

FIGURE 5

154 | P a g e

CHAPTER 4

FIGURE 6

P a g e | 155

Electron microscopy of MreB polymers: Poly-L-Lysine-coated grids At high concentrations (20-30uM) of freshly natively purified MreB in 200mM NaCl, abundant micron-long filaments were visible on PLL-coated grids (Figure 4A). At higher magnification, these filaments appeared to be composed of several straight protofilaments (Figure 4B). At lower concentrations of protein (1-10uM) in 100-200mM salt, MreB filaments were more rare, but still visible (Figure 4C-G). These filaments could be up to one micron in length and composed of two or more protofilaments, each 4- 5 nm in width, which is consistent with the size of a single molecule of MreB. We next looked at the refolded preparation of purified MreB and found that it too is capable of forming long filaments in 100-200mM NaCl with or without the addition of ATP (Figures 5 and 6). Filaments composed of two, three, four and more protofilaments could be found, though filaments of two protofilaments were the most common. Unlike actin, the protofilaments of either native or refolded MreB do not twist around one another but instead lay flat and parallel to one another. Often the filaments were either incompletely adhered to the grid or poorly stained (or both), with sections (especially the ends) that are blurred and poorly resolved (Figures 5 and 6). In addition, the filaments were relatively rare, given the concentration of MreB that was applied (compare with the density of sample in Figure 3). The same concentration of actin or FtsZ completely covers the surface of the grid (data not shown), but with MreB, only a handful of filaments could be observed. Either these filaments were, in fact, only a rare fraction of the sample or the filaments did not readily attach to the grids even after PLL- coating. Thus, extensive analysis of filament dimensions was not practical under these conditions. The effect of the PLL treatment of the EM grids is quite surprising: not only do we see filaments that are not visible on a regular glow-discharge treated EM grid, but we also fail to see the relatively smaller rod-like or amorphous aggregates. Upon closer inspection, however, we realized that in addition to the filaments, we could also detect small, round particles that were of a consistent size and appeared to be made up of either rings or coils (Figure 7A-C). Similarly sized particles could also be observed on PLL- treated grids in the absence of added MreB (Figure 7C), but these particles did not appear to have any particular order or structure, indicating that perhaps these coils and rings

156 | P a g e

CHAPTER 4 could be comprised of MreB or a copolymer of PLL and MreB. In addition, we also detected clusters of particles that resembled the rod-shaped or amorphous aggregates that appear on glow-discharge treated EM grids (Figure 7A, E-F). These particles tended to cluster together unless the sample was applied at very high concentrations of protein or in the presence of high concentrations of salt (0.6M NaCl, Figure 7D-F). All types of structures—filaments, rings/coils, and amorphous aggregates—were present with natively purified and refolded MreB, in the presence and absence of added ATP. The differences between the types of particles that were observed were clearly due to the treatment of the EM grid. When the same exact preparation of refolded MreB was applied to a PLL-treated and a glow-discharged grid, two different outcomes were observed: on PLL, small clusters and long filaments were observed; on regular glow discharge, only rod-like and amorphous aggregates were observed (Figure 7D-E). If that same sample was first incubated in 0.6M NaCl and then applied to a PLL-treated grid, dispersed rod-like and amorphous aggregates were observed (Figure 7F). In the absence of salt, we also observed filaments on PLL-treated EM grids with both natively purified and refolded MreB, with and without the addition of ATP (Figures 8-9). While in salt, MreB filaments are at least two protofilaments wide, in the absence of salt, filaments of only one monomer wide could also be observed. The presence of filaments in the absence of salt was not unexpected, given the results of the pelleting, gel filtration and light scattering (see below) experiments. These data cannot explain, however, why the natively purified preparation fails to pellet in this buffer whereas the refolded preparation does. It is possible that the filaments observed for the native preparation in the absence of salt are not representative of the total sample. It is also possible that these filaments are not as stable as the refolded filaments and depolymerize under the force of centrifugation. Alternatively, it is also possible that this result is an artifact of the PLL: perhaps the sample was induced to polymerize or laterally interact on the surface by the PLL and these polymers are not representative of the sample in solution. While we cannot rule out this possibility, the light scattering data and gel filtration data do suggest that there are indeed polymers present in the solution. In addition, FtsZ and actin do not appear to significantly alter their structure when adhered to PLL-coated EM grids (data not shown). Finally, it is possible that the MreB polymer

P a g e | 157

can change its structure in meaningful and interesting ways when bound to a surface. In vivo, the localization of MreB appears helical along the surface of the cell, whereas the localizations of ParM and MamK, other prokaryotic actin homologs, appear to be fairly linear [109, 112, 137]. Thus, it is possible that the surface plays an important physiological role in regulating MreB structure and dynamics.

FIGURE 7

158 | P a g e

CHAPTER 4

FIGURE 8

P a g e | 159

FIGURE 9

160 | P a g e

CHAPTER 4

Dynamic light scattering of MreB polymers The electron microscopy experiments seem to suggest that there could be multiple forms of polymerized MreB of varying size and structure. Electron microscopy is neither quantitative nor dynamic, however. Furthermore, as we have seen for MreB, the surface preparation and the staining procedures required for EM can have complicated effects on the sample and the resulting data. Thus, it remains unclear whether or not MreB actually forms polymers of varying types and sizes in solution. To develop an orthogonal solution-based method of estimating MreB polymer size, we turned to dynamic light scattering (DLS). In theory, this technique should be able to discriminate particles of varying sizes that are mixed together in a single solution. It measures particle size in solution relatively quickly without disrupting the sample. Whereas many of the established methods for measuring polymerization (pyrene or tryptophan fluorescence, FRET, and right-angle light scattering) can only be used to follow the transition between monomer and polymer, DLS should also be able to detect two polymers that have a fairly significant difference in size and shape. While DLS is not routinely used to study cytoskeletal protein polymerization, we found that the globular and filamentous forms of purified actin are easily distinguishable by this assay (Figure 10A). Thus, we used this technique to determine the sizes of the particles of purified MreB in solution in the hopes of correlating these sizes with those of the particles observed in the EM. To begin, we measured the size of natively purified MreB after dialysis against buffers containing either no salt (G2A buffer) or 100mM NaCl (F1 buffer). After an overnight dialysis in the no-salt buffer, MreB fails to pellet in a high speed centrifugation (See Figure 2); with the DLS, however, we can see that MreB still forms a relatively large complex, displaying a prominent peak around 10-20nm and smaller peak around 100-200nm (Figure 10B). After the high speed centrifugation, only the smallest peak remains. After overnight equilibration in F1 buffer, MreB forms a very large complex that scatters a significant amount of light (out of the range of detection for this assay, Figure 10C). F1 buffer contains salt, Magnesium Chloride and ATP, but the addition of fresh ATP alone to MreB in G2A was found to be sufficient to rapidly increase the size of MreB (Figure 10D). The increase in size that occurred over 20 minutes following fresh ATP addition was not as large, however, as that induced by an overnight equilibration in

P a g e | 161

F1 buffer, indicating that additional aggregation, annealing, or maturation occurs over time. Oddly, adding salt to the G2A sample prior to the addition of ATP did not cause a similar increase in size, at least during this time window (data not shown), indicating that salt may actually slow polymer growth stimulated by ATP. MreB refolded either with or without nucleotide produced similar results: in the absence of salt, it formed a complex that was slightly larger than that of natively purified MreB in a no-salt buffer (~20-30nm radius); in TK buffer, it formed a very large complex (Figure 10E). Remarkably, even though refolded MreB appears to be only slightly larger than natively purified MreB in G-buffer, it is able to pellet with high speed centrifugation whereas natively-purified cannot. In addition to salt and ATP, we found that an increase in temperature alone could trigger an increase in size for refolded nucleotide-free MreB (Figure 10F) in the no-salt buffer. Temperature did not, however, have the same effect on MreB refolded in ADP or natively purified MreB in the no-salt buffer (data not shown), arguing that nucleotide-free MreB may be more primed for polymerization than nucleotide-containing MreB.

162 | P a g e

CHAPTER 4

FIGURE 10

P a g e | 163

For the refolded nucleotide-free MreB preparation, we also measured the size of the most prominent peak at two different temperatures under many of the same conditions that were tested in the high speed centrifugation experiments (Figure 10G). In general, 100-200nM NaCl or KCl, with or without added divalent cation or ATP, significantly increased the size of MreB over that in the no-salt buffer and increasing temperature increased the size even further. While no condition was found to decrease the size of MreB to that corresponding to a monomer, some conditions (basic pH and KI) were able to inhibit the temperature-dependent increase in size.

Determination of Critical Concentration for polymerization At no point during the DLS experiments were we able to detect the hypothetical monomeric form of MreB. Dynamic light scattering is not practical for measuring low concentrations of small proteins, however. Thus, if there is indeed a low concentration of monomeric MreB in these samples, it would probably not be detectable with DLS or any of the assays we have used so far. While we have not been able to identify a solution in which MreB would be present as a monomer at micromolar concentrations, it is still likely that there exists a concentration below which MreB does not form polymers (linear or otherwise). To estimate this ―critical concentration,‖ Ccrit, we measured the amount of polymer present in solutions of varying concentrations of MreB using right angle light scattering (Figure 11). In this assay, monomers or small aggregates (significantly less than the wavelength of light, in this case 400nm) produce only a very weak signal, whereas larger polymers produce large signals. For a simple protein polymer at steady state, the amount of polymer in solution should be linear with increasing concentrations of protein above the critical concentration, and Ccrit can be estimated as the X-intercept of this line (or more accurately as the intersection between the line connecting the concentrations of polymer above Ccrit with that connecting the datapoints from the weak signal below Ccrit). We performed this experiment with natively purified and refolded, nucleotide-free MreB in 0.1M KCl without added ATP. Natively purified MreB produced little to no signal at concentrations below 0.5uM and linearly increasing signal at higher concentrations. Ccrit was estimated to be 0.67uM under these conditions. For refolded nucleotide-free MreB, the results were more complex. At concentrations of 0.8-3uM, the

164 | P a g e

CHAPTER 4 slope of the line corresponded well with that of natively-purified MreB, indicating that the two preparations were likely to be forming similar types of polymers in this concentration range; using these datapoints only, Ccrit was estimated to be 0.37uM. At low concentrations (0.1-0.5uM), however, there was still appreciable signal for nucleotide-free MreB. If all the datapoints are included, Ccrit can be estimated to be 0.19uM. It is possible that nucleotide-free MreB, similar to nucleotide-free actin, forms polymers that are more stable and require a significantly longer time to reach equilibrium. It is also possible that at these low concentrations nucleotide-free MreB forms smaller polymers that scatter less light for a given concentration of protein than the polymers that form at higher concentrations. Unfortunately, without reliable ways of visualizing the types of filaments formed by natively-purified and refolded MreB, particularly at low concentrations, it is difficult to distinguish between these two possibilities.

FIGURE 11

P a g e | 165

Purification of GFP-MreB To develop yet another method for estimating polymer size and characterizing the biochemical behavior of MreB, we fluorescently labeled MreB. We took two approaches to labeling MreB with a fluorescent molecule: one, engineering a fusion to GFP and expressing and purifying the product; and two, chemically labeling MreB with a reactive fluorophore. In Caulobacter, N-terminal GFP-fusions are partially functional. It cannot complement a wild type deletion when expressed from the Pxyl promoter, but its localization pattern when expressed with wild type resembles that of immunofluorescence [39, 117]. In E.coli, a fully functional fusion protein has been made by engineering the coding sequence for mCherry into an internal loop of MreB [224]. We tried to make this same fusion to the Caulobacter protein, but the product was still not able to complement the deletion of the wild type protein (data not shown). Thus, we chose to work with the His-tagged N-terminal GFP-MreB (with the His-tag fused to the N-terminus of GFP). As with wild type, we were able to achieve reasonably high levels of expression of this fusion protein, though again the expression of this protein made E.coli cells round and misshapen (Figure 12A). It is not clear whether it is the exogenous expression of the Caulobacter protein or the significant change in the levels of this protein that makes the E.coli cells change shape and lyse. Much more of GFP-MreB was present in the insoluble fraction upon cell lysis than was unlabeled MreB, indicating that the presence of the GFP fusion may be particularly toxic. We looked at the localization of GFP-MreB during its expression in E.coli and found that it was largely diffuse but with a prominent fraction bound to the membrane (Figure 12A). This membrane localization could be due to either GFP-MreB having an inherent affinity for the membrane or the Caulobacter GFP-MreB binding to host E.coli proteins on the surface of the membrane (i.e. RodZ [223, 224]). The diffuse fluorescence could be due to premature termination of the fusion protein message or cleavage between MreB and GFP; or, this pattern could reflect the true localization of the fusion protein when overexpressed in E.coli. We purified GFP-MreB with a similar protocol as that used to natively purify unlabeled MreB. As with the wild type protein, GFP-MreB was found to have a hydrodynamic radius of ~12nm after dialysis in G-buffer, indicating that the fusion protein can also form polymers (Figure 12B). With the addition of fresh ATP, the size of

166 | P a g e

CHAPTER 4 these polymers increased, though they did not reach sizes as large as the unlabeled protein (compare Figure 12B with Figure 10D). With these same conditions, we used light microscopy to image the sample. In G-buffer, the protein formed a nearly uniform layer of fluorescence on the surface of the glass with only occasional bright puncta visible (data not shown). This localization is consistent with the relatively small size of the protein observed with DLS. After the addition of fresh ATP, there were numerous bright puncta all over the surface of the glass (Figure 12B). Long filaments were not observed. Given the low resolution of light microscopy, we cannot distinguish between small clusters or aggregates and filaments that are up to 200-300nm long. Nonetheless, the formation of visible puncta is also consistent with the increase in size that was observed in the DLS experiments. Some of the puncta appeared elongated and could correspond to short filaments (300-500nm), but filaments greater than one micron in length were not prominent. No change was observed with the addition of varying concentrations of salt, Magnesium or nonhydrolyzable ATP analogs (ATPgammaS, AMPPNP, data not shown). One of the motivations for pursuing this biochemical project was to develop assays with which we could identify and study MreB interacting proteins. Inspired by the recent discovery that ParM could bind to and polymerize off of DNA-coated beads in the presence of ParB and ATP [144], and knowing that MreB is involved in chromosome segregation [40, 219], we thought that it could be possible to develop a biochemical assay for DNA-segregation using DNA-coated beads, purified GFP-MreB, and Caulobacter extract. In the course of our initial experiments toward this end, however, we found that GFP-MreB alone could bind very well to DNA-coated beads (Figure 12C). This binding was not sequence specific and occurred even in the presence of BSA and up to 1M NaCl to block nonspecific binding. MreB contains no DNA-binding domains so the nature of this interaction and its physiological relevance is unclear.

P a g e | 167

FIGURE 12

168 | P a g e

CHAPTER 4

Chemical labeling of MreB to a fluorophore As stated above, GFP-MreB is not fully functional in Caulobacter cells. The DLS data indicates that GFP-MreB polymers do not grow as large as unlabeled polymers with added ATP. In addition, we found that GFP-MreB did not pellet as efficiently in F-buffer as unlabeled MreB (data not shown). Thus, it seems that GFP-MreB may interfere with polymerization such that the polymers are not as large or as stable as unlabeled MreB polymers. In EM experiments, GFP-MreB looked very similar to wild type on regularly glow-discharged grids (data not shown), and PLL-coated grids were not tried. To develop an alternative method for generating a fluorescent MreB, We chemically labeled natively purified MreB with the Lysine-reactive fluorophore, Dylight-488. Unfortunately, since we cannot cycle MreB from the pellet fraction into the soluble fraction (data not shown), protein that is inactivated by the labeling on Lysines remains in the preparation. By varying the reaction conditions, we were able to achieve labeling efficiencies (average % of molecules labeled) from 14-80%. Because we thought it likely that 100% labeling could interfere with polymerization (given the data on GFP-MreB and actin), we generally aimed for 30-40% labeling, although the results did not seem to be significantly affected by this percentage. We performed the labeling in salt concentrations in which MreB should be present in relatively large polymers that pellet at high speed centrifugation. With light microscopy, Dylight-MreB appeared to be approximately the same size as GFP-MreB (Figure 13A). Again no long filaments were observed, even when unlabeled MreB was mixed in with Dylight-MreB. Of course, we had no particular reason to expect that Caulobacter MreB could form long polymers. Even the filamentous polymers that were observed on PLL-coated EM grids were no longer than one micron. Only the very rare filamentous bundles observed on glow-discharged grids were several microns in length (Figure 3D); in fact, the lack of abundant long filaments in the fluorescent samples provides further evidence that the long bundles depicted in Figure 3D are not representative of the bulk of the purified MreB sample. Nonetheless, we thought it would be useful for further biochemical assays to find a condition in which MreB polymers would elongate to lengths detectable by light microscopy. Because the fluorescence microscopy experiment is so simple, we were able to rapidly screen through

P a g e | 169

numerous different conditions in search of one that would yield longer polymers. We tried adding monovalent, divalent, trivalent and tetravalent ions, glycerol, polyethylene glycol, sucrose, ATP, AMPPNP, and GTP. We found no condition that produced a large amount of long filaments (data not shown). Because of our experience with PLL in electron microscopy, we also looked at fluorescent MreB on the surface of PLL-treated glass and found it to be qualitatively similar to untreated glass, indicating that PLL attached to a glass surface is also not capable of inducing long polymers to form (data not shown). Note that in all of these experiments, we were specifically looking for dramatic changes, and thus no sophisticated analysis of particle size was performed. It is possible that some of these conditions do, in fact, affect the finer structure of the polymers or subtly alter their size, but because of the poor resolution, this assay is not appropriate for examining these types of changes. We also looked at the effect of adding PLL and other long charged polymers to MreB in solution. When added to MreB in a low-salt solution, PLL induced the formation of large fluorescent clumps (Figure 13B). In 0.1M NaCl, nearly all of the MreB was recruited into even larger stringy aggregates (Figure 13C). Similar clumps were formed, albeit less efficiently, with the addition of purified Lysozyme (data not shown). The same was not true with the addition of poly-Glutamate, poly-Proline, or poly-Phosphate at similar concentrations and molecular weights (Figure 13D and data not shown). Given our result that GFP-MreB could bind DNA-coated beads, we also experimented with adding DNA to labeled MreB. We observed large bright puncta with the addition of supercoiled plasmid DNA (data not shown). We also found that MreB could coat condensed DNA (isolated mitotic chromosomes, as well as DNA condensed with millimolar concentrations of spermine or spermidine; data not shown). Remarkably, in the presence of linearized DNA, we observed long, straight, and very fluorescent filaments (Figure 13E). These particles were even visible with phase microscopy. In the electron microscope, using glow-discharged grids, these filaments appeared to be nearly crystalline and very electron dense (Figure 13F-H). Note that on these grids we also observed abundant amorphous and rod-like aggregates similar to those observed with unlabeled MreB (Figure 3), indicating that the labeled MreB is likely to be forming similar polymers as unlabeled MreB and that not all of the MreB is binding the DNA and

170 | P a g e

CHAPTER 4 forming these filaments, even with very high ratios of DNA:MreB. The same results were obtained with GFP-MreB and unlabeled MreB, indicating that this behavior is not an artifact of the label. These filaments were stable to treatment with DnaseI and 1M NaCl. While at first we were very excited and intrigued to find that DNA induced the formation of these very large polymers, we later observed that the same filaments could be formed by mixing linearized plasmid DNA with millimolar concentrations of Spermidine (data not shown). This result indicates that it is not the unique ability of MreB to form polymers that is inducing the formation of these large filaments; instead it more likely a property of the charge distribution on the surface of MreB and the linearized DNA fragment. We also used the fluorescent versions of MreB (both GFP-labeled and Dylight- labeled) to determine that MreB does not bind to lipid vesicles made of phosphatidylcholine, phosphatidylglycerol, or a mixture of E.coli polar lipids (data not shown). Thus, the apparent association of MreB with the membrane in vivo is likely to be mediated by other cellular factors.

P a g e | 171

FIGURE 13

172 | P a g e

CHAPTER 4

DISCUSSION

The assembly state of purified Caulobacter MreB Here, we have described the recombinant purification and preliminary characterization of Caulobacter MreB. We used high speed centrifugation, gel filtration chromatography, dynamic light scattering, right angle light scattering, fluorescence microscopy, and electron microscopy to determine that MreB forms a polymer in numerous different solutions conditions, including in the absence of nucleotide. Under no conditions were we able to isolate a high concentration of monomer. While it always behaves as a polymer, variations in the size and structure of MreB polymers were observed. In the absence of salt, MreB formed a relatively small polymer and a larger polymer with added ATP, divalent cation, and salt, and with increased temperature. With electron microscopy, these polymers appeared globular on glow-discharged grids and filamentous on grids pretreated with poly-L-Lysine. Filaments on PLL grids were either single or double protofilaments in the absence of salt, whereas in the presence of salt, filaments were two or more protofilaments wide. Our DLS data suggest that MreB is able to form polymers of varying sizes depending on the MreB preparation and the solution conditions: one with an approximate hydrodynamic radius of 10-30nm, one that is at least 10X larger than that species (with a hydrodynamic radius of 100-500nm), and possibly also even larger species whose sizes cannot be accurately determined with this assay. The smallest particle detected by DLS could correspond to the amorphous and rod-like aggregates that were abundant on glow- discharged EM grids (Figure 3) and on PLL-coated grids at high protein concentrations and high salt (Figure 7E-F), which each have a radius of ~10nm when spherical or are 10-14nm wide by 40-60nm long when elongated. Alternatively (or in addition) this DLS signal could correspond to the small single protofilaments that can be seen on PLL- coated grids in the no-salt buffer (Figures 8-9). The small coils and rings that were observed only on PLL-treated grids have a diameter of approximately 14nm, which is slightly smaller than the observed DLS radius, but in theory this signal could correspond to these particles as well. The other prominent DLS species, with a radius of around 100- 500nm, could correspond to the long filaments that were only observed on PLL-coated

P a g e | 173

grids. The fact that this size range was observed in solution with DLS suggests that these filaments could in fact exist in solution, rather than be an artifact of the PLL-treatment; however, this correlation alone does not prove their existence and indeed this DLS radius could be derived from larger clumps of the amorphous aggregates (like those seen on PLL-coated grids, Figure 7A). The largest DLS radius could be derived from either larger filamentous bundles, like those seen at high concentrations of protein on PLL (Figure 4A), or larger aggregates, like those seen in Figure 3D. Our electron microscopy data could be interpreted in a several different ways. Firstly, it is possible that the solution of MreB contains multiple different types of polymers that differentially adhere to surfaces, with a globular aggregate form that preferentially adheres to the glow-discharged grid and a filamentous form that adheres only to PLL-coated grids. This scenario could result from differences in the charge distribution or hydrophobicity of the exposed surfaces of the two different types of polymers. Alternatively, it is possible that the solution is actually quite homogenous and the preparation of the EM grid surface determines the results. For example, in solution MreB could be present only in the small amorphous aggregates that are visible on regular glow-discharged grids; in this case, the filaments that are observed on PLL could be the result of the interaction between PLL and MreB inducing the formation of a filamentous polymer only on the surface. Conversely, it is possible that the solution is entirely comprised of the filamentous form and that PLL stabilizes these filaments on the surface so that they can be visualized; in this case, the amorphous aggregate structures that are observed on regularly glow-discharged grids could be due to incomplete adherence to the surface or destructive staining. At this time, we cannot rule out any of these scenarios. EM stains, in particular Uranyl Acetate, have been known to cause protein aggregation and unfolding of purified proteins. In addition, while rare, it is not unheard of for a protein to poorly adhere to the surface of a glow-discharged EM grid. In fact, nucleotide- free actin is one pertinent example. While standard preparations of polymerized actin can be easily visualized with standard electron microscopy techniques, actin that has been removed of nucleotide (NFA, nucleotide free actin) can only be visualized with electron microscopy in the presence of a small amount of Phallodin, a drug that binds to the sides of the filament [260, 261]. In other assays, NFA was found to be remarkably stable and

174 | P a g e

CHAPTER 4 polymerize at very low concentrations, so the reason for the Phallodin requirement for electron microscopy imaging is unclear. Given that TM MreB1 [8], ParM [10], MamK [262] and actin form filamentous polymers in solution, it seems plausible that Caulobacter MreB does as well. Nonetheless, we also believe it is possible that MreB is only forming clusters in solution and the filaments observed on PLL are not a prominent form of the protein in solution or result from the surface treatment. Such behavior could be the true behavior of Caulobacter MreB, an artifact of the purification procedure, or due to the absence of some unknown required component or solution condition. This is an important issue that needs to be resolved, as it has important mechanistic implications for the behavior of MreB in cells. Additional methods for visualizing MreB polymers that do not require adherence to a surface or staining with electron-dense molecules, such as cryoelectron microscopy or Atomic Force Microscopy, may be needed to resolve this issue. More sophisticated light-based techniques that distinguish between the shapes of similarly sized polymers, anisotropy for example, could also be informative. Since we have not worked directly with TM MreB1, it is difficult to compare our results directly with those that have been published for TM MreB1. To date only three papers have published electron micrographs of TM MreB1, and these are shown at fairly low magnification [8, 242, 243]. While the authors did not mention that they experienced any difficulty with the standard preparation of the grids, it is possible that the images of long bundled filaments are not completely representative of sample. No images of B.subtilis MreB have been published.

The influence of salt on MreB polymers in vitro Of all the conditions tested, salt and temperature appeared to have the most significant effects on polymer size. In the absence of salt, natively purified MreB fails to pellet in a high speed centrifugation but still flows through the void volume of gel filtration columns and has a hydrodynamic radius of ~12nm. Filaments were quite prominent on PLL-coated grids, including single protofilaments that were >100nm in length. Refolded MreB had a slightly larger hydrodynamic radius and was able to pellet in a high speed spin. With electron microscopy, small single protofilaments were

P a g e | 175

observed. It is not clear why the refolded MreB preparation is able to pellet in this buffer while the natively purified does not, particularly since they seem to form similar structures (if anything the natively purified protein forms longer filaments). As mentioned above, however, electron microscopy may be unreliable with this protein. It is possible that the natively purified protein is less stable and falls apart under the high G- force. It seems surprising, however, that this structure would not fall apart in any of the other solution conditions tested (including 0.6M KI). In the presence of salt, the filaments that were observed on PLL-grids were at least two protofilaments wide. No single filaments were observed, suggesting that salt promotes lateral interactions. In general, we observed much fewer filaments on the PLL grids in the presence of salt, however, and the filaments that were present appeared to be either incompletely adhered or poorly stained, indicating that we may not be able to visualize the entire sample in these conditions and that salt may inhibit the adherence of the filaments to PLL grids. It has been observed that the polymerization of both T.maritima MreB and very recently B.subtilis MreB is inhibited by monovalent salts [241, 263]. We did not obtain similar results: Caulobacter MreB was found to be stable as a polymer in many different concentrations of NaCl and KCl. We did note, however, that high concentrations of NaCl inhibited the temperature-dependent increase in size and the increase in size of natively- purified MreB upon addition of ATP (See Figure 10). It is possible that Caulobacter MreB is not as salt-sensitive as the other homologs. We were never able to isolate a monomer, however, whereas both B.subtilis and T.maritima MreB were monomeric without divalent cation at low temperature. It is possible that our purified protein is in an aggregated, non-filamentous form, and our experiments are reflecting transitions out of that aggregated state into a filamentous state or a larger aggregated state. If so, the biochemical behavior of the Caulobacter homolog could be quite different from other MreB homologs.

The influence of temperature on MreB polymers in vitro Increases in temperature were found to increase the size of Caulobacter MreB polymers (See Figure 10G). This result is similar to those obtained with B.subtilis and T.maritima MreB. With TM MreB1, a change in temperature from 4C to 20C is able to

176 | P a g e

CHAPTER 4 stimulate polymerization even in the absence of salt or divalent cation [241]. The polymers formed by this temperature transition were found to scatter light but less efficiently than polymers formed by the addition of divalent cation, leading the authors to postulate that there are in fact two stages of polymerization: one that is divalent cation- independent, stimulated by temperature and leads the formation of short polymers; another that is divalent cation-dependent and leads to the formation of longer bundled filaments. It could be that Caulobacter MreB polymerizes in two stages as well: the first stage of polymerization is the formation of the stable particle of ~12nm radius, and the second stage is the formation of the ~100nm radius particles or filaments. While T.maritima is adapted to live at 65-80C, Caulobacter is adapted to 25-30C. It is interesting to postulate that the reason that Caulobacter MreB cannot fully depolymerize is that it requires a very low temperature that is impossible to reach without the addition of cryoprotectants, which are likely to further stabilize polymeric MreB.

The effect of nucleotide on MreB polymers in vitro In this work, we did not show any data on the role that the nucleotide cycle (binding, hydrolysis, and release) plays during the polymerization of Caulobacter MreB. We did, however, do some preliminary work with fluorescent analogs of ATP to measure nucleotide binding and exchange. We saw no binding whatsoever of ethenoATP to refolded MreB (data not shown). MANT-ATP exhibited a significant change in fluorescence when added to refolded MreB that was extremely rapid and did not decay even after 48 hours (data not shown). These data are difficult to interpret, however. Since we cannot depolymerize Caulobacter MreB, the conditions of this assay were such that MreB should have been a large polymer. The nucleotide binding pocket is thought to be hidden and inaccessible in actin filaments, and nucleotide binding and exchange is thought to occur in a monomer form. In fact, monomeric TM MreB1 binds well to ethenoATP [241]. The lack of binding of ethenoATP to Caulobacter MreB, therefore, is consistent with nucleotide exchange not occurring in the polymeric state. The binding of MANT-ATP may or may not be reflecting nucleotide exchange. Particularly since the kinetics of binding and release between MANT-ATP and Caulobacter MreB are so unusual, we think it is possible that the MANT-ATP nonspecifically associates with the

P a g e | 177

surface of MreB polymers rather than binding to the nucleotide binding pocket. In addition, we measured no appreciable ATPase activity (with a phosphate release assay) for nucleotide-free (refolded) MreB (data not shown). Again, since we were not able to depolymerize MreB in order to ensure nucleotide exchange, the lack of activity is likely due to the lack of nucleotide uptake. Nevertheless, the structure of the tubulin homolog FtsZ is capable of allowing GTP exchange when in a filamentous state [99]. Thus, the binding of nucleotide to MreB may not necessarily follow the same process as that of actin, and more work will be needed to address this topic. By refolding MreB in the absence of nucleotide, we showed that Caulobacter MreB does not require nucleotide to form polymers and filaments. The same is true for actin. As mentioned above, actin purified in the presence of sucrose can be removed of nucleotide and polymerizes extremely well [260, 261]. In this case, the nucleotide is required for turnover of the filaments, as NFA filaments are remarkably stable in the polymer form. In fact, NFA has a much lower critical concentration than ATP- or ADP- bound actin, and the structure of the filament is slightly altered. The authors of this work proposed that the low critical concentration was due to the artificial stabilization of the ―closed‖ conformation of the nucleotide-binding cleft by the high concentrations of sucrose. Many members of the actin superfamily sharing the same structural fold exhibit nucleotide-dependent structural changes [106, 226]. To what extent this is true for MreB and how nucleotide affects polymerization is still unclear. Very recently, it was shown that B.subtilis MreB also polymerizes regardless of the type or presence of nucleotide [263]. Critical concentrations for polymerization were equivalent in ATP, ADP, AMP- PNP at the same temperature. For TM MreB1, no polymerization was observed in the absence of nucleotide, but critical concentrations for ATP and ADP bound MreB were only modestly different [241]. We did not measure the critical concentrations of MreB in different nucleotides (in part because it is not clear that nucleotide exchange occurs without cycling through the monomeric state, as discussed above). We did, however, observe a modest difference in critical concentrations of refolded vs. natively purified MreB (Figure 11), which could be due to the presence of bound nucleotide in the natively purified protein. We observed a consistently high 260:280 ratio in the natively purified

178 | P a g e

CHAPTER 4 protein preparation, though we have not yet been able to identify this component. In general, it seems that the presence or type of nucleotide does not significantly affect the amount of polymer present at steady state. Thus, it is very unlikely that MreB undergoes dynamic instability like another prokaryotic actin, ParM. It seems likely that cellular interacting partners are involved in regulating the polymerization cycle or that the nucleotide cycle of MreB regulates its interactions with other cellular components. The bound nucleotide could, of course, also affect filament structure (length, twist, or number of protofilaments). There has not yet been a thorough study of the structure of MreB filaments containing bound nucleotide at high resolution for any of the MreB homologs. The drug A22 has been shown in vivo to delocalize MreB [40]. In vitro, it binds directly to TM MreB1 and prevents the assembly of TM MreB1 into long rigid polymers; but, its inhibition is incomplete, as small polymers are still visible with fluorescence microscopy [208]. The critical concentration increases from 0.5uM (ATP-bound) to 2uM (A22-bound), which is similar to the critical concentration of ADP-bound MreB (1.7uM) [208, 241]. Thus, for TM MreB1, A22 appears to induce an ADP-like state in MreB. We tried adding A22 to Caulobacter MreB at various stages in our purification and biochemical experiments in the hopes of inducing a monomeric state. Even when A22 was added to the E.coli cells expressing Caulobacter MreB prior to cell lysis and throughout the purification procedure, MreB was not monomeric as judged by gel filtration and dynamic light scattering (data not shown). If, however, A22 is only inducing an ADP-like state, and the ADP-bound state of MreB is still stable as a polymer, we would not expect to see drastic differences in the behaviors with and without A22. Given the precedents set by actin and microtubules, we think it is likely that the nucleotide cycle plays an important role in the function of MreB in the cell, and future work on this topic will be crucial for understanding this bacterial actin homolog. In the last chapter we described a series of point mutations that alter the localization and perhaps function of MreB. Many of these mutations surround the nucleotide-binding pocket and thus are likely to be involved in regulating the nucleotide cycle. In this chapter we have described the initial characterization of wild type Caulobacter MreB, and it is now possible to use similar techniques to characterize purified mutant variants and analyze their biochemical behaviors. We think it is likely that at least some of these

P a g e | 179

mutants will have altered biochemical behaviors, an increased or decreased propensity to form filaments for example. Ideally, we could use also biochemical techniques to assess the ability of these mutants to bind, hydrolyze and release nucleotide. These types of experiments would allow us to directly assess how changes in the nucleotide cycle affect the polymerization cycle and link changes in biochemical behavior with cellular phenotypes. We think it is also possible, however, that changes in the nucleotide cycle do not drastically alter the polymerization of MreB, since Caulobacter MreB polymerizes with or without nucleotide and B.subtilis MreB polymerization is not affected by the type or presence of nucleotide. In this case, we may find that the mutant proteins form polymers that are indistinguishable from wild type in vitro. This result may indicate that the mutant phenotype is the result of changes in the interaction with partner proteins in the cell. With the initial characterization of Caulobacter MreB performed, we can also now think about developing biochemical methods for the identification of regulators of MreB polymerization and other interaction partners. In particular, it seems likely that there could be a cellular factor that is involved in regulating the disassembly of MreB and/or nucleotide exchange. Preliminary work on developing a biochemical assay for MreB disassembly will be described in the next chapter.

180 | P a g e

CHAPTER 4

MATERIALS AND METHODS

Reagents Chemicals were purchased from Sigma (unless otherwise indicated). A22 was synthesized according to [40]. While it was not typically used, we did attempt to use A22 to depolymerize MreB at a final concentration of 10µg/ml. When nucleotide was added to buffers, it was dissolved directly in the buffer, and the pH was adjusted accordingly. When nucleotide was added to MreB directly, we added small amounts of stock solutions prepared in 25mM Tris, pH 7.5. Solutions containing urea and potassium iodide were prepared immediately prior to use. Reducing agent (β-mercaptoethanol or DTT) was added to the solutions just prior to use. Purified actin and FtsZ were generous gifts of M. Footer and E. Goley, respectively.

Buffers (in alphabetical order) BW buffer: 10mM Tris, pH 7.5 at 20C, 1mM EDTA, 2M NaCl CK buffer: 100mM Sodium carbonate, pH 8.3, 50mM KCl (no DTT!)

G1 buffer: 2mM Tris, 0.5mM EDTA, 0.5mM ATP, 0.2mM CaCl2, 1mM DTT, pH 8 at 4C G2 buffer: 2mM Tris, 0.5mM EDTA, 1mM DTT, pH 8 at 4C G2A buffer: 2mM Tris, 0.5mM EDTA, 0.5mM ATP, 1mM DTT, pH 8 at 4C

F1 buffer: 100mM Tris, pH 7.4 at 20C, 100mM NaCl, 6mM CaCl2, 4mM MgCl2, 2mM ATP. Based on the ―polymerization buffer‖ used in Ref [8].

F2 buffer: 100mM Tris, pH 7.4 at 20C, 100mM NaCl, 4mM MgCl2, 2mM ATP. Same as F1 buffer except for the Calcium. Ni Buffer A: 50mM HEPES, 250mM NaCl, 20mM imidazole, 10% glycerol, pH 7.9 at 4C. 1mM β-mercaptoethanol was added immediately prior to use. Ni Buffer B: 50mM HEPES, 500mM NaCl, 1M imidazole, 10% glycerol, pH 7.9 at 4C Q Buffer A: 20mM Tris, 10mM NaCl, 0.5mM EDTA, 0.5mM DTT, pH 7.6 at 4C. Q Buffer B: Q buffer A+ 1M NaCl. S Buffer A: 40mM HEPES, 1mM EDTA, 10% glycerol, pH 7.1 at 4C. 1mM DTT was added immediately prior to use.

P a g e | 181

S Buffer B: S Buffer A + 1M NaCl. pH adjusted to 7.1 at 4C. TEAK1 Buffer: 30mM triethanolamine, 100mM KCl, 1mM DTT, pH 7.8 TEAK2 buffer: 20mM triethanolamine, 20mM KCl, 0.5mM DTT, pH 7.8 TP buffer: 10mM Tris, 0.1M Sodium phosphate, pH 6.8 at 4C TPU buffer: TP buffer + 8M urea, pH 8. Prepared on day of experiment. TPU wash buffer: TPU buffer but pH adjusted to 6.3. Prepared on day of experiment. TPU elute buffer: TPU buffer but pH adjusted to 4.5. Prepared on day of experiment. TK buffer: 50mM Tris, 100mM KCl, 1mM EDTA, pH 8 at 4C TKG buffer: TK buffer + 10% glycerol

TKM buffer: 30mM Tris, pH 7.5 at 20C, 25mM KCl, 1mM MgCl2, 1mM DTT. Based on that used in Ref [111]. TN buffer: 100mM Tris, pH 8 at 4C, 200mM NaCl, 0.5mM EDTA, 1mM DTT

Cycling of endogenous MreB in Caulobacter extract Wild type CB15N Caulobacter was grown overnight at 30C in PYE rich media to saturation. This culture was then diluted 100X into fresh, prewarmed media and allowed to grow to late log phase (OD660nm=0.6-0.7). Cells were harvested in a JA10 rotor at 10,000 rpm for 20 min at 4C. Supernatent was discarded, and pellet was washed with TKM buffer and respun. Final pellets were resuspended in 0.01X original volume of TKM buffer, flash frozen in liquid nitrogen and stored at -80C. To lyse the cells, we added to the defrosted slurry: lysozyme to 0.2mg/ml, magnesium chloride to 2mM and DnaseI to 5µg/ml and incubated on ice for 20min. The cells were then sonicated with a microtip sonicator at the maximum level for 4 pulses of 10 sec. The lysate was spun for 20min at 16,000xg at 4C (the ―low speed spin‖), and the supernatant was removed to a new tube. For the ―high speed spin‖, we spun the low speed supernatant at 100,000rpm (300,000xg) in a TLA100.4 rotor for 15min at 4C. Supernatents were removed to a new tube, and pellets were partially resuspended in G2A buffer (lacking EDTA, though in retrospect, this probably would have enhanced depolymerization). This high-speed pellet pellet (shiny, yellowish-brown, containing MreB) was quite difficult to resuspend. A pipet was used to scrape the contents out of the tube and into a new tube, and several

182 | P a g e

CHAPTER 4 rounds of pipetting and sonication (gentle pulses) were able to get the contents mostly resuspended (in retrospect, should have used a homogenizer to fully resuspend this pellet). The resuspended pellet was dialyzed (12-14,000 MWCO) against fresh buffer overnight at 0C. The sample was then subjected to another high speed spin (as above). At this point, much of the MreB is found in the supernatant. To cycle it back into the pellet, we added a mixture of components to make the buffer contents equal that of F1 buffer. The sample was incubated at room temperature for 20-30 min and then 4C for another 2 hours. The high speed centrifugation was repeated, and samples from each step were run on SDS-PAGE. In attempt to increase the expression of MreB in Caulobacter, we used a strain with an extra copy of mreB at the xylX site to overproduce MreB in a xylose-dependent manner (LS3810) [117]. However, we found that overexpression caused more MreB to pellet in the low speed supernatant (data not shown), and thus we did not end up using these cells.

Cloning Standard molecular biology techniques were used to clone mreB and gfp-mreB into expression plasmids for induction in Caulobacter and E.coli. The expression plasmid containing Caulobacter mreB in pET28a and the E.coli strain (C43) containing this construct was a gift of Z. Gitai. The strain of Caulobacter containing his-mreB at the xylX site for xylose-inducible expression was also a gift of Z. Gitai. For the expression of gfp- mreB, we constructed a variant of the original fusion protein. PCR was used to amplify mreB from Caulobacter with primers containing BglII and HindIII restriction sites. These enzymes were used to clone mreB into pXGFP4-C1, which fuses GFP to the N-terminus. The gfp-mreB sequence was moved from this plasmid into pET28a (Novagen) with NdeI and HindIII. This construct was electroporated into BL21 E.coli.

Growth of E.coli and induction of His-MreB E.coli strains were grown in LB media while shaking at 37C. Starter cultures of 20-50ml LB were grown overnight after inoculation with either a frozen stock or colony. This culture was then used to inoculate 2L of pre-warmed LB media, and the culture was

P a g e | 183

allowed to grow until an optical density of 0.4-0.5 (600nm). IPTG was added to a final concentration of 0.5mM, and the cells were grown for another 2-2.5 hours. Cells were harvested with a 20min spin at 10,000rpm in a JA10 rotor equilibrated at room temperature. Typically, we harvested 2x2L of culture at once: the first culture was spun, the supernatant was discarded, and then the second culture was added on top of the first culture. Cells were washed with ~0.1X the original volume of Ni Buffer A and respun. The final pellet was resuspended in 0.01X the original volume with Ni Buffer A, flash frozen in liquid nitrogen, and stored at -80C until further use. We experimented with longer induction times but found that the yield did not increase with induction times longer than 3 hours, probably because the cells begin to lyse. We did not try altering the temperature of induction.

Native purification of recombinant MreB Frozen E.coli cell slurry was defrosted in a water bath at 37C . β-mercaptoethanol, PMSF, and Leupeptin were added. Lysis was performed with either a pre-cooled French press or Avestin EmulsiFlex-C5 homogenizer at ~10,000psi. To remove cell debris, the lysate was centrifuged for 30 min at 23,000xg (14,000rpm in Fiberlite F15) at 4C. The lysate was further clarified with a centrifugation at 55,000rpm in a Beckman Ti60 rotor at 4C for 30-60min. In earlier preps, we used a shorter time for this spin, whereas in later preps, we used longer times. The only difference that we could see is that more EF-Tu can be removed with longer spins at this step. The supernatant was then filtered first through a 5µm filter, followed by a 0.45µm filter, and then loaded into a sample loop. Chromotography was performed with an Akta Purifier FPLC (GE Biosciences) equipped with an in-line spectrophotometer. Absorbance was followed at 256nm and 280nm. We poured a 2.5ml column of NiNTA (Qiagen) and equilibrated it Ni Buffer A. Recharging the resin was performed between preps (according to the Qiagen manual). After loading the sample, the Ni column was washed until the absorbance reached baseline. Sample was eluted with a gradient of 0-40% B over 15CV. Fractions containing MreB, as detected by molecular weight and Western blot, were pooled and dialyzed (12-14,000MWCO) for 2-3 hours into S Buffer A, changing the buffer about halfway. The sample was centrifuged at 23,000xg for 20min to pellet the

184 | P a g e

CHAPTER 4 small amount of protein that precipitates during dialysis. Sample was syringe-filtered as described above and loaded onto the next column. Cation exchange chromatography was performed with HiTrap S (GE Biosciences) at pH 7. Sample was eluted with a gradient of 0-100% B over 15CV. After the Ni and S columns, MreB is quite pure. Because we have problems concentrating the protein (see notes below), we typically used the most pure fractions from the S column for our experiments. If necessary, we performed gel filtration chromatography with Superose 6 or Superdex S200. For the gel filtration experiment at high pH, we used a pre-packed Superose 6 HR 10/30 column (GE Biosciences) equilibrated in 50mM CAPS, 1M NaCl, 1mM EDTA, 1mM DTT at pH 10, and the sample was run at a flow rate of 0.5ml/min. Note that the predicted pI of MreB is ~6.5, which should indicate that anion exchange, rather than cation exchange, chromatography would be appropriate. We found, however, that MreB can bind to both S and Q columns at pH 7-7.5. Since the S column provided the best enrichment, as judged by Coomassie stained SDS-PAGE gels, we used this column for most of our preps. Care must be taken with MreB at pH ~7, however. In the low salt S load buffer at pH 7, MreB sometimes precipitates, or the solution becomes slightly hazy. We typically never allowed dialysis to continue longer than three hours for this reason. This behavior was observed with or without glycerol, detergent (0.01% Triton X-100) or ATP. We observed that MreB is particularly prone to precipitation in phosphate buffers so these were avoided. Some further enrichment can be achieved with a Q column (not shown). We also found that MreB fails to bind to Heparin or Blue sepharose columns. Note that while the listed protocol is the most common procedure, we experimented with modifications to this procedure and obtained similar results. In particular, the removal of glycerol from the S-buffer does not affect the outcome in any way that we were able to measure.

Denaturing purification of recombinant MreB Cells were grown and lysed as described above. The clarified lysate was precipitated with 40% Ammonium Sulfate and pelleted. The supernatant was discarded and the pellet was resuspended with an equal volume of TPU buffer (containing 8M

P a g e | 185

urea). We found that adding urea to an AS pellet allowed for much more efficient and complete denaturation than adding urea to a pellet of whole cells. To equilibrate the sample further, we then dialyzed the protein against the same buffer (TPU) for 3 hours (12-14,000 MWCO). After dialysis the sample was spun for 15min at 23,000xg or filtered to remove particulates. Then, it was loaded onto the Ni-NTA column equilibrated in TPU buffer. The column was washed with TPU wash buffer, and then eluted with 100% TPU elute buffer. MreB is quite pure after this step, but it can actually still contain residual 260nm absorbance. To completely remove the nucleotide, switch buffers, begin removing the urea, and dilute the sample, we ran a Superdex S200 gel filtration column in 6M urea in TK buffer. MreB elutes ~59ml and has a 260/280nm ratio of 0.65. If the sample is incompletely denatured, some MreB can be seen to flow through at the void volume (47ml). This fraction also has a higher 260/280 ratio. EDTA was found to enhance nucleotide removal. If this gel filtration step was removed, the protein behaved more similarly to the natively purified protein (soluble in low salt buffers), probably indicating that contaminating nucleotide is still present. To refold the protein, we used a series of dialysis steps at 0C. We mixed TK buffer containing 6M urea with TKG buffer (lacking urea) at an appropriate ratio to bring the final concentration of urea to 4M or 2M. We then dialyzed the MreB against the 4M urea buffer overnight. If nucleotide was added during refolding, it was added at this step and included in the remaining buffers. The next day the sample was transferred to the 2M urea buffer for 4 hours, followed by the 0M urea TKG buffer for another 6 hours. The buffer was then changed and the sample was dialyzed overnight in fresh TKG. Note that the concentration of glycerol also gradually increases during this method (0%, 2%, 5%, 10%), which has the effect of modestly increasing the concentration of MreB during dialysis. The protein was then aliquoted, frozen in liquid nitrogen, and stored at -80C. In some cases, we used 200mM NaCl (TN) instead of 100mM KCl (TK). Glycerol is not required for proper folding and did not change the results, although we typically included it as a cryoprotectant and to concentrate the protein. We could refold the protein in the absence of salt and nucleotide without observing significant precipitation, but in this case it still formed polymers that were similar to those refolded in the presence of salt.

186 | P a g e

CHAPTER 4

SDS-PAGE Standard methods for denaturing protein electrophoresis were used. SDS-PAGE sample buffer was added and the samples were boiled prior to loading in the gel. For MreB, we used 10-12% acrylamide gels.

Measurement of protein concentration and concentrating purified MreB Protein concentration was measured with a Bradford assay in a microtiter plate according to the protocol in the BioRad manual using a standard curve of Bovine γ- Globulin. We were not able to concentrate MreB using typical spin columns (regenerated cellulose), as a significant amount of MreB was found to irreversibly bind these surfaces. We used either anion exchange chromatography or dialysis against glycerol to at least modestly increase the concentration of MreB.

Thrombin digestion Thrombin does not cleave purified MreB, probably because the cleavage site is occluded in the MreB polymers. However, we were able to cleave the His-tag from MreB when it was still bound to the Ni-NTA resin. This experiment is not trivial, however, as MreB will precipitate at high concentration when no longer bound to the resin. For this experiment, we performed the Ni affinity chromatography in batch, rather than on the FPLC. The cells and lysate were obtained as in the native purification described above. Ni resin (2ml slurry) was added to clarified lysate (80ml, from a 4L culture of cells), and the mixture was incubated with mild agitation at 4C, 50min. The mixture was then added to a disposable column, to separate the resin and bound components from the rest of the materials in the lysate. The resin was then washed with ~10ml of 0.5X Ni Buffer A. The final volume of resin and buffer was brought to a volume of ~4ml, and Thrombin (20U, Sigma) was added. The column was then sealed, and then placed on an end-over at room temperature (~20C) to allow digest to occur for 2-6 hours. After digestion, the sample was allowed to flow through the column, and additional buffer was added to wash the resin. At this point, cleaved MreB should flow through, leaving uncleaved MreB bound to the Ni-resin. If the concentration of MreB is too high, this flow through sample will be

P a g e | 187

partially precipitated. In our experience much of the cleaved MreB remains bound to the Ni-resin, however, either because it is associating with the uncleaved MreB or because it is precipitating on the column. To remove the cleaved MreB still bound to the resin, we added 8M urea in Ni Buffer A and collected the flow through. The urea was able to release much of the remaining cleaved MreB, while leaving uncleaved MreB bound to the resin. We continued with both the soluble, flow through cleaved MreB (natively purified), and the denatured cleaved MreB that was released from the beads only with urea. We ran the natively purified sample over Superdex S200 in 20mM TEA, pH 8, 5mM KCl, 1mM EDTA, 1mM DTT and found that the sample eluted at 47ml (void volume, data not shown). With the denatured MreB, we first continued with a variation of the standard denaturing purification protocol by gel filtering the sample in 6M urea in 20mM Tris, ph 8 (4C), 10mM KCl, 1mM EDTA, 10% glycerol. We then refolded the protein with slow dialysis into a buffer containing progressively lower concentrations of urea (4M, then 2M, then 0M urea in 20mM TEA, pH 8, 5mM KCl, 1mM EDTA, 1mM DTT). We then performed gel filtration chromatography as listed for the cleaved natively purified MreB. Cleaved refolded MreB was found in the void volume in the absence of nucleotide (47ml).

Purification of His-MreB from Caulobacter The Caulobacter strain bearing his-mreB at the xylX locus was grown overnight at 30C in PYE+Kanamycin (5µg/ml) in a 10ml starter culture. This culture was then used to inoculate 2L of pre-warmed PYE, and the culture was grown for another 5.5hrs. Xylose was added to a final concentration of 0.3% at an OD (660nm) of 0.075, and the culture was allowed to grow while inducing His-MreB expression. At an OD=0.46, the cells were harvested, washed, and lysed as described for the recombinant expression in E.coli. Lysates were subjected to the Ni-affinity chromatography, following the same procedure as that of the Native purification of recombinant MreB. Coomassie stained SDS-PAGE and Western blots were used to follow MreB. After the Ni column, we used anion exchange chromatography with a HiTrap Q column. Samples were equilibrated in Q buffer A, loaded and then eluted from the Q column with a gradient of 0-100% Q Buffer

188 | P a g e

CHAPTER 4

B over 10CV. MreB-containing fractions were gel filtered with Superdex S200 in Q buffer A. MreB eluted in the void volume of this run, even at a low concentration, after being purified from Caulobacter. In this prep, only His-tagged MreB was retrieved from Caulobacter extract. We could distinguish endogenous MreB from His-tagged MreB with Western blots of the extract and determined that only His-tagged MreB binds to the Ni-resin. For Western blot detection of MreB, we used rabbitt anti-serum raised against purified Caulobacter MreB (Josman) as the primary and HRP-conjugated anti-rabbit as the secondary. The nitrocellulose containing the protein sample from the gel was incubated in a solution of 2% milk in TBST buffer for 30 min (or overnight at 4C). The blot was then incubated in a solution of primary antibody (diluted 1:10,000) in TBST for 30 min at room temperature and then washed 3x10min in TBST. The blot was then incubated in a solution of secondary antibody (diluted 1:10,000) in TBST for 30 min at room temperature and then washed 3x10min in TBST. HRP reagent was applied according to the manufactor’s suggestions (Sigma) and the signal was captured on audoradiography film (Denville Scientific).

Native purification of GFP-MreB GFP-MreB was purified according to the same native protocol described above for unlabeled MreB. Usually we did not continue with further chromatography after the Ni column. GFP-MreB eluted in the void volume of a Superdex S200 gel filtration column in either G-buffer1 or Ni Buffer A (data not shown).

Centrifugation We used ―low-speed‖ and ―high-speed‖ centrifugations, which correspond to 16,000xg (standard epindorf microcentrifuge) for 10-20min and 300,000xg (TLA100) for 15min, respectively. For the supernatant, a stock solution of SDS-PAGE sample buffer was added. For the pellet, 5X SDS-PAGE sample buffer was added directly to the tube and allowed to incubate for ~10min at room temperature. Then, a pipet tip was used to aggressively scrape the sides of the tube, and a combination of vortexing and pipetting up and down was used to resuspend the pellet. The final concentration of SDS-PAGE

P a g e | 189

sample buffer and the total volume were adjusted to match that of the supernatant. All samples were then boiled and run in equal amounts on SDS-PAGE.

Electron Microscopy To prepare the EM grids, we first cleaned 300-400mesh Copper grids (EMS) with a brief wash in Acetone and then allowed them to dry on chromatography paper. A thin layer of formvar was prepared by dipping a glass microscope slide into a solution of Formvar and then quickly removing it and allowing it to air dry. The formvar was released from the slide by scoring the edges and carefully floating the layer off into a dish of clean distilled water. The Cu grids were dropped onto this layer of formvar and then the entire layer was peeled off the water by quickly adsorbing it to a piece of parafilm. After the residual water dried, we coated these Formvar-coated Cu grids with a thin layer of Carbon in a metal evaporator under high vacuum. Grids were stored under desiccation at room temperature for no longer than six months. To glow discharge the surface, we transferred the desired number of grids to a piece of chromatography paper in a glass petri dish. This dish was placed in a metal evaporator and discharged under vacuum in either air or Argon. To discharge in Amylamine, a drop of solution was placed on a piece of chromatography paper and placed next to the dish containing the EM grids in the metal evaporator. Grids were discharged for ~30 sec to 1min immediately prior to sample preparation. To coat EM grids with poly-L-lysine, a premade solution of PLL at 10mg/ml (Sigma) was added to a glow discharged grid and allowed to bind for 5 min. Excess sample was then blotted away and then the grid was allowed to dry completely (2-16 hours) before sample addition. For negative staining, we prepared solutions of 1% Uranyl Acetate, PTA or Ammonium Molybdate. The pH of PTA and Ammonium Molybdate was 7.0 or 8.0, respectively. These solutions were syringe filtered through 0.2µm filters immediately prior to use. Protein (3µl) was added to the prepared grid, allowed to bind for 30 sec to 1min, and then blotted away using Whatman chromatography paper. Then, a drop of stain was added to the grid and then almost immediately blotted away. Grids were then dried completely (2-16 hrs). We experimented with varying the amount of time that the sample

190 | P a g e

CHAPTER 4 was allowed to adhere to the grid prior to staining and found no significant difference in the amount of protein that was visible. Imaging was performed on a JEOL JEM-1230 transmission electron microscope at 80-120kV.

Dynamic light scattering For the dynamic light scattering experiments presented in Figure 10, we used a DynaPro 801 molecular sizing instrument (Protein Solutions, Charlottsville, VA) equipped with a temperature controlled microsampler and controlled by Dynamics V6 software. Protein samples were added to 12µl sample cuvettes (cat #003279) pre- equilibrated in the sample chamber at the desired temperature. For each acquisition, scattering was measured every 1sec for 10 sec. For each measurement, we averaged 20- 30 acquisitions. The standard ―optimal‖ resolution setting was used, and samples were modeled as spheres or coils (neither considerably changed the output). The solvent setting was chosen to best match that of our sample (water, 10% glycerol or 1% NaCl, though this parameter did not affect the result considerably). We reported the radius as a function of % mass, though the % intensity readings were similar (data not shown). When we observed increases in the sizes of the polymers, we also observed an increase in the overall total intensity and polydispersion (data not shown). To account for this increase in intensity, it was sometimes necessary to lower the laser power or dilute the sample.

Right angle light scattering To measure light scattering at 90˚, we used a Fluorolog-3 spectrofluorometer (Jobin Yvon Horiba) controlled by DataMax v. 2.20 software. Excitation and emission wavelengths were set to 400nm. Slit widths were set to 1nm. At the start of the experiment, the voltage of the PMT was adjusted so that the sample with the highest concentration of MreB would produce a scattering signal that was near the maximum of the linear range of the instrument. To prepare the MreB samples, stock solutions of MreB (native or refolded) were dialyzed against TEAK1 buffer overnight. The concentration was measured after dialysis with a Bradford Assay. Dilutions into TEAK1 buffer were made to produce final concentrations of: 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.8, 1.0, 1.2, 1.5, 2, 2.5, 3.0, 3.5, and 4µM.

P a g e | 191

Samples were gently sonicated in a bath sonicator (3x10 sec pulses), and then allowed to equilibrate overnight (18-24 hr) at room temperature in the dark. Samples were prepared in triplicate. For each sample, we took the median value of the amount of scattering measured every 5sec for a period of 60sec (1sec integration). The mean and standard deviation of the replicates for each concentration was calculated, normalized to the highest value for the experiment, and then plotted as a function of concentration.

Chemical coupling of MreB to Dylight A stock solution of Dylight-488 NHS-ester (amine-reactive fluorophore, ThermoScientific/Pierce) was prepared in dimethyl formamide (DMF) and then serially diluted into CK buffer. The concentration of dye was determined by measuring its absorbance at 493nm, using an extinction coefficient of 70,000M-1cm-1. Purified MreB was dialyzed into CK buffer. MreB and dye were mixed together and incubated at 4C protected from light. Dye was removed with several rounds of dialysis into TEAK2 buffer (contains DTT). The efficiency of labeling was measured by comparing the absorbance at 493 to that at 280nm (see Dylight manual for more information). The labeling efficiency could be tuned by varying the concentration of protein, ratio of Dylight:MreB, and incubation time. With 0.1mg/ml MreB, a 50X molar excess of Dylight, and an overnight incubation we achieved 31% labeling. In another experiment, we used 0.4mg/ml MreB, 20X molar excess of Dylight, and a 1hr or 4 hour incubation, we achieved 50% or 70% labeling.

Preparation of DNA-coated beads PCR was used to amplify 1Kb regions of the Caulobacter chromosome. An EcoRI site was added to both forward and reverse primers. DNA was purified with Qiagen spin columns and then digested with EcoRI (Invitrogen). The reaction was cleaned again with another Qiagen spin column and eluted in water. The Klenow fragment (Invitrogen) was used to fill in the overhangs created by the EcoRI site with dNTPs containing biotin-dATP. Dynabeads M270 Streptavidin-coupled magnetic polystyrene beads (Dynal Biotech) were washed twice in BW buffer and then mixed with biotinylated DNA overnight at room temperature. After binding, beads were washed with

192 | P a g e

CHAPTER 4

BW buffer. Binding of DNA to the beads was evaluated with Ethidium bromide staining.

Imaging of Dylight-MreB To image Dylight-MreB under various solution conditions, we used a stock solution of MreB at 10µM in TEAK2 buffer. Mixtures were prepared with MreB at a final concentration of 1µM. These were allowed to incubate for ~10-20 min at room temperature and then imaged on glass. We used plasmid DNA of various sizes: pUC19, pET28a, pXGFP4C1, and pMR10. EcoRI was used to linearize the plasmids. The final concentration of the linearized DNA in the mixtures was 30-100µM. For the polymerized amino acids, we prepared 1% solutions in 10mM Tris, pH 8 (~10mg/ml) and then diluted these solutions so that the final concentration in the mixture with MreB was ~0.001%. We used poly-L-lysine (Sigma P7890, avg MW 15,000), poly-L-glutamate (Sigma P4886, avg MW 34-64,000), or poly-L-proline (Sigma P2129, avg MW 15-26,000).

Light microscopy For the imaging, we used a upright fluorescence microscope (Zeiss, Thornwood, NY) equipped with a plan-apo 100x phase 3 objective lens, conventional epi-fluorescence filter set and a 512 x 512 pixel back illuminated frame transfer camera (Phometrics, Tucson, AZ)

P a g e | 193

194 | P a g e

Chapter 5 Biochemical fractionation of Caulobacter cell extract to identify regulators of MreB assembly

Natalie Dye, Julie Theriot, and Lucy Shapiro

P a g e | 195

ABSTRACT

Cytoskeletal organization and dynamic behavior are regulated by numerous interacting proteins. For the bacterial actin homolog MreB, there are no known regulators of polymerization or nucleotide-dependent binding partners. In this work, we develop a biochemical assay for MreB disassembly. Our assay uses high speed centrifugation to separate MreB polymers from small oligomers and monomers, followed by a quantitative Western blot to measure the amount of MreB in the supernatant. With this assay, we show that Caulobacter extract contains a heat and protease-sensitive activity that destabilizes MreB polymers, releasing more MreB into the high speed supernatant than buffer alone. We partially fractionated the destabilizing activity with various chromatographic techniques. While we were unable to identify the precise molecular component that is responsible for this activity, we provide evidence to suggest that this activity belongs to a DNA- or RNA-binding protein. We predict that this approach will be useful in the future for identifying novel effectors of MreB assembly in vitro.

196 | P a g e

CHAPTER 5

INTRODUCTION

Like their eukaryotic counterparts, prokaryotic cytoskeletal homologs are dynamic in cells. The localization of FtsZ is cell cycle regulated [53, 158]. Much of the cellular FtsZ pool accumulates at midcell just prior to division. The FtsZ molecules that are not recruited to midcell exhibit rapid movement and oscillate from pole to pole in helical-like intermediates [167]. Photobleaching of YFP-FtsZ in E.coli and B.subtilis indicates that the filaments of FtsZ at the division plane are capable of rapid recovery (a half time as short as 8-9s, [166, 264]). Mutants of FtsZ that are deficient in GTP hydrolysis exhibit slower dynamics and form aberrant structures in vivo [166, 167, 264- 266], indicating that the nucleotide cycle is important for regulating the behavior of FtsZ, as it is for actin and microtubules. In addition, the localization and dynamic behavior of FtsZ is sensitive to developmental and environmental signals [168, 267]. Importantly, there are now many known interacting partners for FtsZ, including proteins that inhibit FtsZ polymerization by stimulating GTP hydrolysis (for example, MinC and MipZ) and proteins that stabilize FtsZ polymers and promote lateral interactions (for example ZipZ, ZapA, and EzrA) [49, 52, 53, 92]. As in eukaryotic cells, the presence and activities of these interacting partners contribute to the cellular localization, organization and function of FtsZ in prokaryotic cells. As discussed in previous chapters, the localization of MreB is also cell cycle regulated in Caulobacter and potentially also E.coli [39, 117, 268]. As with FtsZ, rapid movement of fluorescent MreB has been observed in B.subtilis, and photobleaching studies in E.coli and B.subtilis indicate that the helical structures are capable of reasonably rapid turnover [177, 178, 227]. We and others have shown that mutations in MreB can affect these dynamic behaviors and function (Chapter 3). In addition, the movement of single molecules of YFP-MreB in Caulobacter has been characterized [228]. In this experiment, two types of motion were described: rapid diffusion and slow, directed movement over distances of ~300-400nm. It has been suggested that this slow movement corresponds to subunits of MreB polymers treadmilling through stationary filaments (See Chapter 6). Lastly, MreB that is expressed in S.pombe forms linear filamentous structures that do not at all correspond to the organization of MreB in

P a g e | 197

bacterial cells, indicating that the cellular environment and components of the bacterial cell are critical for the organization and dynamic behavior of MreB [269]. As yet, no regulators of MreB polymerization have been identified, though many different avenues have been pursued. The mreB gene is found in many bacterial species in an operon with mreC, and mreD. We showed in Chapter 2 that MreC is not required for the proper localization or dynamic behavior of MreB in Caulobacter, though these two genes have similar null phenotypes and therefore are likely to interact in the same functional cell shape determination pathway. Others have reported similar findings in E.coli [231]. Furthermore, since MreC is largely periplasmic, it is not likely to play a critical role in regulating MreB (which lies in the cytoplasm) [39, 143]. MreD is also not a likely regulator of MreB assembly, since it is an integral without any soluble cytoplasmic domains. Bacterial two-hybrid, cell biology, and biochemical experiments have shown that MreB can interact with many of the enzymes required for the synthesis of peptidoglycan precursors [140, 205-207], but again these proteins are effectors of the cell shape pathway and not likely to directly regulate MreB assembly. The gene rodZ was identified in a genetic screen for morphological mutants [223, 224, 270]. It has now been shown that RodZ and MreB interact directly and that RodZ is required to link MreB filaments to the membrane [223, 224, 270]. Structural and biochemical studies have indicated that the interaction between RodZ and MreB is not nucleotide-dependent, however, and that RodZ can bind to both monomer and filament forms [214]. Lastly, co-immunoprecipitation, co-pelleting, and affinity chromatography experiments have been performed with MreB but have not identified any probable candidate regulators (K.Amman, Z.Gitai, personal communication, and unpublished observations). In this work, we chose a biochemical approach to identifying potential regulators of MreB assembly. Rather than assaying for binding, we chose to assay for activity. To measure changes in the assembly state of MreB, we used high speed centrifugation to separate small and large forms of MreB and then performed a quantitative Western blot on the supernatant. With this assay, we show that Caulobacter extract has a destabilizing activity that can be enriched with biochemical fractionation. Rather than identifying binding partners, this approach has the potential to identify functional interacting partners

198 | P a g e

CHAPTER 5 of MreB and could therefore be used to find unexpected proteins that regulate MreB assembly in vitro.

RESULTS

Purified MreB responds to Caulobacter extract We have previously shown that purified recombinant MreB polymers, particularly those that have been purified with a denaturing procedure, reliably pellet with high speed centrifugation in a wide variety of solution conditions (See Chapter 4). Thus, we reasoned that we could use this same centrifugation procedure to assess whether the assembly state of purified MreB could be modified with the addition of exogenous Caulobacter extract in vitro (Figure 1A). We used Ammonium sulfate precipitation to fractionate a crude Caulobacter lysate containing a high concentration of protein (~50- 100mg/ml, See Methods). These fractions were equilibrated in a medium ionic strength buffer (See Methods) and then added to an excess of purified MreB. After a 15 min incubation at 4C, the reactions were subjected to a high speed centrifugation (>300,000xg, 15min, 4C). Supernatants were then assayed for the presence of MreB with a quantitative Western blot. In the absence of extract, very little MreB can be detected in the supernatant, as expected from the results presented in Chapter 4. In the presence of particular Ammonium Sulfate (AS) fractions, however, we detected a significant increase in the amount of MreB in the supernatant over that in buffer alone (Figure 1B). This effect could be abolished by first boiling the extract or treating the extract with proteases prior to the incubation with MreB polymers, indicating that this activity is protein- dependent (Figure 1B and data not shown). These data are consistent with the hypothesis that Caulobacter extracts contain one or more proteins that can promote the disassembly of preassembled, purified MreB polymers. We will hereafter refer to this activity as the ―MDF activity,‖ for MreB Disassembly Factor. We observed that MDF activity peaks slightly in the 0-25% AS fraction (contains proteins that are soluble up to 25% AS), decreases almost entirely in the 35-45% AS fraction (contains proteins that precipitate between 35 and 45% AS), and reaches a maximum in the 55-70% Ammonium Sulfate fraction (soluble up to 55% ammonium

P a g e | 199

sulfate). This pattern was replicated several times, on different days and using different, though similarly prepared, Caulobacter extracts. We also observed qualitatively similar patterns with and without the addition of ADP and ATP (data not shown). Given that AS precipitation is a rather crude fractionation method, we were somewhat surprised to see two peaks in the activity. This behavior could be the result of two independent proteins with MDF activity that are well separated with AS precipitation or the presence of a separate activity in fraction 35-45% that is antagonistic to the MDF activity. To distinguish between these two possibilities, we performed a mixing experiment. If it were true that there were two proteins with MDF activity (one in 0-25%, one in 55-70% and none in 35-55%), a 1:1 mixture of the 55-70% and 35-45% fractions would result only in a slight decrease in activity (due to dilution). Instead, we observed that MDF activity was completely abolished (Figure 1C). This drop in activity was not due solely to the dilution of the fraction, as it retained partial activity when diluted 2-fold with buffer. This result suggests that there are both destabilizing and stabilizing activities present in the Caulobacter extract. Because a stabilizing activity decreases the already small amount of soluble purified MreB, the pelleting assay we have used here is not ideal for following this particular activity. Thus, we chose to focus on the destabilizing activity that is present in the 55-70% AS fraction.

The identification of prominent proteins in an active MDF fraction (Scheme 1) To further fractionate the MDF activity, we applied the 55-70% AS fraction to a DEAE ion exchange column and eluted the bound proteins in two steps (250mM NaCl wash, followed by a gradient up to 1M NaCl). By Coomassie stain, we detected many proteins in the flow through and both elutions, but the MDF activity was clearly enriched in the late-eluting fraction (Figure 2C). This late eluting fraction also had a considerable amount of contaminating nucleic acids, as indicated by a high absorbance at 260nm (data not shown). Given that MreB can bind and aggregate DNA (Chapter 4), we thought it important to try to remove this material from the extract fraction. To achieve this, we added purified RnaseA, DnaseI and millimolar concentrations of divalent cation (as well as protease inhibitors) to the active DEAE elute fraction. With a simple Ethidium Bromide spot test, we detected a considerable loss of nucleic acid after nuclease

200 | P a g e

CHAPTER 5 treatment, although not all was removed (Figure 2D).

FIGURE 1

P a g e | 201

FIGURE 2

202 | P a g e

CHAPTER 5

We then applied the nuclease-treated fraction to another ion exchange column at a slightly lower pH (Q, a stronger anion exchanger than DEAE) and eluted with a salt gradient. After the Resource Q, the MDF activity was found to be limited to two fractions (Figure 2F). This first fractionation scheme—AS precipitation, DEAE at pH 7.5, nuclease treatment, and Q at pH 7—produces a considerable enrichment of MDF activity (Figure 2B). After analyzing the Coomassie stained gel (Figure 2F), we chose four of the most prominent bands that appeared to be enriched in the MDF active fractions and extracted these proteins from the gel. These proteins were identified by mass spectrometry to be (in order of descending molecular weight): RNA polymerase β and β’-subunits, DNA Topoisomerase I, RNA polymerase α-subunit, and Transaldolase (Table 1). The most abundant component, which is at least 10X more concentrated than any of the remaining bands, was found to be Transaldolase.

A modified fractionation scheme separates Transaldolase and MDF activities (Scheme 2) Transaldolase was the most prominent band in our active fractions at the end of Scheme 1. This protein catalyzes the conversion of the 6-carbon sugar fructose-6- phosphate and the 4-carbon sugar erythrose-4-phosphate to a 7-carbon sugar sedoheptulose-7-phosphate and a 3-carbon sugar glyceraldehyde-3-phosphate in the pentose-phosphate pathway. Because it has a known metabolic enzymatic activity, it is not a likely candidate for a protein that can regulate MreB in vivo. Nonetheless, we could not deny that was enriched in fractions of Caulobacter extract with MDF activity in vitro. In addition, aldolase (related distantly to transaldolase) is a bona fide actin-binding

P a g e | 203

protein [271-275]. In Apicomplexan parasites (i.e. Plasmodium and Toxoplasma), aldolase is required in vivo to link actin filaments to a membrane-bound force generating adhesion complex (TRAP) during cell motility [276-278]. The binding site between TRAP and aldolase is also present in WASp family members, which are actin nucleators in mammalian cells [277]. Thus, we thought it possible that a protein with metabolic enzymatic activity and function could also ―moonlight‖ as a regulator of MreB. We repeated the fractionation of the MDF activity with slight modifications to the previous procedure. This method will be referred to as Scheme 2 (Figure 3). This method is the same as Scheme 1 until the nuclease treatment. Here, after dialyzing the active fraction with nucleases, we introduced a gel filtration step, followed by the Q column at a lower pH than used in Scheme 1 (pH 6 rather than pH 7). During the course of this fractionation experiment, we assayed fractions for both MDF activity and the enzymatic activity of Transaldolase (See Methods). We found that the Transaldolase activity follows that of the MDF activity through the first two steps (AS and DEAE, data not shown). After the gel filtration step, however, we noticed a slight difference between the two peaks of specific activity (Figure 3C). Whereas the specific activity of Transaldolase peaks at fractions 16-17, that of MDF peaks around fraction 19. Furthermore, fraction 11 has measurable activity for MDF but not Transaldolase. We continued the procedure with fractions 15-17. After the dialysis into a pH 6 buffer and elution from the Q column, no MDF activity was detectable in any of the eluted fractions or the flow through, but considerable Transaldolase activity was present in fraction 39 (Figure 3D and data not shown). Note that a considerable amount of 260nm absorbance was separated from the peak of MDF activity after the gel filtration column (Figure 3B), suggesting that MDF activity does not require nucleic acid. The peak of Transaldolase elution indicates that this small 23kD protein must exist in a large complex in vitro, as it elutes fairly early from this column (~380,000kDa, as estimated from calibration runs with known standards, data not shown). Other aldolase family members are known to form large (10- 12mer) homocomplexes in vitro [279]. The MDF activity also elutes fairly early, indicating that it is also either a complex of proteins or one protein that oligomerizes in vitro.

204 | P a g e

CHAPTER 5

FIGURE 3

P a g e | 205

Transaldolase and MDF activity The data from Scheme 2 suggest that the activities for the Transaldolase reaction and MDF can be separated from one another and that Transaldolase cannot be the sole source of the MDF activity. Nonetheless, it seemed possible that the MDF activity was the result of the collaborative effort of Transaldolase and one or more proteins, which are lost after the dialysis into pH 6. With the purified Transaldolase eluted from the Q column (fraction 39), we investigated its interaction with MreB. We found that MreB actually pelleted slightly more in the presence of Transaldolase than in its absence (Figure 4A), suggesting that Transaldolase and MreB could physically interact but with the result of stabilizing rather than destabilizing MreB polymers. This activity was observed with less than 20 min of incubation time and was independent of added magnesium, ADP or ATP. It has been shown that the enzymatic activity of aldolase is inhibited by actin [272]. We tested the enzymatic activity of Transaldolase with and without MreB and found it to be unchanged (Figure 4B). MreB was also not found to significantly change the ability of Transaldolase to pellet in a low or high speed spin: the activity in the supernatents of these spins was unchanged by the presence of MreB (Figure 4B). Thus, the presence of Transaldolase slightly changes the behavior of MreB but the two proteins are probably not tightly physically bound in a stoichiometric ratio. Lastly, we created a transaldolase deletion strain to definitively assess the importance of Transaldolase in the MDF activity. This strain is viable but grows with a significantly longer doubling time (data not shown). The enzymatic activity of Transaldolase is nearly completely lost from the extract (Figure 4Ci). If Transaldolase is required for MDF activity, this activity should also be diminished or at least altered in extracts made from a knockout strain lacking Transaldolase. We found the opposite to be true: MDF activity still fractionates with the 55-70% AS pellet, with or without the addition of exogenous purified Transaldolase (Figure 4Cii). Therefore, we conclude that Transaldolase is not required for MDF activity.

206 | P a g e

CHAPTER 5

FIGURE 4

Nucleic acid affects the fractionation of, but is not required for, MDF activity Having verified that Transaldolase was a red-herring, we next investigated other chromatographic methods of fractionating the late-eluting fraction of the DEAE column (Figure 5). In this next round of fractionation, we focused on purifying the activity away from both contaminating protein and nucleic acid. In Scheme 3 (Figure 5A), we first performed the AS precipitation and DEAE columns as in Scheme 1. Then, we tested the ability of the late-eluting fraction to be separated by hydrophobic interaction columns (HIC, butyl or high substituted phenyl), S-sepharose, Blue-sepharose, or Hydroxyapatite. We found that the S, Blue, and Hydroxyapitite columns were unable to significantly increase the specific activity of the MDF activity or remove nucleic acid (data not

P a g e | 207

shown). The HIC columns, however, were more promising. The activity was partially bound by the Butyl HIC (equal amounts in FT and fraction 8, Figure 5B). While this column did not increase the specific activity, much of the nucleic acid contamination was removed with this step (data not shown). MDF activity also bound well to the Phenyl HIC and appeared to elute earlier than most of the protein, resulting in a large increase in specific activity (Figure 5C). However, these eluted fractions still had a large amount of nucleic acid (data not shown).

FIGURE 5

208 | P a g e

CHAPTER 5

After Scheme 5, we still lacked a method for reducing nucleic acid contamination while increasing specific activity. While the Butyl column eliminates nucleic acid, not all of the MDF binds. Conversely, while the Phenyl column increases specific activity, nucleic acids remain. In Scheme 1 we treated the fractions with nucleases, but a considerable amount of nucleic acid remained. The MDF activity was actually found to increase slightly after nuclease treatment (data not shown), again indicating that nucleic acid is not required for MDF activity; its presence may actually decrease activity. Additionally, when we attempted to remove more of the nucleic acid prior to the DEAE step (with addition of nucleases and precipitation), the MDF activity was found to elute in the first salt step (250mM NaCl, data not shown). These results suggest that nucleic acid affects the fractionation of the MDF activity, perhaps because the protein responsible for the MDF activity is also a DNA- or RNA-binding protein. We developed yet another fractionation protocol that is orthogonal to the original DEAE scheme and fully separates DNA from protein (Scheme 4, Figure 6). In this method, we still used the AS precipitation, as it provides a significant enrichment of MDF activity. Rather than continuing to the DEAE column, we next increased the salt concentration to 2M NaCl, which should dissociate most DNA- and RNA-binding proteins from nucleic acid, and then precipitated the nucleic acid with PEG8000. Under these conditions most of the protein is soluble, while the nucleic acid is insoluble. After a centrifugation step to separate the two components, we took the protein fraction and performed Butyl HIC. The combination of these two steps fully removed nucleic acid contamination and the MDF activity was preserved (Figure 6C). We attempted to further fractionate this activity with a Q column and achieved a small increase in specific activity (Figure 6D). This fraction still has many protein components, however, and needs further fractionation before the identity of the MDF activity can be identified.

P a g e | 209

FIGURE 6

210 | P a g e

CHAPTER 5

DISCUSSION

In this work, we have shown that Caulobacter extract is capable of affecting the ability of purified MreB to pellet in a high speed centrifugation. Using a quantitative Western blot to measure MreB in the supernatant, we followed the ability of fractionated extract to destabilize MreB (an activity we named ―MDF activity‖). This activity was enriched by anion exchange columns and two hydrophobic interaction columns (Butyl and high substituted Phenyl). The activity eluted relatively early in size exclusion chromatography (an estimated molecular weight of ~300,000kDa) and was not enriched with cation exchange, hydroxyapitite, or blue sepharose. Lastly, we found that nucleic acid was not required for MDF activity but affected the fractionation of the activity, perhaps indicating that the activity is due to a DNA- or RNA-binding protein.

What is the inhibitor? Unfortunately, we were not able to identify the precise molecular component or components that are responsible for the MDF activity. The major component of the active fraction at the end of Scheme 1 was found to be Transaldolase. Additional experiments confirmed that Transaldolase is not absolutely required for this activity, however. We also think it is unlikely that the components of the RNA polymerase (bands labeled 1 and 3 in Figure 2C) are solely responsible for the MDF activity for two primary reasons: the first is that these bands appear to peak slightly earlier than the MDF activity in the gel filtration of Scheme 2 (Figure 3); the second is that these bands are still present in the eluted fractions of the Resource Q following the gel filtration, whereas there is no MDF activity in these fractions (Figure 3D). These bands are also not visible in the fractions eluted from the Phenyl column (Figure 5D); although, it is difficult to see all but the most prominent components of these fractions since the protein concentration was fairly low. Nevertheless, it has been shown recently that all three of these RNAP components can be co-immunoprecipitated from E.coli cell lysates, suggesting that there is a physical connection between these proteins [280]. As such, it may be worth directly looking at the interaction between these proteins in Caulobacter extract. Recently, it was also shown that E.coli MreB can interact with the ParC subunit of

P a g e | 211

Topoisomerase IV [118]. We did not identify this specific protein in our preps; however band 2 is a related protein, Topoisomerase I (labeled 2). This band is not particularly enriched in fractions 8 and 9 of the Q (Figure 2F). It is possible, however, that there are actually two proteins around the same molecular weight that cannot be resolved well with this high percentage acrylamide gel and have very different elution profiles from the Q column (one that peaks in fraction 5). A band at this molecular weight (97.9kD) elutes in a broad peak from the gel filtration, from fraction 7-27, which is consistent with the elution of the MDF activity (although here again it is difficult to tell the difference between what looks like two proteins around the same molecular weight, Figure 3D). Also, a band of this size is largely missing from the Q elution after the gel filtration (Figure 3D). Notably, the pI for this protein is predicted to be 8.68. Therefore, it is expected to be positively charged at the pH used for the DEAE column (which should only bind negatively-charged proteins). The fact that it bound to this column and eluted at a very high salt concentration is probably due to its ability to bind DNA. The pIs for the RNAP subunits and transaldolase are all very low (5.25, 5.4, 4.97, and 4.53, respectively), and therefore they are expected to bind well to DEAE. Intriguingly, we found that the fractionation of the MDF activity was also affected by the presence of DNA, in that the activity eluted earlier from the DEAE when nucleases were used to clear the sample of nucleic acid prior to the column. Therefore, it seems plausible that DNA Topoisomerase I could be the source of the MDF activity. Of course, it is also possible that none of these proteins that were identified are the source of the activity or that a complex of these or other proteins is required to produce MDF activity. A related in vitro activity-based approach has been used to identify factors responsible for the rapid depolymerization of Listeria comet tails [281]. In that work, it was shown that a complex of Aip1, coronin, and cofilin is required for full activity [281]. Additional work will be required to confirm the protein or complex of proteins that produce the MDF activity.

Suggestions for future work In future work, it may be possible to directly exploit the fact that the MDF activity seems to follow nucleic acid by performing DNA or RNA affinity chromatography [282,

212 | P a g e

CHAPTER 5

283]. Notably, there are inhibitors of FtsZ polymerization that nonspecifically localize to the nucleoid: Noc in B.subtilis and SlmA in E.coli [72, 73]. In addition, there does appear to be some connection between MreB and the nucleoid in vivo. In B.subtilis and E.coli, MreB is localized towards the center of the cell and away from the poles, which matches the localization of nucleoid [59, 116, 121, 178, 268]. In B.subtilis, GFP-MreB is no longer dynamic after the depletion of ParE, a subunit of Topoisomerase IV [178]. This depletion also results in the formation of anucleate cells, and GFP-MreB no longer appears to form filaments in these cells lacking DNA [178]. In the future, it may be advantageous to combine different techniques to determine the source of the MDF activity and potential regulators. In particular, we showed in the last chapter that endogenous MreB in Caulobacter extract can undergo one round of cycling (pellet-supernatent-pellet, Chapter 4, Figure 1). There does appear to be some similarity between the patterns of bands that follow MreB through this cycling and those that are present in MDF active fractions. It may be worth identifying all of the components of these samples and looking for ones that appear in both. Lastly, it would be worth developing alternative biochemical activity assays. In this work, we chose a conceptually very simple assay for MreB disassembly. The pelleting experiment is reliable, quick and non-disruptive, but the use of the quantitative Western blot to measure the activity is problematic. First, it is slow. We optimized the procedure as best we could but still could not do the assay any faster than three hours. The time required to assay a fraction is important because proteins in a crude extract can be degraded over time. The number of fractionation steps we could achieve in a day was limited by the time required for the assays rather than the time required for the actual chromatography. Second, the dynamic range of this assay is very narrow. Without knowing a priori what the activity of any given fraction will be, it is very difficult to guess what amount of extract to add to the assay to match that very narrow range. As a consequence, we often had a redo the assays, which is significant given that the assay is slow. Third, this assay does not seem capable of identify factors that are nucleotide- dependent. The amount of activity in the AS-fractions seemed to be similar, irrespective of added nucleotide or the use of MreB refolded in the presence of ATP or ADP (data not shown). As mentioned in Chapter 2, we think it is likely that Caulobacter contains

P a g e | 213

proteins that are capable of recognizing nucleotide-dependent conformational states of MreB and/or affect the nucleotide binding, hydrolysis, and exchange of MreB. In the future, it would be interesting to directly develop assays for factors that affect the rate of nucleotide hydrolysis or binding for MreB.

Is this activity relevant for MreB function in vivo? The major assumption that we made going into this work was that the assembled state of purified recombinant MreB is similar to its in vivo state or at least that the in vivo effectors of MreB can recognize preassembled polymers of MreB in vitro. Without knowing the precise structure of MreB in vivo (or even in vitro, See Chapter 4), it is difficult to assess the validity of this assumption. If it proves to be untrue, we could be assaying an activity that is an in vitro artifact and not useful for finding regulatory proteins that are physiologically important. Even in that worst-case scenario, however, it is possible that this activity assay and fractionation procedure could prove useful for further characterizing or manipulating the in vitro behavior of purified MreB. For example, the interaction between DnaseI and actin is not terribly relevant for understanding actin behavior and function inside cells. Nonetheless, it has been exploited for technical use: affinity columns of DnaseI can be used to purify actin, and labeled DnaseI can be used to specifically mark one end of the polar actin filament. Thus, DnaseI is an important tool for which there is no parallel yet for MreB.

AUTHOR CONTRIBUTIONS

I performed the experiments presented in this Chapter. Julie Theriot and Lucy Shapiro provided considerable intellectual input on this project.

ACKNOWLEDGEMENTS

We are particularly indebted to Matt Footer, whose biochemical expertise was invaluable to this project. We would also like to thank Aaron Straight and Kristina Godek for practical advice and reagents.

214 | P a g e

CHAPTER 5

MATERIALS AND METHODS

Buffers (in alphabetical order) BTE Buffer: 20mM Bis Tris Propane, 20mM NaCl, 1mM EDTA, 1mM DTT; pH was adjusted to 7.0 at 4C. BTEN Buffer: BTE buffer +1M NaCl and pH readjusted to 7.0. BTE6 Buffer: 20mM Bis Tris Propane, 10mM NaCl, 1mM EDTA, 1mM DTT; pH was adjusted to 6.0 at 4C. BTEN6 Buffer: BTE6 buffer +1M NaCl, and pH readjusted to 6.0. BUTYL Buffer A: 30mM Tris, 1M Ammonium sulfate, 2M NaCl, 1mM EDTA, pH adjusted to 7.5 at 4C. 1mM β-mercaptoethanol was added immediately prior to use. BUTYL Buffer B: 30mM Tris, 25mM NaCl, 1mM EDTA, pH adjusted to 7.5 at 4C. DEAE Buffer A: 30mM Tris buffer, 25 mM NaCl, 1mM EDTA, 1mM DTT. After all components were mixed and equilibrated at 4C, the pH was adjusted to 7.5. DEAE Buffer B: DEAE Buffer A + 1M NaCl and pH readjusted to 7.5. HA Buffer A: 5mM Sodium phosphate, pH 7. HA Buffer B: 0.5M Sodium phosphate, pH 7. HIC Buffer A: 50mM Sodium phosphate, 1M Ammonium sulfate, pH adjusted to 7.0 at 4C. HIC Buffer B: 50mM Sodium phosphate at pH 7. PEG Buffer A: 30mM Tris, 1.7M NaCl, 1mM EDTA, pH adjusted to 7.0 at 4C. SB Buffer A: 20mM HEPES, pH 7.5. SB Buffer B: SB Buffer A+1M NaCl. pH adjusted to 7.5. TBST: 100mM Tris, pH 7.4 at 20C, 150mM NaCl, 0.05% Tween 20. TEA-EDTA: 100mM Triethanolamine (Sigma), 20mM EDTA. pH adjusted to 7.8 at room temperature. WTB buffer: 10mM CAPS buffer, pH 11.0 at RT, 10% methanol.

P a g e | 215

Purification of MreB In this work, we strictly used recombinant Caulobacter MreB purified with a denaturing protocol (See Chapter 4). We observed only small differences in MDF activity in AS-fractionated extracts using MreB that had been refolded with ADP, ATP or no nucleotide (data not shown). Still, it is worth considering that refolded MreB is probably more stable in polymer form than natively purified MreB and does not rapidly exchange nucleotide. MreB was dialyzed into DEAE buffer A prior to use in the MDF assays.

Preparation of Caulobacter extract A starter culture of wild type CB15N Caulobacter (LS101) was grown from a frozen stock or colony to saturation (20-24 hrs) in 20ml of PYE while shaking at 30C. Cultures were then diluted 10-fold into prewarmed PYE media (added to 2L of PYE in a 6L flask) and allowed to reach exponential phase (0.4-0.6) while shaking at 30C. Cultures were then harvested with centrifugation at 10,000rpm in a JA10 rotor at 4C for 30 minutes. Supernatents were discarded. Pelleted cells were partially resuspended in 0.1X the original volume of DEAE Buffer A and respun. Supernatents were discarded. Pelleted cells were resuspended in 0.01-0.005X the original volume, rapidly frozen in liquid nitrogen, and stored at -80C until further use. For lysis, we defrosted the cell slurry by placing the tube in water at room temperature. Once defrosted, cells were transferred to ice and protease inhibitors (PMSF and Leupeptin) were added. Cells were lysed with a pre-cooled Avestin EmulsiFlex-C5 homogenizer at ~10,000psi. To remove cell debris, the lysate was centrifuged for 30 min at 23,000xg (14,000rpm in Fiberlite F15) at 4C. The pellet was discarded. The volume of the supernatant was measured with a graduated cylinder.

Protease treatment A stock solution of Pronase was prepared in water at 10mg/ml and frozen in small aliquots at -80C. For use, the stock solution was diluted 10-fold to 1mg/ml in water. To 100µl of clarified extract, we added 5µl of 1mg/ml Pronase and incubated the solution at 37C for 50min. Samples were spun at 16,000xg for 5 min. Supernatent was added to a new tube.

216 | P a g e

CHAPTER 5

Nuclease treatment and Ethidium Bromide spot testing To digest the nucleic acid in our sample we added the following (to the final concentration listed): RnaseA (Sigma) to 4µg/ml, calcium chloride to 3mM, magnesium chloride to 5mM, and DnaseI (Sigma) to 2.5µg/ml. Incubations were performed at 4C. The reaction appeared to qualitatively plateau after ~10min and was not considerably improved with an incubation at room temperature (10min, data not shown). For a quick and easy assay of the presence of nucleic acid, we used Ethidium bromide staining. A 5mg/ml stock solution of Ethidium bromide was diluted 2000X and then mixed 1:1 with the sample or buffer as a control (1µl of each). This sample was spotted onto a piece of plastic wrap and imaged with UV illumination.

Ammonium sulfate precipitation The soluble lysate was transferred to a beaker on ice in a temperature-controlled 4C room. A magnetic stir bar was added, and the sample was allowed to stir at a low setting (avoiding foaming). To help the Ammonium sulfate dissolve rapidly and evenly, we crushed aliquots of the salt with a mortar and pestle to create a fine powder. For each fraction, small increments of the powdered Ammonium sulfate were added to the sample, allowing the salt to slowly dissolve before adding more. After the last addition of Ammonium sulfate for that fraction, the sample was allowed to equilibrate while gently spinning on ice for at least 30 min (up to 3 hour). The pH of the sample was monitored and adjusted back to 7.5 as needed. The sample was then transferred to a centrifuge tube and spun at 16,000xg for 30 min at 4C. The supernatant was removed, and its volume was remeasured. The pellet was stored at -80C until further use. The concentration of Ammonium sulfate in the supernatant was slowly adjusted to the new desired concentration in the method described above. This procedure was repeated until all desired fractions had been obtained. To assay the Ammonium sulfate fractions, pellet samples were defrosted and resuspended in a small volume of DEAE Buffer A. The sample was then desalted with either dialysis (3500kD molecular weight cut off) or, more commonly, with a desalting column (G25 or PD10, GE). After resolubilization, the sample was respun (as above) to

P a g e | 217

remove particulates that failed to resolubilize in low concentrations of Ammonium sulfate.

Determination of protein concentration Protein concentration was measured with a Bradford assay in a microtiter plate according to the protocol in the BioRad manual using a standard curve of Bovine γ- Globulin.

DOC/TCA precipitation When necessary, we concentrated the protein samples by DOC/TCA precipitation prior to SDS-PAGE. To the sample, we added 0.015X volume of 1% deoxycholate (DOC) and incubated for 10min at room temperature. Then, 100% TCA (trichloroacetic acid) was added at 0.1X the sample volume. Samples were incubated on ice briefly and then spun at 16,000xg for 10min at 4C. Supernatent was removed and discarded. Pellets were resuspended in 10µl of 1M Tris base (not pH corrected, pH~10), and then SDS- PAGE sample buffer was added to the desired final concentration.

SDS-PAGE Standard methods for denaturing protein electrophoresis were used. The percentages of acrylamide used are indicated in the figure legends.

MDF activity assay For the MDF assay, 30µl of purified MreB at 2.5µM equilibrated in DEAE Buffer A and 10µl of extract sample were mixed and incubated at 4C for 15min. High speed centrifugation was performed in a TLA100 rotor at 100,000rpm (>300,000xg) for 15min at 4C. Supernatents were removed to a new tube, and SDS sample buffer was added. Samples were boiled, and then 2µl were electrophoresed on a 12% acrylamide gel. On the same gel as the MDF assays, we ran samples of MreB at known concentrations for the Western standard curve, as well as an MDF reaction containing buffer alone (no extract sample). Protein was transferred to nitrocellulose paper with a semi-dry transfer apparatus (the W.E.P. company) in WTB buffer.

218 | P a g e

CHAPTER 5

Western blots were performed using rabbit anti-serum raised against purified Caulobacter MreB (Josman) as the primary and HRP-conjugated anti-rabbit as the secondary. The nitrocellulose containing the protein sample from the gel was incubated in a solution of 5% milk in TBST buffer for 30 min (or overnight at 4C). The film was then incubated in a solution of primary antibody (diluted 1:10,000) in TBST+1% milk for 30 min at room temperature and then washed 3x10min in TBST. The film was then incubated in a solution of secondary antibody (diluted 1:10,000) in TBST+1% milk for 30 min at room temperature and then washed 3x10min in TBST. HRP reagent was applied according to the manufactor’s suggestions (Sigma), and the signal was captured on audoradiography film (Denville Scientific). To quantitate the MDF activity, we digitally scanned the exposed films and measured the intensity of each band with KodakMI software. A standard curve of intensity was generated with the samples of MreB run at known concentration and used to calculate the amount of MreB in the unknowns. The amount of soluble MreB in the presence of buffer alone was subtracted from that in the MDF assays. The protein concentration of the extract sample was calculated with a Bradford assay. To calculate specific activity, we divided the increase in the amount of MreB in the supernatant over buffer alone by the amount of extract protein (in mg) that was added to the assay. One unit is defined as 1nmol of soluble MreB (over buffer alone) released into the high speed supernatant (in the 15min incubation time). This assay is problematic for several reasons (some of which are addressed in the Discussion). To improve the reproducibility, we found it critical to perform at least two replicates for each experiment. If more than one gel was required to run all samples, we made sure to include a standard curve of MreB on each gel and use that curve to quantitate only those samples on that gel. This procedure helped to control for variations in the transfer step. The mean and standard deviations for these replicates are reported. We did not take any steps to remove the endogenously expressed MreB from the extract, but Western blots indicate that MreB is not present at high concentration in the extract samples (data not shown). By adding a relatively high concentration of purified MreB, we hoped to compete binding partners away from the endogenous form. Fractions were pooled based on peaks in the absorbance at 280nm. Only fractions

P a g e | 219

that appeared to have protein are shown.

Scheme I protocol Ammonium sulfate was added incrementally (as described above) to a clarified Caulobacter extract to a final concentration of 50%. The sample was spun at 10,000rpm in a JA10 rotor at 4C, and the pellet was discarded. Additional ammonium sulfate was added to 70% and spun again. This time supernatant was discarded and pellet was saved. The pellet was resuspended slowly in ~20ml of DEAE Buffer A and then dialyzed in 3500kD MWCO dialysis tubing against the same buffer for three hours with a buffer change halfway. The sample was spun for 10min at 23,000xg. Pellet (small, containing residual insoluble matter) was discarded, and the supernatent was filtered through a 0.45µm pore size syringe or bottle top filter and then loaded in a 50ml sample loop. Next, the sample was loaded onto a DEAE column (110ml) using an Akta Purifier FPLC (G.E. Biosystems) at a flow rate of 4ml/min. Absorbance was measured in an online flow cell at 256nm, to monitor nucleic acid content, and 280nm to monitor protein content. After the sample was loaded, the column was washed with two column volumes (CV). To elute the bound protein, we increased the concentration of DEAE Buffer B to 20% in 0.1CV and then from 20-25% in 3CV. A large peak of mostly protein (high 280nm signal) was released at this step. In the second step, we increased the concentration of DEAE Buffer B from 25-100% over 10CV. A huge peak of both nucleic acid and protein is released around halfway through this gradient. For the elutions, we collected 10ml fractions. Pools were made for the flow through, the first peak of elution, and the second peak of elution. At this point, the sample was either Ammonium sulfate precipitated (at 70%) for storage at -80C or dialyzed overnight (3500kD MWCO) with nucleases. By spot tests with Ethidium Bromide, we found that the nuclease reaction was complete in as little as 10min, but the dialysis step did not seem to be sufficient to remove all of the nucleotides (data not shown). Therefore, we did another ammonium sulfate precipitation in attempt to remove smaller soluble nucleotides prior to the next chromatography step. For the next step, the sample was equilibrated in BTE buffer and then loaded in a superloop over a 1ml Resource Q column at a flow rate of 7ml/min. The column was

220 | P a g e

CHAPTER 5 washed with 10CV, and then a gradient of 0-100% BTEN was applied. Fractions (1ml) were collected and pooled. For the gel shown in Figure 2, the samples were concentrated 5-fold with DOC/TCA precipitation.

Scheme II protocol The Ammonium sulfate and DEAE steps were formed as described in Scheme I. For the gel filtration step, we first concentrated the DEAE elute pool about 20-fold (after dialysis overnight with nucleases) with an Ammonium sulfate precipitation. Pellets were resuspended in the BTE buffer and filtered with a syringe 0.45µm filter. We used a self- poured gel filtration column of Superdex S200 resin, 127ml volume. Sample was run through the column at a flow rate of 1ml/min, and absorbance was monitored at 256nm and 280nm. For the gel shown in Figure 3, we used DOC/TCA precipitation to concentrate the samples 5-fold. For the Resource Q column at pH 6, we pooled elute fractions 15-17 from the gel filtration column and dialyzed that sample against BTE6 Buffer (3500kD MWCO). Sample was syringe filtered as above and loaded onto the Resource Q column equilibrated in BTE6 buffer. Column was washed with 5CV of BTE6 buffer. Sample was eluted with a gradient of 0-100% BTEN6 Buffer over 20CV. For the gel shown in Figure 3, the samples were concentrated 5-fold with DOC/TCA precipitation.

Scheme III protocol Ammonium Sulfate precipitation and DEAE chromatography was performed as in Scheme 1 and 2, except that the sample was desalted with a G25 column rather than dialysis prior to the DEAE step. After elution from the DEAE column, the sample was split into two. To both aliquots, we slowly and carefully added ammonium sulfate from a 3M stock solution to a final concentration of 1M. No precipitation was observed, but the sample was syringe filtered anyway to remove small particulates. HiTrap Butyl and Phenyl_highsub FF HIC columns were used (GE Biosciences). Both columns were run with the same method: sample was loaded in HIC buffer A at a flow rate of 1ml/min; column was washed with 5CV; sample was eluted with a gradient of 0-100% HIC buffer B over 10CV.

P a g e | 221

The AS precipitation and DEAE chromatography steps were repeated with another lysate sample, and then the DEAE elute was applied to Blue sepharose, S sepharose and hydroxyapitite columns. For these columns the sample was first desalted with a PD10 column. Then the sample was applied to either HiTrap Blue or S (1ml, FF, GE Biosciences) in SB buffer and eluted in a gradient elution of 0-100% over 10CV. For the hydroxyapitite, we used an HA Ultragel custom poured 4ml column. Sample was equilibrated in HA Buffer A, loaded, and eluted with a gradient of 0-100% B over 5CV.

Scheme IV protocol Ammonium sulfate precipitation was performed as in Scheme 1. Pellets were resuspended in ~1X volume of PEG Buffer A and incubated at 4C for 20min. A stock solution of 30% (w/v) PEG8000 in water was added slowly to the sample to a final concentration of 10%. Because PEG solutions are so viscous, we monitored the addition of PEG by weight. The sample immediately became cloudy, indicating precipitation of nucleic acids. The sample was allowed to equilibrate at 4C for 30min with mild agitation and then centrifuged for 15min at 10,000xg. Supernatent was removed and transferred to a new tube. The pellet, quite large and white, was discarded. To the remaining supernatant, we adjusted the concentration of NaCl, ammonium sulfate, and EDTA to match the concentration in BUTYL Buffer A by diluting from stock solutions. The sample was spun for 10min at 15,000rpm in a JA21 rotor. This spin causes the sample to separate into two visible layers, a very viscous PEG layer floating above the aqueous sample. The bottom layer was removed and filtered with 0.45µm syringe filter. Butyl HIC chromatography was performed as in Scheme III. For the Q column, a PD10 column was used to desalt the eluted fractions from the Butyl HIC. A 1ml HiTrap Q column was used with DEAE Buffers A and B. Samples were eluted with a 0-100% gradient of Buffer B.

Transaldolase assay To measure the enzymatic activity of Transaldolase, we used the method of Tsolas and Joris [284] . This method links the Transaldolase reaction with the oxidation of NADH by glyceraldehyde dehydrogenase. The loss of NADH is followed by the loss

222 | P a g e

CHAPTER 5 of absorbance at 340nm in a spectrophotometer. Reagents: D-fructose-6-phosphate (Sigma): dissolved to 0.14M in water Erythrose-4-phosphate (Sigma): 10mM in water NADH (aka DPNH): 10mM. Dissolved fresh at each use in 0.001N NaOH (inhibitor develops with time and when dissolved in water or acid) and protected from light. α-Glycerophosphate dehydrogenase-Triose phosphate isomerase: mixture of the two enzymes at 10mg/ml (Roche). Reaction Setup: 940µl of TEA-EDTA buffer + 20µl Fructose-6-phosphate + 20µl Erythrose-4- phosphate + 10µl NADH + 1µl GDH-TIM enzyme mix + 10µl buffer or extract sample (added immediately prior to measuring). Samples were monitored at 340nm every 10 sec for 600 sec (0.5 sec integration time). We used an extinction coefficient for 6300M-1cm-1 to convert absorbance into amount of NADH. Activity is reported as the rate of nmol of NADH oxidized per minute. Specific activity is that activity divided by the total milligrams of protein added to the assay.

Transaldolase knockout The gene encoding Transaldolase in Caulobacter is cc3614. To create an in frame deletion, we amplified ~500bp of sequence upstream and ~500bp of sequence downstream of this gene. These sequences were cloned these sequences into the integrating vector pNPTS138 (LS1914) with a triple ligation. The resulting plasmid was electroporated into wild type CB15N Caulobacter. By selecting with Kanamycin, we obtained the single integration event. To select for the double homologous recombination event, we first grew the strain overnight in the absence of Kanamycin. We then plated the cells on sucrose and selected for cells that were sucrose resistant and Kanamycin- sensitive. Deletions were confirmed with PCR amplification of the entire locus. Deletion strains had much smaller colonies and took 3 days to reach high turbidity in liquid culture.

P a g e | 223

Mass Spectrometry Identification of proteins present in gel slices was performed at the Protein and Nucleic Acid (PAN) facility at the Beckman Center at Stanford University. Additional data for the proteins identified in each band slice indicated in Figure 2F are presented in Table 1.

Table 1: Mass spectrometry data from the gel slices in Figure 2.

Band Protein Name Accession # (gi) Protein MW (Da) Protein PI Peptide Count 1 DNA-directed 16124757 151024 5.25 27 RNA polymerase beta subunit 1 DNA-directed 16124758 154952 6.4 22 RNA polymerase beta’ subunit 2 DNA 16126690 97891 8.68 28 topoisomerase I 3 DNA-directed 16125521 37363 4.97 13 RNA polymerase alpha subunit 4 Transaldolase 16127844 23186 4.53 6

224 | P a g e

Chapter 6 Does MreB Form an Actin-like Cytoskeleton in Bacteria?

Natalie Dye and Julie Theriot

P a g e | 225

ABSTRACT

In 2001, Jan Lowe and colleagues published the crystal structure of MreB from the thermophilic bacterium T.maritima, demonstrating that this protein has a tertiary structure that is remarkably similar to that of eukaryotic actin [8]. In the same year, Jeff Errington and colleagues demonstrated that GFP-labeled homologs of MreB in B.subtilis localize to helical patterns on the cytoplasmic face of the membrane [116]. Such a localization pattern had never before been observed in bacteria. Coupled with a well- established role for MreB in the maintenance of cell shape in rod-like bacteria [38, 116, 285], the existence of what appeared to be a continuous filamentous structure inside cells was suggestive of MreB being the major component of a true bacterial cytoskeletal filament. But in the nine years since these landmark papers were published, more data has accumulated that make MreB seem less and less actin-like. Is it possible that this widespread and appealing analogy between MreB and actin is actually more limiting than helpful for understanding the function of this protein in bacterial cells?

226 | P a g e

CHAPTER 6

WHY DO WE THINK THAT MREB IS CYTOSKELETAL?

The similarity between the MreB and actin tertiary structures is quite remarkable [8] and likely indicates that these proteins are ancestrally related. As yet, however, it has not been conclusively proven that MreB is actually a cytoskeletal protein in bacterial cells as actin is in eukaryotic cells and there may be other models to explain the existing data. Let us operationally define a ―cytoskeletal‖ protein to be one that forms a linear, filamentous homopolymer that provides mechanical support, facilitates the active transport of other molecules, and/or generates force. MreB is not the only bacterial protein that is thought to be ―cytoskeletal‖ [92]. ParM is another prokaryotic member of the actin superfamily and is encoded on plasmid R1 in E.coli and other Enterobacteria. MreB and ParM are quite distinct in sequence and structure; they are no more similar to each other than to eukaryotic actin [10]. ParM is required for plasmid R1 segregation and can be seen with fluorescence microscopy to form a linear filament between two clusters of segregating plasmids in vivo [109, 112, 148]. In vitro, ParM forms helical filaments that perform nucleotide-dependent dynamic instability [10, 111, 113, 148], and DNA segregation has been reconstituted in vitro with purified ParM, DNA bound to polystyrene beads, and ParB (a protein that binds DNA and ParM) [286]. The dynamics of ParM polymerization in this in vitro assay are consistent with the dynamics of polymerization and DNA segregation in vivo [111, 112, 286]. ParM filaments have also been visualized in cells at high resolution using cryoelectron microscopy [287]. Thus, there is clear evidence to show that ParM forms a filament both in vivo and in vitro and that it generates force to facilitate the active transport of DNA. It is a true bacterial cytoskeletal protein. Much less is known about MreB, which is encoded on the chromosome and is conserved in bacterial species with a non-spheroidal shape [43, 116]. Other than its structural similarity to actin and ParM, what is the evidence that MreB is cytoskeletal? Recombinantly purified MreB homologs from T.maritima, E.coli, and B.subtilis polymerize in vitro [8, 118, 241, 242, 263]. Unlike actin and ParM, which form filaments of two protofilaments twisted around each other, MreB forms structures of bundled straight protofilaments that can be several microns long and are not dynamic once formed

P a g e | 227

[8, 118, 241-243]. MreB, ParM, and actin share very little sequence or structural similarity at the predicted subunit interfaces [8, 10, 288]. Additionally, several loops in the actin structure, which are thought to be important for folding, polymerization, and dynamics, are missing in MreB [288, 289]. Thus, it is perhaps unsurprising that MreB filaments do not resemble those of actin. In cells, MreB does not appear to make one long straight bundle as it does in vitro. Instead, fluorescence microscopy (using both GFP fusion proteins and immunofluorescence) shows that MreB localizes to a roughly helical pattern in B.subtilis, E.coli, Caulobacter and other bacteria [39, 43, 59, 116, 117, 121, 224, 231, 268, 290, 291]. If MreB inherently forms a long rigid filamentous polymer, this type of helical pattern could be generated by forcing the polymer to interact with the membrane and be confined to a cylindrical geometry [292, 293]. But in the last few years dozens of other molecular components of the bacterial cell have also been observed to have a similar, roughly helical localization, including the Min proteins [59], chemoreceptors [294], the Sec secretion apparatus [295], proteins involved in cell wall synthesis [46, 142, 143, 180, 206, 224, 225, 231, 296], outer membrane proteins [193, 297], RnaseE [298], viral replisomes [299], replication initiation factor DnaA [300], and even LPS in E.coli and lipids in B.subtilis [193, 301]. Many of these patterns do not completely overlap with that of MreB nor do they require MreB for their maintenance. How then do we know that the helical pattern of MreB is generated by its own unique ability to form filamentous polymers in the cell? Is the helical pattern of MreB really indicative of a cytoskeletal structure or is it reflecting a more general property of the organization and geometry of the bacterial cell that we do not yet understand? No one has yet been able to image MreB filaments in wild type cells at high resolution using such methods as cryoelectron microscopy. Thus, definitive evidence for MreB forming a filament in vivo is currently absent. While the purified protein is capable of polymerization in vitro, the filaments are much longer than the cell and are not dynamic, whereas in the cell, dynamic movement, reorganization and turnover of MreB-containing structures has been demonstrated with live cell fluorescence microscopy and photobleaching techniques [117, 177, 178, 227, 228]. Thus, if MreB forms a filament in vivo, the cellular environment and potential binding partners are likely to cause MreB to form filaments that are very different in structure than those seen

228 | P a g e

CHAPTER 6 in vitro. It is also possible that there are no filaments at all in cells and that the localization and dynamics of MreB are driven by other mechanisms. Clear evidence for a traditional cytoskeletal function is also currently absent for MreB. As with other cytoskeletal proteins in eukaryotic cells, there is considerable genetic and pharmacological evidence to indicate that MreB has a role in morphology in both Gram-positive and Gram-negative branches of bacterial phylogeny [39, 40, 43, 116, 117, 121, 141, 187, 205], but there is currently no direct evidence to suggest that MreB is required for structural support. Treatment of Caulobacter with A22 rapidly delocalizes MreB but does not immediately cause the cells to change shape; rather, cell shape changes gradually over time as the cell grows [40]. Thus, it is thought that MreB influences cell shape by determining where and when new peptidoglycan cell wall grows as the cell elongates [43, 116, 121]. The mechanism by which MreB influences morphology is not well understood. There is no evidence to support the existence of motor proteins analogous to myosin and kinesin that track along filaments of MreB, nor is there any direct evidence to suggest that MreB functions as a polymerization rachet like actin in the movement of Listeria monocytogenes [302] or like ParM in the movement of plasmid DNA [286]. If we assume as a null hypothesis that MreB is NOT an actin-like cytoskeletal protein, we cannot reject this null with existing data. Of course, this area of research is still active. The lack of clear, positive evidence is not definitive proof that MreB is NOT an actin-like cytoskeletal element, and it may be true that MreB is a cytoskeletal filament that is fundamentally different from eukaryotic actin. But are there also additional models to explain the existing data that do not require MreB to be cytoskeletal at all?

P a g e | 229

MECHANISMS FOR SLOW, PERSISTENT MOTION

In 2006, an important paper addressing the dynamics of MreB in Caulobacter crescentus was published [228]. In this work, the authors expressed YFP-MreB at a very low level from an inducible promoter so that they could follow the motion of single molecules in live cells. Two types of motion of MreB were observed: rapid diffusion and slow directed motion. In the presence of the drug A22, which disrupts the helical pattern of MreB in the cell, only the rapid diffusive motion is observed, indicating that this motion corresponds to MreB that is dispersed and delocalized. The slow directed motion has an average rate of 6 nm/s and persists for an average of 392 nm. The tracks of motion are roughly perpendicular to the long axis of the cell or at a slight angle, with no apparent preference for direction towards either pole or either side of the cell. Currently, the popular view is that MreB is an actin-like cytoskeletal polymer and thus the slow directed motion corresponds to monomers of MreB that are treadmilling through a filamentous polymer of MreB (Fig 1B) [50, 204, 228, 293, 303]. But had this experiment been done prior to 2001, when the hypothesis that MreB forms an actin-like cytoskeleton was established, would another hypothesis have been favored? Let us now consider all of the plausible ways to explain the motion of MreB in cells, including models that do not invoke the existence of an actin-like cytoskeleton in bacteria. The persistent, directed motion of MreB defies the laws of diffusion and therefore argues for the existence of a large scale structure that influences the motion of MreB, either directly or indirectly. Because the tracks of MreB motion are found throughout the cell and of a finite length, this structure must be present in all regions of the cell in segments of at least 300-400 nm. These segments must be oriented roughly perpendicular to the long axis of the cell. Lastly, the model must be able to explain the directed movement of MreB at a rate of ~6 nm/s. We can think of three structures in a bacterial cell that could meet these requirements: a cytoskeleton, the peptidoglycan cell wall, and the nucleoid.

230 | P a g e

CHAPTER 6

FIGURE 1

P a g e | 231

A cytoskeletal structure For the motion to be explained by cytoskeletal treadmilling, we must assume that MreB filaments have the equivalent of a plus and minus end with different kinetics of assembly and disassembly [304]. We also must assume that the filaments are bounded by at least one fixed anchor point that can maintain attachment to a growing and shrinking filament end; otherwise, the monomer would stay stationary, even during the process of treadmilling (Fig 1A-B) [305, 306]. Treadmilling of microtubules in vivo has been observed both in the presence and absence of barriers. In the mitotic spindle, microtubules are anchored at the minus end by the microtubule organizing center (MTOC) and at the plus end by a kinetochore complex that associates with the chromosomes or with overlapping microtubules originating from the opposite spindle pole. In this case, the assembly and disassembly of microtubules, combined with the activity of motor proteins, drive the poleward flux of monomers within a filament at rates that range between ~8-80 nm/s [307-313]. In interphase, however, the minus ends of microtubules can detach from the MTOC and the plus ends are no longer bound to the chromosomes or any other load. In this case, treadmilling-style assembly and disassembly of microtubules results in translocation of the entire filament while each individual monomer stays stationary [314]. Actin filament treadmilling has been measured in vitro to be about 0.5-1.5 nm/s [315-317], and the presence of accessory factors can greatly accelerate this rate [318]. There has yet to be any direct demonstration of treadmilling for individual actin filaments in vivo, though flux of monomers through larger structures of interconnected filamentous actin has been demonstrated, including in branched actin networks of motile cells at ~25 nm/s [310, 319, 320] and bundled paracrystalline actin of stereocilia at ~0.05 nm/s [321]. Thus, the speed of MreB movement is within an order of magnitude of the rates that have been observed for the treadmilling-based movement of eukaryotic cytoskeletal elements. It is not unreasonable to assume that MreB filaments assemble in a polar fashion and that there are fixed barriers, but there also is no direct supporting evidence for these assumptions. In fact, the assembly of ParM is bidirectional, so unidirectional treadmilling of this polymer is not expected to occur [111, 286]. Since the tracks of MreB motion are short, the barriers for the MreB filaments cannot be the ends of the cell, unless

232 | P a g e

CHAPTER 6 the protofilaments are bound to a continuous bundled structure that runs from pole to pole. Since two molecules can be seen moving in opposite directions in the same cell [228], this structure has to have at least two bundles of opposite polarity. The tracks of MreB motion are nearly perpendicular to the length axis, however. For these tracks to be a part of a larger helical structure, the pitch would have to be very tight, and the fluorescence images of bulk MreB in cells are not consistent with this prediction [39, 117, 250]. Thus, it does not seem likely that the cell poles can be the barriers. There are transmembrane proteins that have been shown to interact with MreB through bacterial two-hybrid assays and colocalization studies (RodZ, MreC, Pbp2) [141-143, 180, 205, 223-225, 227, 231] and could function as filament anchors, though none have been shown to have the specific filament end-binding properties that would be required to support this kind of treadmilling. It is also possible that the MreB motion could be explained by an association with other bacterial cytoskeletal elements. Two candidates are FtsZ, a tubulin homolog, and ParA, a Walker ATPase that has been proposed to have cytoskeletal function [92]. Single molecules of FtsZ in E.coli do not exhibit this directed motion, however [322]. ParA is not known to interact with MreB, and there is not yet any information about its single molecule behavior.

The peptidoglycan cell wall The peptidoglycan cell wall is a network of glycan polymers crosslinked by short peptides. Growth of the cell wall involves the incorporation of new glycan strands sporadically throughout the entire network of existing crosslinked glycans, requiring enzymes to polymerize the glycans, hydrolyze the links between old strands, crosslink the new strand with the old strands, and degrade old glycans [12, 13]. The cell wall can be considered a single large macromolecule; thus, the enzymes that synthesize new cell wall must be able to move relative to this stationary object, much like the DNA replisome moves relative to the nucleoid [323]. Also like the replisome, it has been proposed several times that the enzymes that are required for peptidoglycan synthesis exist in large complexes [13, 25, 26, 141, 143, 180, 205, 324]. There is clear evidence that MreB is required for the synthesis of new cell wall [39, 43, 116, 117, 121, 141, 187, 205]. Might

P a g e | 233

the movement of MreB be explained by a connection between MreB molecules in the cytoplasm to the enzymes that synthesize the cell wall (Fig 1C)? For this model to be correct, we have to assume that MreB is coupled to the cell wall and that the synthetic enzymes are capable of processive motion for an average of 390 nm at a rate of ~6 nm/s. Since MreB exists in the cytoplasmic compartment of the cell, it would have to be bound to a transmembrane protein, which could be one of peptidoglycan synthetic enzymes (candidates include Pbp2) or an intermediary protein that is not an enzyme itself but binds to members of a synthetic complex or directly to peptidoglycan (candidates include MreC and RodZ). The movement of the synthetic enzymes should be dictated by their inherent processivity and the orientation and length of the glycans. In Gram-negative organisms such as Caulobacter, the glycan strands are thought to lie in the plane of the membrane and be oriented roughly perpendicular to the long axis of the cell [13, 18, 325], which matches the orientation of the MreB tracks. In E.coli the average steady state length of glycan strands is 20-30 disaccharides, though strands up to 100 subunits have been isolated [13, 326-328]. In vitro transglycosylase reactions with Lipid II substrate and purified E.coli PBP1A, a bifunctional transglycosylase, suggest that this enzyme can processively synthesize glycans for up to 50 subunits, with minimal accumulation of shorter products [329, 330]. Each subunit disaccharide is approximately 1 nm in length [12], so processive synthesis of glycans up to 50 nm in length in E.coli is too short to account for the persistent directed motion of MreB in Caulobacter for >300 nm. Nonetheless, it is possible that transglycosylases, when bound to the intact peptidoglycan sacculus and complexed with other enzymes, are able to exhibit processive motion that exceeds the lengths of individual glycan strands. Alternatively, the transglycosylase from Caulobacter may be even more processive than that of E.coli and able to make longer glycan strands. It is also possible that the transglycosylase is not the cell wall-associated enzyme that is directly driving the motion of MreB. Autolytic enzymes and the transpeptidase Pbp2 are other candidates, but measurements of processivity have not been made for these enzymes in the context of the entire peptidoglycan mesh in live cells. The exact rates with which these enzymes polymerize and crosslink new glycan strands have not been directly measured in live intact cells. We can estimate this rate,

234 | P a g e

CHAPTER 6 however, to approximate the rate with which these enzyme complexes could be moving. A newborn Caulobacter cell has a circumference of ~1570 nm and a length of ~1500 nm. Assuming 2 nm lateral spacing between strands, this corresponds to approximately 1.2x106 glycan disaccharides in total. To double in size in 180 min, a cell needs to make at least 109 disaccharides/s. Let us assume that peptidoglycan is synthesized by a complex of all the required enzymes. To determine the rate of movement for a single complex, we would need to know how many active complexes there are per cell, and for Caulobacter this information is not yet available. In E.coli, there are 58+/-8 molecules of the elongation-specific transpeptidase Pbp2 in a cell that is growing in minimal media [48]. If we assume that Caulobacter has roughly the same amount of Pbp2/cell as E.coli, that each synthetic complex has one Pbp2 molecule, and that all of these complexes are active, each complex would move a rate of 1.9 nm/s, which is comparable to the observed rate of MreB single-molecule movement.

The nucleoid The DNA occupies almost the entire space of the Caulobacter cytoplasm. In Caulobacter, the origin of replication is at the flagellated pole, the terminus is at the other end of the cell, and the intervening sequences are regularly spaced between these loci [82]. The genome is therefore a highly organized structure that could in theory affect the intracellular organization and motion of other macromolecules in the bacterial cell. Could this structure be part of the mechanism that controls the localization and dynamic behavior of MreB? MreB has been implicated in chromosome segregation in Caulobacter, and regions of the chromosome co-immunoprecipitate with MreB [40]. In E.coli, MreB is thought to directly interact with ParC, a topoisomerase subunit [118]. In B.subtilis, MreB is excluded from regions of the cell that are devoid of DNA, and the localization of MreB is affected by the deletion of parE, another topoisomerase subunit [178]. Thus, it seems possible that MreB has a functional and physical connection to the DNA. One mechanism for nucleoid-dependent motion would involve an association between MreB and the replisome. In Caulobacter, the replisome moves from one pole to the other at a rate of ~0.13 nm/s [323]. This rate was measured in growing cells, however, and the rate of cell

P a g e | 235

growth is only slightly slower than the rate of observed replisome movement. Therefore, the rate of replisome translocation is likely to be much too slow to account for the persistent motion of MreB. Additionally, only one replisome is seen in the cell at any given time, whereas in the single molecule experiment, more than one track of persistent MreB motion could be seen in the same cell [228]. Thus, the motion of MreB cannot be due to direct binding between MreB and the replisome components. Another theoretical model for nucleoid-dependent directed motion of MreB involves a DNA-binding protein that binds to DNA and then spreads along the DNA to make a nucleoprotein filament (Fig 1D). If MreB were associated with the DNA, either directly or indirectly, it is conceivable that it could be pushed along the DNA, through a Brownian rachet-type mechanism [149], by another protein that is binding to and spreading along the DNA strand. Nucleoid-associated proteins (NAPs) are abundant and bind to loci throughout the chromosome [331]. It has been proposed that H-NS binds first to high affinity sites and then polymerizes on DNA to make a nucleoprotein filament [332-334]. Caulobacter does not have an H-NS homolog, however, and spreading of H- NS has only been examined with short segments of DNA (<1 kB). If the 4 Mbp genome of Caulobacter were linearly packed into a 1.5 um cell, 1 kB of DNA would correspond to ~0.4 nm. An NAP would have to spread for almost 1 MB of DNA (almost a quarter of the genome) to account for the ~400 nm of persistent motion of MreB. Furthermore, there are currently no known examples of an NAP being able to function as a polymerization rachet to direct the motion of other proteins. Thus, it does not seem likely that the processive polymerization of NAPs on DNA is sufficient to explain the motion of MreB. Lastly, MreB has been implicated in chromosome segregation in Caulobacter [40, 117, 219], which occurs through an active transport mechanism that is poorly understood [81, 82, 335]. If MreB is physically connected to DNA as it is being segregated, MreB motion would be expected to match that of DNA as it is being segregated. In fact, segregation in Caulobacter occurs at 3.5-7.8 nm/s [82], which is remarkably close to the speed of single molecule MreB. The orientation of the movement is in the wrong axis, however: DNA moves parallel to the long axis of the cell, whereas MreB moves roughly perpendicular to this axis.

236 | P a g e

CHAPTER 6

CONCLUDING REMARKS

The models listed above provide reasonable explanations for the persistent motion of MreB. Given the existing data, it seems unlikely that the nucleoid is driving the motion of MreB. We cannot, however, distinguish between the other two plausible models. These two models make very different, but testable, predictions. In the cytoskeletal model, a transmembrane anchor protein remains stationary relative to the processive movement of MreB, whereas in the peptidoglycan model, this protein moves with the same rate and directional persistence as MreB. The peptidoglycan model predicts that the motion of MreB would be very sensitive to antibiotics that target cell wall synthesis and genetic perturbations that affect the localization or activity of the cell wall synthetic machinery. In the cytoskeletal model, MreB movement may also be affected by antibiotics that target the cell wall but probably only on long timescales, after large scale disruptions of cell wall structure. The cytoskeletal model requires MreB to polymerize into a polarized filament. If this model is true, the filaments of MreB would each be only ~390 nm long. The direction of the movement of MreB suggests that the MreB cytoskeleton would be more accurately described as a series of rings or arcs rather than a continuous helix, and more precise quantitation of the fluorescence images of MreB in cells may be able to distinguish between these two types of structures. The peptidoglycan model does not require MreB to form filaments at all, leaving us to wonder about the true role of MreB in shape determination and how the helical localization pattern is established. The model that MreB is a cytoskeletal protein analogous to actin is enticing because it implies that prokaryotes and eukaryotes, despite their evolutionary distance, have similar mechanisms for intracellular organization and growth. However, the peptidoglycan-based model is arguably just as interesting because it suggests that prokaryotes use this actin homolog in a fundamentally different way. If we continue to think of MreB only in the context of actin, perhaps we will miss an opportunity to discover new and different biology.

P a g e | 237

ACKNOWLEDGEMENTS

We would like to thank Ethan Garner and Guy Ziv for reminding us that the single molecule data on YFP-MreB in Caulobacter can only be explained by a treadmilling model if there are fixed barriers to polymerization. We also heartily thank K.C. Huang and all the members of the Theriot lab for extensive discussions on the subject and Lucy Shapiro for thoughtful feedback on the manuscript.

238 | P a g e

Table 1. Characteristics of the different types of structures that could potentially drive the slow, directed motion of MreB.

Structure Rates Length Subcellular localization Orientation* References

MreB in Caulobacter 6 nm/s ~400nm Throughout Perpendicular [228]

Treadmilling eukaryotic cytoskeleton Microtubules in mitotic spindle 8-80 nm/s Microns N/A N/A [307-313] Actin filaments in vitro 0.5-1.5 nm/s Microns N/A N/A [315-317] Actin networks in motile cells ~25 nm/s Microns N/A N/A [310, 319, 320] Actin bundles in stereocilia 0.05 nm/s Microns N/A N/A [321]

Coupling to peptidoglycan growth E.coli PBP1A Transglycosylation in vitro 50 nm Throughout Perpendicular [329, 330] Length of glycan strands in E.coli 20-30 nm avg; up Throughout Perpendicular [13, 18, 325-328] to 100nm Estimated speed of enzyme complex (in vivo for ~1.9 nm/s Throughout Perpendicular This work Caulobacter growing in minimal media)

Nucleoid-dependent motion Replisome 0.13 nm/s 1-2 microns Cell cycle dependent (single Parallel [323] focus that moves from end of cell to other) Nucleoprotein polymerization < 1kB (0.5nm) Throughout [332-334] Chromosome segregation 3.5-7.8 nm/s 1-2 microns Throughout Parallel [82]

* Orientation of movement, relative to the long axis of the cell.

240 | P a g e

Chapter 7 Concluding Remarks

Natalie Dye

P a g e | 241

The last decade has been an exciting time of growth and discovery in the field of bacterial cell biology. The widespread existence of ―cytoskeletal‖ proteins was a remarkable, paradigm-shifting discovery. Understanding how these proteins function could provide important insight into fundamental cell biological processes in these organisms and into the evolution of the eukaryotic cytoskeleton and cellular life. The work that I have described in this thesis represents a significant contribution to our understanding of the role that MreB plays in the determination of cell shape in Caulobacter. But of course, so many unanswered questions remain. At the beginning of this project, the prevailing model for cytoskeletal-mediated peptidoglycan synthesis was something similar to the ―dynamic scaffolding‖ mechanism presented in Chapter 1 (the Push and the Pull of the Bacterial Cytoskeleton). Several groups proposed that the enzymes of the periplasmic peptidoglycan synthetic machinery (referred to as the ―brick-layers‖ in Chapter 1) exist in complexes, one specific for division and one specific for elongation [17]. In this model, the cytoskeletal proteins FtsZ and MreB self-assemble into higher-order structures in the cell and direct the localization and activity of these complexes, where FtsZ would be specific to division and MreB for elongation. For MreB, the helical pattern would arise as a consequence of forcing an inherently stiff filamentous polymer to be confined to the cylindrical geometry [292, 293]. The underlying assumption here is that without filamentous structures of FtsZ and MreB, these brick-layer complexes do not form or are diffusely organized; diffuse growth ensues and shape is perturbed. We have shown in Caulobacter that the brick-layer proteins MreC and Pbp2 do not colocalize with MreB and are not recruited to the poles in polar MreB mutants. It has also been shown by other groups that in B.subtilis and E.coli, at least some brick-layer molecules can maintain a punctate/helical localization after removal of MreB and do not always colocalize with MreB [46, 47, 231]. Furthermore, it has been shown in E.coli that the synthesis of peptidoglycan is not actually diffuse in the absence of MreB and FtsZ [122]; rather, synthesis occurs in large clusters, leading to the formation of bulges and eventual lysis. This result is somewhat counterintuitive, as it implies that the brick-layer proteins, in the absence of cytoskeletal elements, are inherently predisposed to cluster together in the cell rather than spread out diffusely across the whole surface. Indeed, we

242 | P a g e

CHAPTER 7 also showed that Pbp2 becomes concentrated at midcell, rather than helically dispersed along the sidewalls, in the absence of MreB (or MreC, Chapter 2). As discussed in Chapter 6, there is (as yet) no direct evidence to suggest that MreB forms a continuous, filamentous structure in cells. In vitro, we observed that purified Caulobacter MreB is capable of forming large stable polymers but that the filamentous structures that were expected can only be observed under special circumstances (Chapter 4). From these recent data, we conclude that the structural scaffolding model must be incorrect or at least incomplete: MreB cannot be a scaffold to direct the localization of the brick-layers, and the dynamic pattern of MreB localization in the cell must be influenced by cellular factors. Further, these results suggest that the function of MreB must be to prevent cell wall synthesis from being too clustered, rather than limit it to a defined helical path. We (my coworkers and I) would like to propose an alternative highly-speculative model wherein MreB, rather than serving as a structural scaffold, provides spatial information that functions to spread out peptidoglycan synthesis at regular intervals. We predict there are cellular factors that can control the subcellular localization, polymeric structure, nucleotide state, and/or activity of MreB. We propose that the pattern of MreB localization in the cell is generated by a chemical-based reaction diffusion mechanism analogous to the Min system for positioning FtsZ in E.coli. In this model, the spontaneous polymerization of MreB would lead to the formation of clusters of MreB polymers that may or may not be filamentous. The disassembly and regular spacing of these clusters would be controlled by a long range inhibitory cellular component. Since we observed the energetics of polymerization to be irrespective of nucleotide (Chapter 4), cluster formation is probably not solely controlled by nucleotide hydrolysis and likely involves other factors. As with FtsZ, we predict that the spatial positioning of MreB is achieved through negative regulation of polymerization, which is fundamentally different from the regulation of eukaryotic cytoskeletal elements. The preliminary data we presented in Chapter 5 supports the existence of a biochemical activity for the disassembly of MreB polymers in Caulobacter extract. This spatial patterning of MreB would ensure that the synthesis of new cell wall occurs in regularly spaced out intervals along the lateral surface. Note that this spatial patterning does not require MreB to form a continuous structure.

P a g e | 243

How exactly the spatial patterning of MreB is translated to the peptidoglycan synthetic machinery is still mysterious. We can think of both biochemical and biophysical mechanisms. First, it is possible that MreB directly binds to the cytoplasmic ―brick-maker‖ enzymes (those that synthesize the repeating disaccharide unit) and functions to cluster these enzymes together into microdomains on the surface. Very recently, others have shown that MreB can interact with many of the cytoplasmic peptidoglycan synthetic enzymes in a bacterial two-hybrid assay [207]. It seems possible that MreB, by interacting with several of these enzymes, could function as a platform to concentrate these enzymes into small domains that would lead to efficient export of the disaccharide subunit. There is some evidence to suggest that the brick-layer enzymes are localized by their ability to bind the substrate [195]. Thus, by localizing the export of the peptidoglycan precursor molecules, MreB would indirectly influence the localization, or more likely the activity, of the brick-layer proteins. In an alternative biochemical model, MreB could locally inhibit the brick-layer proteins directly, thereby controlling their localization through negative regulation. The mechanism for this type of repulsion is unclear, however, particularly since the brick-layer molecules are either periplasmic or transmembrane with only very small cytoplasmic domains. This mechanism could involve RodZ, which has both cytoplasmic and periplasmic domains and binds directly to MreB [214]. The ability of MreB to polymerize may indeed indicate that a biophysical, rather than biochemical, mechanism underlies its function. For example, MreB could function by physically blocking where new cell wall synthesis can occur. Though MreB is cytoplasmic, many of the brick-layer molecules are transmembrane and must diffuse in the membrane space. By polymerizing into small patches on the surface, MreB may physically block the diffusion of these enzymes, thereby determining where synthesis cannot occur. Indeed, it has been suggested that the intermediate filament Crescentin generates curvature through this type of negative regulation of cell wall growth, physically blocking growth on one side of the cell and indirectly promoting growth on the opposite side [124]. Alternatively (or in conjunction with the above mechanism), MreB may structurally alter the properties of the cell wall locally. As mentioned in Chapter 1, Arthur Koch’s Surface Stress Theory proposes that the synthesis of new cell wall occurs

244 | P a g e

CHAPTER 7 where the wall is the most stressed [14]. He suggests that the cell wall synthetic machinery is most active in areas of high tension. Perhaps patches of MreB polymers are capable of resisting stress locally and thereby indirectly activating cell wall synthesis at nearby sites. These biophysical mechanisms, if true, would indicate that this prokaryotic actin functions in a very different way than eukaryotic actin and tubulin in the determination of cell shape. Of course the above-mentioned models are not mutually exclusive. Furthermore, a model that posits that MreB is only inhibitory cannot explain the polar MreB mutants, in which MreB appears to be associated with the areas of active elongation in the polar and near-pole regions. Additionally, the dynamic behavior of MreB, from the level of single- molecules to the whole cell pattern, is likely to be an important part of its function, and determining how this behavior is altered in the MreB mutants is likely going to be enlightening for determining the mechanism of MreB action. Lastly, it will be interesting to determine how much similarity there is between different model organisms. As noted in Chapter 1, the regulation of FtsZ positioning in E.coli, B.subtilis, and Caulobacter is thematically similar, but there are important differences. A similar situation may exist for MreB. At almost every turn, this project has yielded unexpected results. Such a project can be horribly frustrating at times and terribly interesting and exciting at others. I anticipate that future work will continue to bring new fascinating surprises. The data and plausible mechanisms that I have discussed make experimentally testable predictions. The mutant strains and biochemical methods developed here are tools that should be valuable to future investigators. To all future scientists that pick up this story, I wish you the best of luck.

P a g e | 245

246 | P a g e

References

P a g e | 247

1. Dye, N.A. and L. Shapiro, The push and pull of the bacterial cytoskeleton. Trends Cell Biol, 2007. 17(5): p. 239-45. 2. Goley, E.D., A.A. Iniesta, and L. Shapiro, Cell cycle regulation in Caulobacter: location, location, location. J Cell Sci, 2007. 120(Pt 20): p. 3501-7. 3. Matsumoto, K., et al., Lipid domains in bacterial membranes. Mol Microbiol, 2006. 61(5): p. 1110-7. 4. Pogliano, J., Dynamic cellular location of bacterial plasmids. Curr Opin Microbiol, 2002. 5(6): p. 586-90. 5. Shapiro, L., H.H. McAdams, and R. Losick, Generating and exploiting polarity in bacteria. Science, 2002. 298(5600): p. 1942-6. 6. Shapiro, L., H.H. McAdams, and R. Losick, Why and how bacteria localize proteins. Science, 2009. 326(5957): p. 1225-8. 7. Lowe, J. and L.A. Amos, Crystal structure of the bacterial cell-division protein FtsZ. Nature, 1998. 391(6663): p. 203-6. 8. van den Ent, F., L.A. Amos, and J. Lowe, Prokaryotic origin of the actin cytoskeleton. Nature, 2001. 413(6851): p. 39-44. 9. van den Ent, F. and J. Lowe, Crystal structure of the cell division protein FtsA from Thermotoga maritima. EMBO J, 2000. 19(20): p. 5300-7. 10. van den Ent, F., et al., F-actin-like filaments formed by plasmid segregation protein ParM. Embo J, 2002. 21(24): p. 6935-43. 11. Young, K.D., The selective value of bacterial shape. Microbiol Mol Biol Rev, 2006. 70(3): p. 660-703. 12. Ghuysen, J.M. and R. Hackenbeck, Bacterial cell wall. New comprehensive biochemistry v. 27. 1994, Amsterdam ; New York: Elsevier. xxiv, 581 p. 13. Holtje, J.V., Growth of the stress-bearing and shape-maintaining murein sacculus of Escherichia coli. Microbiol Mol Biol Rev, 1998. 62(1): p. 181-203. 14. Koch, A.L., Bacterial growth and form. 2nd ed. 2001, Dordrecht ; Boston: Kluwer Academic Publishers. 470 p. 15. Neidhardt, F.C. and R. Curtiss, Escherichia coli and Salmonella : cellular and molecular biology. 2nd ed. 1996, Washington, D.C.: ASM Press. 2 v. (xx, 2822 p.). 16. Cabeen, M.T. and C. Jacobs-Wagner, Skin and bones: the bacterial cytoskeleton, cell wall, and cell morphogenesis. J Cell Biol, 2007. 179(3): p. 381-7. 17. Cabeen, M.T. and C. Jacobs-Wagner, Bacterial cell shape. Nat Rev Microbiol, 2005. 3(8): p. 601-10. 18. Gan, L., S. Chen, and G.J. Jensen, Molecular organization of Gram-negative peptidoglycan. Proc Natl Acad Sci U S A, 2008. 105(48): p. 18953-7. 19. Meroueh, S.O., et al., Three-dimensional structure of the bacterial cell wall peptidoglycan. Proc Natl Acad Sci U S A, 2006. 103(12): p. 4404-9. 20. Vollmer, W. and J.V. Holtje, The architecture of the murein (peptidoglycan) in gram-negative bacteria: vertical scaffold or horizontal layer(s)? J Bacteriol, 2004. 186(18): p. 5978-87. 21. Labischinski, H., et al., Conformational and topological aspects of the three- dimensional architecture of bacterial peptidoglycan. Ann Inst Pasteur Microbiol, 1985. 136A(1): p. 45-50. 22. Verwer, R.W., et al., Oriented fragmentation of Escherichia coli sacculi by

248 | P a g e

sonication. J Bacteriol, 1980. 141(1): p. 327-32. 23. Verwer, R.W., et al., Arrangement of glycan chains in the sacculus of Escherichia coli. J Bacteriol, 1978. 136(2): p. 723-9. 24. Cooper, S., Bacterial growth and division : biochemistry and regulation of prokaryotic and eukaryotic division cycles. 1991, San Diego: Academic Press. xxv, 501 p. 25. Holtje, J.V., A hypothetical holoenzyme involved in the replication of the murein sacculus of Escherichia coli. Microbiology, 1996. 142 ( Pt 8): p. 1911-8. 26. Alaedini, A. and R.A. Day, Identification of two penicillin-binding multienzyme complexes in Haemophilus influenzae. Biochem Biophys Res Commun, 1999. 264(1): p. 191-5. 27. Bertsche, U., et al., Interaction between two murein (peptidoglycan) synthases, PBP3 and PBP1B, in Escherichia coli. Mol Microbiol, 2006. 61(3): p. 675-90. 28. Schwarz, U. and W. Leutgeb, Morphogenetic aspects of murein structure and biosynthesis. J Bacteriol, 1971. 106(2): p. 588-95. 29. Koch, A.L., How bacteria grow and divide in spite of internal hydrostatic pressure. Can J Microbiol, 1985. 31(12): p. 1071-84. 30. Koch, A.L., The bacterium's way for safe enlargement and division. Appl Environ Microbiol, 2000. 66(9): p. 3657-63. 31. Moore, R.L., The biology of Hyphomicrobium and other prosthecate, budding bacteria. Annu Rev Microbiol, 1981. 35: p. 567-94. 32. Staley, J.T., Prosthecomicrobium and Ancalomicrobium: new prosthecate freshwater bacteria. J Bacteriol, 1968. 95(5): p. 1921-42. 33. Whittenbury, R. and C.S. Dow, Morphogenesis and differentiation in Rhodomicrobium vannielii and other budding and prosthecate bacteria. Bacteriol Rev, 1977. 41(3): p. 754-808. 34. Satta, G., R. Fontana, and P. Canepari, The two-competing site (TCS) model for cell shape regulation in bacteria: the envelope as an integration point for the regulatory circuits of essential physiological events. Adv Microb Physiol, 1994. 36: p. 181-245. 35. Begg, K.J. and W.D. Donachie, Cell shape and division in Escherichia coli: experiments with shape and division mutants. J Bacteriol, 1985. 163(2): p. 615- 22. 36. Spratt, B.G., Distinct penicillin binding proteins involved in the division, elongation, and shape of Escherichia coli K12. Proc Natl Acad Sci U S A, 1975. 72(8): p. 2999-3003. 37. Dai, K. and J. Lutkenhaus, ftsZ is an essential cell division gene in Escherichia coli. J Bacteriol, 1991. 173(11): p. 3500-6. 38. Wachi, M., et al., Mutant isolation and molecular cloning of mre genes, which determine cell shape, sensitivity to mecillinam, and amount of penicillin-binding proteins in Escherichia coli. J Bacteriol, 1987. 169(11): p. 4935-40. 39. Figge, R.M., A.V. Divakaruni, and J.W. Gober, MreB, the cell shape-determining bacterial actin homologue, co-ordinates cell wall morphogenesis in Caulobacter crescentus. Mol Microbiol, 2004. 51(5): p. 1321-32. 40. Gitai, Z., et al., MreB actin-mediated segregation of a specific region of a bacterial chromosome. Cell, 2005. 120(3): p. 329-41.

P a g e | 249

41. Wang, Y., B.D. Jones, and Y.V. Brun, A set of ftsZ mutants blocked at different stages of cell division in Caulobacter. Mol Microbiol, 2001. 40(2): p. 347-60. 42. de Pedro, M.A., et al., Murein segregation in Escherichia coli. J Bacteriol, 1997. 179(9): p. 2823-34. 43. Daniel, R.A. and J. Errington, Control of cell morphogenesis in bacteria: two distinct ways to make a rod-shaped cell. Cell, 2003. 113(6): p. 767-76. 44. Tiyanont, K., et al., Imaging peptidoglycan biosynthesis in Bacillus subtilis with fluorescent antibiotics. Proc Natl Acad Sci U S A, 2006. 103(29): p. 11033-8. 45. Markiewicz, Z., B. Glauner, and U. Schwarz, Murein structure and lack of DD- and LD-carboxypeptidase activities in Caulobacter crescentus. J Bacteriol, 1983. 156(2): p. 649-55. 46. Scheffers, D.J., L.J. Jones, and J. Errington, Several distinct localization patterns for penicillin-binding proteins in Bacillus subtilis. Mol Microbiol, 2004. 51(3): p. 749-64. 47. Scheffers, D.J. and M.G. Pinho, Bacterial cell wall synthesis: new insights from localization studies. Microbiol Mol Biol Rev, 2005. 69(4): p. 585-607. 48. Dougherty, T.J., et al., Direct quantitation of the number of individual penicillin- binding proteins per cell in Escherichia coli. J Bacteriol, 1996. 178(21): p. 6110- 5. 49. Rothfield, L., A. Taghbalout, and Y.L. Shih, Spatial control of bacterial division- site placement. Nat Rev Microbiol, 2005. 3(12): p. 959-68. 50. Thanbichler, M. and L. Shapiro, Getting organized--how bacterial cells move proteins and DNA. Nat Rev Microbiol, 2008. 6(1): p. 28-40. 51. Errington, J., R.A. Daniel, and D.J. Scheffers, Cytokinesis in bacteria. Microbiol Mol Biol Rev, 2003. 67(1): p. 52-65, table of contents. 52. Harry, E., L. Monahan, and L. Thompson, Bacterial cell division: the mechanism and its precison. Int Rev Cytol, 2006. 253: p. 27-94. 53. Thanbichler, M. and L. Shapiro, MipZ, a spatial regulator coordinating chromosome segregation with cell division in Caulobacter. Cell, 2006. 126(1): p. 147-62. 54. Adler, H.I., et al., MINIATURE escherichia coli CELLS DEFICIENT IN DNA. Proc Natl Acad Sci U S A, 1967. 57(2): p. 321-326. 55. Lutkenhaus, J., Assembly dynamics of the bacterial MinCDE system and spatial regulation of the Z ring. Annu Rev Biochem, 2007. 76: p. 539-62. 56. Lutkenhaus, J., Min oscillation in bacteria. Adv Exp Med Biol, 2008. 641: p. 49- 61. 57. Raskin, D.M. and P.A. de Boer, MinDE-dependent pole-to-pole oscillation of division inhibitor MinC in Escherichia coli. J Bacteriol, 1999. 181(20): p. 6419- 24. 58. Raskin, D.M. and P.A. de Boer, Rapid pole-to-pole oscillation of a protein required for directing division to the middle of Escherichia coli. Proc Natl Acad Sci U S A, 1999. 96(9): p. 4971-6. 59. Shih, Y.L., T. Le, and L. Rothfield, Division site selection in Escherichia coli involves dynamic redistribution of Min proteins within coiled structures that extend between the two cell poles. Proc Natl Acad Sci U S A, 2003. 100(13): p. 7865-70.

250 | P a g e

60. Hu, Z. and J. Lutkenhaus, Topological regulation of cell division in Escherichia coli involves rapid pole to pole oscillation of the division inhibitor MinC under the control of MinD and MinE. Mol Microbiol, 1999. 34(1): p. 82-90. 61. Howard, M. and K. Kruse, Cellular organization by self-organization: mechanisms and models for Min protein dynamics. J Cell Biol, 2005. 168(4): p. 533-6. 62. Howard, M. and A.D. Rutenberg, Pattern formation inside bacteria: fluctuations due to the low copy number of proteins. Phys Rev Lett, 2003. 90(12): p. 128102. 63. Huang, K.C., Y. Meir, and N.S. Wingreen, Dynamic structures in Escherichia coli: spontaneous formation of MinE rings and MinD polar zones. Proc Natl Acad Sci U S A, 2003. 100(22): p. 12724-8. 64. Kruse, K., M. Howard, and W. Margolin, An experimentalist's guide to computational modelling of the Min system. Mol Microbiol, 2007. 63(5): p. 1279- 84. 65. Kulkarni, R.V., et al., Pattern formation within Escherichia coli: diffusion, membrane attachment, and self-interaction of MinD molecules. Phys Rev Lett, 2004. 93(22): p. 228103. 66. Meinhardt, H. and P.A. de Boer, Pattern formation in Escherichia coli: a model for the pole-to-pole oscillations of Min proteins and the localization of the division site. Proc Natl Acad Sci U S A, 2001. 98(25): p. 14202-7. 67. Loose, M., et al., Spatial regulators for bacterial cell division self-organize into surface waves in vitro. Science, 2008. 320(5877): p. 789-92. 68. Edwards, D.H. and J. Errington, The Bacillus subtilis DivIVA protein targets to the division septum and controls the site specificity of cell division. Mol Microbiol, 1997. 24(5): p. 905-15. 69. Levin, P.A., et al., Identification of Bacillus subtilis genes for septum placement and shape determination. J Bacteriol, 1992. 174(21): p. 6717-28. 70. Marston, A.L., et al., Polar localization of the MinD protein of Bacillus subtilis and its role in selection of the mid-cell division site. Genes Dev, 1998. 12(21): p. 3419-30. 71. Varley, A.W. and G.C. Stewart, The divIVB region of the Bacillus subtilis chromosome encodes homologs of Escherichia coli septum placement (minCD) and cell shape (mreBCD) determinants. J Bacteriol, 1992. 174(21): p. 6729-42. 72. Bernhardt, T.G. and P.A. de Boer, SlmA, a nucleoid-associated, FtsZ binding protein required for blocking septal ring assembly over Chromosomes in E. coli. Mol Cell, 2005. 18(5): p. 555-64. 73. Wu, L.J. and J. Errington, Coordination of cell division and chromosome segregation by a nucleoid occlusion protein in Bacillus subtilis. Cell, 2004. 117(7): p. 915-25. 74. Kocks, C., et al., L. monocytogenes-induced actin assembly requires the actA gene product, a surface protein. Cell, 1992. 68(3): p. 521-31. 75. Kocks, C., et al., Polarized distribution of Listeria monocytogenes surface protein ActA at the site of directional actin assembly. J Cell Sci, 1993. 105 ( Pt 3): p. 699- 710. 76. Tilney, L.G., P.S. Connelly, and D.A. Portnoy, Actin filament nucleation by the bacterial pathogen, Listeria monocytogenes. J Cell Biol, 1990. 111(6 Pt 2): p.

P a g e | 251

2979-88. 77. Tilney, L.G. and D.A. Portnoy, Actin filaments and the growth, movement, and spread of the intracellular bacterial parasite, Listeria monocytogenes. J Cell Biol, 1989. 109(4 Pt 1): p. 1597-608. 78. Rafelski, S.M. and J.A. Theriot, Bacterial shape and ActA distribution affect initiation of Listeria monocytogenes actin-based motility. Biophys J, 2005. 89(3): p. 2146-58. 79. Smith, G.A., D.A. Portnoy, and J.A. Theriot, Asymmetric distribution of the Listeria monocytogenes ActA protein is required and sufficient to direct actin- based motility. Mol Microbiol, 1995. 17(5): p. 945-51. 80. Jensen, R.B. and L. Shapiro, The Caulobacter crescentus smc gene is required for cell cycle progression and chromosome segregation. Proc Natl Acad Sci U S A, 1999. 96(19): p. 10661-6. 81. Thanbichler, M. and L. Shapiro, Chromosome organization and segregation in bacteria. J Struct Biol, 2006. 156(2): p. 292-303. 82. Viollier, P.H., et al., Rapid and sequential movement of individual chromosomal loci to specific subcellular locations during bacterial DNA replication. Proc Natl Acad Sci U S A, 2004. 101(25): p. 9257-62. 83. Kawai, F., et al., Cardiolipin enrichment in spore membranes and its involvement in germination of Bacillus subtilis Marburg. Genes Genet Syst, 2006. 81(2): p. 69-76. 84. Kawai, F., et al., Cardiolipin domains in Bacillus subtilis marburg membranes. J Bacteriol, 2004. 186(5): p. 1475-83. 85. Mileykovskaya, E. and W. Dowhan, Visualization of phospholipid domains in Escherichia coli by using the cardiolipin-specific fluorescent dye 10-N-nonyl acridine orange. J Bacteriol, 2000. 182(4): p. 1172-5. 86. Huang, K.C., R. Mukhopadhyay, and N.S. Wingreen, A curvature-mediated mechanism for localization of lipids to bacterial poles. PLoS Comput Biol, 2006. 2(11): p. e151. 87. Ramamurthi, K.S., et al., Geometric cue for protein localization in a bacterium. Science, 2009. 323(5919): p. 1354-7. 88. Ramamurthi, K.S. and R. Losick, Negative membrane curvature as a cue for subcellular localization of a bacterial protein. Proc Natl Acad Sci U S A, 2009. 106(32): p. 13541-5. 89. Rudner, D.Z., Q. Pan, and R.M. Losick, Evidence that subcellular localization of a bacterial membrane protein is achieved by diffusion and capture. Proc Natl Acad Sci U S A, 2002. 99(13): p. 8701-6. 90. Charles, M., et al., Polar targeting of Shigella virulence factor IcsA in Enterobacteriacae and Vibrio. Proc Natl Acad Sci U S A, 2001. 98(17): p. 9871- 6. 91. Robbins, J.R., et al., The making of a gradient: IcsA (VirG) polarity in Shigella flexneri. Mol Microbiol, 2001. 41(4): p. 861-72. 92. Shih, Y.L. and L. Rothfield, The bacterial cytoskeleton. Microbiol Mol Biol Rev, 2006. 70(3): p. 729-54. 93. Lutkenhaus, J.F., H. Wolf-Watz, and W.D. Donachie, Organization of genes in the ftsA-envA region of the Escherichia coli genetic map and identification of a

252 | P a g e

new fts locus (ftsZ). J Bacteriol, 1980. 142(2): p. 615-20. 94. Ward, J.E., Jr. and J. Lutkenhaus, Overproduction of FtsZ induces minicell formation in E. coli. Cell, 1985. 42(3): p. 941-9. 95. Erickson, H.P., Atomic structures of tubulin and FtsZ. Trends Cell Biol, 1998. 8(4): p. 133-7. 96. Mukherjee, A. and J. Lutkenhaus, Guanine nucleotide-dependent assembly of FtsZ into filaments. J Bacteriol, 1994. 176(9): p. 2754-8. 97. Mukherjee, A. and J. Lutkenhaus, Dynamic assembly of FtsZ regulated by GTP hydrolysis. EMBO J, 1998. 17(2): p. 462-9. 98. Mukherjee, A. and J. Lutkenhaus, Analysis of FtsZ assembly by light scattering and determination of the role of divalent metal cations. J Bacteriol, 1999. 181(3): p. 823-32. 99. Oliva, M.A., S.C. Cordell, and J. Lowe, Structural insights into FtsZ protofilament formation. Nat Struct Mol Biol, 2004. 11(12): p. 1243-50. 100. Romberg, L. and P.A. Levin, Assembly dynamics of the bacterial cell division protein FTSZ: poised at the edge of stability. Annu Rev Microbiol, 2003. 57: p. 125-54. 101. Romberg, L., M. Simon, and H.P. Erickson, Polymerization of Ftsz, a bacterial homolog of tubulin. is assembly cooperative? J Biol Chem, 2001. 276(15): p. 11743-53. 102. Scheffers, D. and A.J. Driessen, The polymerization mechanism of the bacterial cell division protein FtsZ. FEBS Lett, 2001. 506(1): p. 6-10. 103. Erickson, H.P., Modeling the physics of FtsZ assembly and force generation. Proc Natl Acad Sci U S A, 2009. 106(23): p. 9238-43. 104. Bi, E.F. and J. Lutkenhaus, FtsZ ring structure associated with division in Escherichia coli. Nature, 1991. 354(6349): p. 161-4. 105. Li, Z., et al., The structure of FtsZ filaments in vivo suggests a force-generating role in cell division. EMBO J, 2007. 26(22): p. 4694-708. 106. Bork, P., C. Sander, and A. Valencia, An ATPase domain common to prokaryotic cell cycle proteins, sugar kinases, actin, and hsp70 heat shock proteins. Proc Natl Acad Sci U S A, 1992. 89(16): p. 7290-4. 107. Gerdes, K. and S. Molin, Partitioning of plasmid R1. Structural and functional analysis of the parA locus. J Mol Biol, 1986. 190(3): p. 269-79. 108. Jensen, R.B. and K. Gerdes, Partitioning of plasmid R1. The ParM protein exhibits ATPase activity and interacts with the centromere-like ParR-parC complex. J Mol Biol, 1997. 269(4): p. 505-13. 109. Moller-Jensen, J., et al., Bacterial mitosis: ParM of plasmid R1 moves plasmid DNA by an actin-like insertional polymerization mechanism. Mol Cell, 2003. 12(6): p. 1477-87. 110. Moller-Jensen, J., R.B. Jensen, and K. Gerdes, Plasmid and chromosome segregation in prokaryotes. Trends Microbiol, 2000. 8(7): p. 313-20. 111. Garner, E.C., C.S. Campbell, and R.D. Mullins, Dynamic instability in a DNA- segregating prokaryotic actin homolog. Science, 2004. 306(5698): p. 1021-5. 112. Campbell, C.S. and R.D. Mullins, In vivo visualization of type II plasmid segregation: bacterial actin filaments pushing plasmids. J Cell Biol, 2007. 179(5): p. 1059-66.

P a g e | 253

113. Orlova, A., et al., The structure of bacterial ParM filaments. Nat Struct Mol Biol, 2007. 14(10): p. 921-6. 114. Feucht, A., et al., Cytological and biochemical characterization of the FtsA cell division protein of Bacillus subtilis. Mol Microbiol, 2001. 40(1): p. 115-25. 115. Lara, B., et al., Cell division in cocci: localization and properties of the Streptococcus pneumoniae FtsA protein. Mol Microbiol, 2005. 55(3): p. 699-711. 116. Jones, L.J., R. Carballido-Lopez, and J. Errington, Control of cell shape in bacteria: helical, actin-like filaments in Bacillus subtilis. Cell, 2001. 104(6): p. 913-22. 117. Gitai, Z., N. Dye, and L. Shapiro, An actin-like gene can determine cell polarity in bacteria. Proc Natl Acad Sci U S A, 2004. 101(23): p. 8643-8. 118. Madabhushi, R. and K.J. Marians, Actin homolog MreB affects chromosome segregation by regulating topoisomerase IV in Escherichia coli. Mol Cell, 2009. 33(2): p. 171-80. 119. Srivastava, P., et al., Changes in nucleoid morphology and origin localization upon inhibition or alteration of the actin homolog, MreB, of Vibrio cholerae. J Bacteriol, 2007. 189(20): p. 7450-63. 120. Wagner, J.K., C.D. Galvani, and Y.V. Brun, Caulobacter crescentus requires RodA and MreB for stalk synthesis and prevention of ectopic pole formation. J Bacteriol, 2005. 187(2): p. 544-53. 121. Kawai, Y., K. Asai, and J. Errington, Partial functional redundancy of MreB isoforms, MreB, Mbl and MreBH, in cell morphogenesis of Bacillus subtilis. Mol Microbiol, 2009. 73(4): p. 719-31. 122. Varma, A. and K.D. Young, In Escherichia coli, MreB and FtsZ direct the synthesis of lateral cell wall via independent pathways that require PBP 2. J Bacteriol, 2009. 191(11): p. 3526-33. 123. Ausmees, N., J.R. Kuhn, and C. Jacobs-Wagner, The bacterial cytoskeleton: an intermediate filament-like function in cell shape. Cell, 2003. 115(6): p. 705-13. 124. Cabeen, M.T., et al., Bacterial cell curvature through mechanical control of cell growth. EMBO J, 2009. 28(9): p. 1208-19. 125. Koonin, E.V., A superfamily of ATPases with diverse functions containing either classical or deviant ATP-binding motif. J Mol Biol, 1993. 229(4): p. 1165-74. 126. Gerdes, K., J. Moller-Jensen, and R. Bugge Jensen, Plasmid and chromosome partitioning: surprises from phylogeny. Mol Microbiol, 2000. 37(3): p. 455-66. 127. Barilla, D., et al., Bacterial DNA segregation dynamics mediated by the polymerizing protein ParF. Embo J, 2005. 24(7): p. 1453-64. 128. Lim, G.E., A.I. Derman, and J. Pogliano, Bacterial DNA segregation by dynamic SopA polymers. Proc Natl Acad Sci U S A, 2005. 102(49): p. 17658-63. 129. Bouet, J.Y., et al., Polymerization of SopA partition ATPase: regulation by DNA binding and SopB. Mol Microbiol, 2006. 130. Ebersbach, G., et al., Regular cellular distribution of plasmids by oscillating and filament-forming ParA ATPase of plasmid pB171. Mol Microbiol, 2006. 61(6): p. 1428-42. 131. Leonard, T.A., P.J. Butler, and J. Lowe, Bacterial chromosome segregation: structure and DNA binding of the Soj dimer--a conserved biological switch. Embo J, 2005. 24(2): p. 270-82.

254 | P a g e

132. Suefuji, K., R. Valluzzi, and D. RayChaudhuri, Dynamic assembly of MinD into filament bundles modulated by ATP, phospholipids, and MinE. Proc Natl Acad Sci U S A, 2002. 99(26): p. 16776-81. 133. Ebersbach, G. and K. Gerdes, Bacterial mitosis: partitioning protein ParA oscillates in spiral-shaped structures and positions plasmids at mid-cell. Mol Microbiol, 2004. 52(2): p. 385-98. 134. Marston, A.L. and J. Errington, Dynamic movement of the ParA-like Soj protein of B. subtilis and its dual role in nucleoid organization and developmental regulation. Mol Cell, 1999. 4(5): p. 673-82. 135. Adachi, S., K. Hori, and S. Hiraga, Subcellular positioning of F plasmid mediated by dynamic localization of SopA and SopB. J Mol Biol, 2006. 356(4): p. 850-63. 136. Becker, E., et al., DNA segregation by the bacterial actin AlfA during Bacillus subtilis growth and development. Embo J, 2006. 25(24): p. 5919-31. 137. Komeili, A., et al., Magnetosomes are cell membrane invaginations organized by the actin-like protein MamK. Science, 2006. 311(5758): p. 242-5. 138. Pradel, N., et al., Polar positional information in Escherichia coli spherical cells. Biochem Biophys Res Commun, 2007. 353(2): p. 493-500. 139. Pradel, N., et al., Biogenesis of actin-like bacterial cytoskeletal filaments destined for positioning prokaryotic magnetic organelles. Proc Natl Acad Sci U S A, 2006. 103(46): p. 17485-9. 140. Carballido-Lopez, R., et al., Actin homolog MreBH governs cell morphogenesis by localization of the cell wall hydrolase LytE. Dev Cell, 2006. 11(3): p. 399-409. 141. Kruse, T., J. Bork-Jensen, and K. Gerdes, The morphogenetic MreBCD proteins of Escherichia coli form an essential membrane-bound complex. Mol Microbiol, 2005. 55(1): p. 78-89. 142. Dye, N.A., et al., Two independent spiral structures control cell shape in Caulobacter. Proc Natl Acad Sci U S A, 2005. 102(51): p. 18608-13. 143. Divakaruni, A.V., et al., The cell-shape protein MreC interacts with extracytoplasmic proteins including cell wall assembly complexes in Caulobacter crescentus. Proc Natl Acad Sci U S A, 2005. 102(51): p. 18602-7. 144. Garner, E.C., et al., Reconstitution of DNA Segregation Driven by Assembly of a Prokaryotic Actin Homolog. Science, 2007. 315: p. 1270-1274. 145. Popp, D., et al., Concerning the dynamic instability of actin homolog ParM. Biochem Biophys Res Commun, 2007. 353(1): p. 109-14. 146. Fogel, M.A. and M.K. Waldor, A dynamic, mitotic-like mechanism for bacterial chromosome segregation. Genes Dev, 2006. 20(23): p. 3269-82. 147. Mohl, D.A. and J.W. Gober, Cell cycle-dependent polar localization of chromosome partitioning proteins in Caulobacter crescentus. Cell, 1997. 88(5): p. 675-84. 148. Moller-Jensen, J., et al., Prokaryotic DNA segregation by an actin-like filament. Embo J, 2002. 21(12): p. 3119-27. 149. Theriot, J.A., The polymerization motor. Traffic, 2000. 1(1): p. 19-28. 150. Ebersbach, G. and K. Gerdes, Plasmid segregation mechanisms. Annu Rev Genet, 2005. 39: p. 453-79. 151. Jensen, R.B. and K. Gerdes, Mechanism of DNA segregation in prokaryotes: ParM partitioning protein of plasmid R1 co-localizes with its replicon during the

P a g e | 255

cell cycle. Embo J, 1999. 18(14): p. 4076-84. 152. Westermann, S., et al., The Dam1 kinetochore ring complex moves processively on depolymerizing microtubule ends. Nature, 2006. 440(7083): p. 565-9. 153. Fiebig, A., K. Keren, and J.A. Theriot, Fine-scale time-lapse analysis of the biphasic, dynamic behaviour of the two Vibrio cholerae chromosomes. Mol Microbiol, 2006. 60(5): p. 1164-78. 154. Fogel, M.A. and M.K. Waldor, Distinct segregation dynamics of the two Vibrio cholerae chromosomes. Mol Microbiol, 2005. 55(1): p. 125-36. 155. Briegel, A., et al., Multiple large filament bundles observed in Caulobacter crescentus by electron cryotomography. Mol Microbiol, 2006. 62(1): p. 5-14. 156. Thompson, S.R., G.H. Wadhams, and J.P. Armitage, The positioning of cytoplasmic protein clusters in bacteria. Proc Natl Acad Sci U S A, 2006. 103(21): p. 8209-14. 157. Paredez, A.R., C.R. Somerville, and D.W. Ehrhardt, Visualization of cellulose synthase demonstrates functional association with microtubules. Science, 2006. 312(5779): p. 1491-5. 158. Sun, Q. and W. Margolin, FtsZ dynamics during the division cycle of live Escherichia coli cells. J Bacteriol, 1998. 180(8): p. 2050-6. 159. Jensen, S.O., L.S. Thompson, and E.J. Harry, Cell division in Bacillus subtilis: FtsZ and FtsA association is Z-ring independent, and FtsA is required for efficient midcell Z-Ring assembly. J Bacteriol, 2005. 187(18): p. 6536-44. 160. Pichoff, S. and J. Lutkenhaus, Unique and overlapping roles for ZipA and FtsA in septal ring assembly in Escherichia coli. Embo J, 2002. 21(4): p. 685-93. 161. Pichoff, S. and J. Lutkenhaus, Tethering the Z ring to the membrane through a conserved membrane targeting sequence in FtsA. Mol Microbiol, 2005. 55(6): p. 1722-34. 162. Corbin, B.D., et al., Z-ring-independent interaction between a subdomain of FtsA and late septation proteins as revealed by a polar recruitment assay. J Bacteriol, 2004. 186(22): p. 7736-44. 163. Rico, A.I., et al., Role of two essential domains of Escherichia coli FtsA in localization and progression of the division ring. Mol Microbiol, 2004. 53(5): p. 1359-71. 164. Hamoen, L.W., et al., SepF, a novel FtsZ-interacting protein required for a late step in cell division. Mol Microbiol, 2006. 59(3): p. 989-99. 165. Ishikawa, S., et al., A new FtsZ-interacting protein, YlmF, complements the activity of FtsA during progression of cell division in Bacillus subtilis. Mol Microbiol, 2006. 60(6): p. 1364-80. 166. Anderson, D.E., F.J. Gueiros-Filho, and H.P. Erickson, Assembly dynamics of FtsZ rings in Bacillus subtilis and Escherichia coli and effects of FtsZ-regulating proteins. J Bacteriol, 2004. 186(17): p. 5775-81. 167. Thanedar, S. and W. Margolin, FtsZ exhibits rapid movement and oscillation waves in helix-like patterns in Escherichia coli. Curr Biol, 2004. 14(13): p. 1167- 73. 168. Ben-Yehuda, S. and R. Losick, Asymmetric cell division in B. subtilis involves a spiral-like intermediate of the cytokinetic protein FtsZ. Cell, 2002. 109(2): p. 257- 66.

256 | P a g e

169. Mukherjee, A., C. Cao, and J. Lutkenhaus, Inhibition of FtsZ polymerization by SulA, an inhibitor of septation in Escherichia coli. Proc Natl Acad Sci U S A, 1998. 95(6): p. 2885-90. 170. Bazylinski, D.A. and R.B. Frankel, Magnetosome formation in prokaryotes. Nat Rev Microbiol, 2004. 2(3): p. 217-30. 171. Mitchison, T.J., Evolution of a dynamic cytoskeleton. Philos Trans R Soc Lond B Biol Sci, 1995. 349(1329): p. 299-304. 172. Lan, G., et al., Condensation of FtsZ filaments can drive bacterial cell division. Proc Natl Acad Sci U S A, 2009. 106(1): p. 121-6. 173. Lleo, M.M., P. Canepari, and G. Satta, Bacterial cell shape regulation: testing of additional predictions unique to the two-competing-sites model for peptidoglycan assembly and isolation of conditional rod-shaped mutants from some wild-type cocci. J Bacteriol, 1990. 172(7): p. 3758-71. 174. Moller-Jensen, J. and J. Lowe, Increasing complexity of the bacterial cytoskeleton. Curr Opin Cell Biol, 2005. 17(1): p. 75-81. 175. Doi, M., et al., Determinations of the DNA sequence of the mreB gene and of the gene products of the mre region that function in formation of the rod shape of Escherichia coli cells. J Bacteriol, 1988. 170(10): p. 4619-24. 176. Kruse, T., et al., Dysfunctional MreB inhibits chromosome segregation in Escherichia coli. Embo J, 2003. 22(19): p. 5283-92. 177. Carballido-Lopez, R. and J. Errington, The bacterial cytoskeleton: in vivo dynamics of the actin-like protein Mbl of Bacillus subtilis. Dev Cell, 2003. 4(1): p. 19-28. 178. Defeu Soufo, H.J. and P.L. Graumann, Dynamic movement of actin-like proteins within bacterial cells. EMBO Rep, 2004. 5(8): p. 789-94. 179. Lee, J.C., et al., Heterospecific expression of the Bacillus subtilis cell shape determination genes mreBCD in Escherichia coli. Curr Microbiol, 2003. 47(2): p. 146-52. 180. Leaver, M. and J. Errington, Roles for MreC and MreD proteins in helical growth of the cylindrical cell wall in Bacillus subtilis. Mol Microbiol, 2005. 57(5): p. 1196-209. 181. Soufo, H.J. and P.L. Graumann, Actin-like proteins MreB and Mbl from Bacillus subtilis are required for bipolar positioning of replication origins. Curr Biol, 2003. 13(21): p. 1916-20. 182. Lee, J.C. and G.C. Stewart, Essential nature of the mreC determinant of Bacillus subtilis. J Bacteriol, 2003. 185(15): p. 4490-8. 183. Skerker, J.M. and M.T. Laub, Cell-cycle progression and the generation of asymmetry in Caulobacter crescentus. Nat Rev Microbiol, 2004. 2(4): p. 325-37. 184. Meisenzahl, A.C., L. Shapiro, and U. Jenal, Isolation and characterization of a xylose-dependent promoter from Caulobacter crescentus. J Bacteriol, 1997. 179(3): p. 592-600. 185. Divakaruni, A.V., et al., The Cell-Shape Protein MreC Interacts with Extracytoplasmic Proteins Including Cell Wall Assembly Complexes in Caulobacter crescentus. Proc Natl Acad Sci U S A, 2005. 186. Shaner, N.C., et al., Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein. Nat Biotechnol,

P a g e | 257

2004. 22(12): p. 1567-72. 187. Iwai, N., K. Nagai, and M. Wachi, Novel S-benzylisothiourea compound that induces spherical cells in Escherichia coli probably by acting on a rod-shape- determining protein(s) other than penicillin-binding protein 2. Biosci Biotechnol Biochem, 2002. 66(12): p. 2658-62. 188. Quardokus, E., N. Din, and Y.V. Brun, Cell cycle regulation and cell type-specific localization of the FtsZ division initiation protein in Caulobacter. Proc Natl Acad Sci U S A, 1996. 93(13): p. 6314-9. 189. Leventon, M.E., Grimson, W.E.L., Faugeras, O., Statistical shape influence in geodesic active contours. Proc. IEEE Conf. Computer Vision and Pattern Recognition, 2000. 1: p. 316-323. 190. Tsai, A., et al., A shape-based approach to the segmentation of medical imagery using level sets. IEEE Trans Med Imaging, 2003. 22(2): p. 137-54. 191. Levy, S. Interactive 3-D Visualization of Particle Systems with Partiview. in Astrophysical Supercomputing Using Particle Simulations. 2001: Astronomical Society of the Pacific. 192. Espeli, O., et al., SetB: an integral membrane protein that affects chromosome segregation in Escherichia coli. Mol Microbiol, 2003. 50(2): p. 495-509. 193. Ghosh, A.S. and K.D. Young, Helical disposition of proteins and lipopolysaccharide in the outer membrane of Escherichia coli. J Bacteriol, 2005. 187(6): p. 1913-22. 194. Morlot, C., et al., The D,D-carboxypeptidase PBP3 organizes the division process of Streptococcus pneumoniae. Mol Microbiol, 2004. 51(6): p. 1641-8. 195. Pinho, M.G. and J. Errington, Recruitment of penicillin-binding protein PBP2 to the division site of Staphylococcus aureus is dependent on its transpeptidation substrates. Mol Microbiol, 2005. 55(3): p. 799-807. 196. Ely, B., Genetics of Caulobacter crescentus. Methods Enzymol, 1991. 204: p. 372-84. 197. Evinger, M. and N. Agabian, Envelope-associated nucleoid from Caulobacter crescentus stalked and swarmer cells. J Bacteriol, 1977. 132(1): p. 294-301. 198. Good, P., Permutation tests: A practical guide to resampling methods for testing hypotheses. 1994, New York: Springer-Verlag. 199. Pitman, E., Significance tests which may be applied to samples from any populations. J Royal Statist Soc, 1937. 4: p. 119-130. 200. Wheeler, R.T. and L. Shapiro, Differential localization of two histidine kinases controlling bacterial cell differentiation. Mol Cell, 1999. 4(5): p. 683-94. 201. Abramoff, M.D., Magelhaes, P.J., Ram, S.J., Image Processing with ImageJ. Biophotonics International, 2004. 11(7): p. 36-42. 202. Yoo, T.S., M.J. Ackerman, W.E. Lorensen, W. Schroeder, V. Chalana, S. Aylward, D. Metaxes, R. Whitaker., Engineering and Algorithm Design for an Image Processing API: A Technical Report on ITK-- The Insight Toolkit., in Proc. of Medicine Meets Virtual Reality, J. Westward, Editor. 2002, IOS Press: Amsterdam. p. 586-592. 203. Ye, Q.Z. Proceedings of the 9th International Conference on Pattern Recongnition (ICPR '88). 1988. Rome, Italy: IEEE Computer Society Press. 204. Carballido-Lopez, R., The bacterial actin-like cytoskeleton. Microbiol Mol Biol

258 | P a g e

Rev, 2006. 70(4): p. 888-909. 205. Divakaruni, A.V., et al., The cell shape proteins MreB and MreC control cell morphogenesis by positioning cell wall synthetic complexes. Mol Microbiol, 2007. 66(1): p. 174-88. 206. Kawai, Y., R.A. Daniel, and J. Errington, Regulation of cell wall morphogenesis in Bacillus subtilis by recruitment of PBP1 to the MreB helix. Mol Microbiol, 2009. 71(5): p. 1131-44. 207. White, C.L., A. Kitich, and J.W. Gober, Positioning cell wall synthetic complexes by the bacterial morphogenetic proteins MreB and MreD. Mol Microbiol. 208. Bean, G.J., et al., A22 disrupts the bacterial actin cytoskeleton by directly binding and inducing a low-affinity state in MreB. Biochemistry, 2009. 48(22): p. 4852-7. 209. Thomas, J.H., N.F. Neff, and D. Botstein, Isolation and characterization of mutations in the beta-tubulin gene of Saccharomyces cerevisiae. Genetics, 1985. 111(4): p. 715-34. 210. Dorn, J.F., et al., Yeast kinetochore microtubule dynamics analyzed by high- resolution three-dimensional microscopy. Biophys J, 2005. 89(4): p. 2835-54. 211. Aaron, M., et al., The tubulin homologue FtsZ contributes to cell elongation by guiding cell wall precursor synthesis in Caulobacter crescentus. Mol Microbiol, 2007. 64(4): p. 938-52. 212. Pincus, Z. and J.A. Theriot, Comparison of quantitative methods for cell-shape analysis. J Microsc, 2007. 227(Pt 2): p. 140-56. 213. Charbon, G., M.T. Cabeen, and C. Jacobs-Wagner, Bacterial intermediate filaments: in vivo assembly, organization, and dynamics of crescentin. Genes Dev, 2009. 23(9): p. 1131-44. 214. van den Ent, F., et al., Bacterial actin MreB assembles in complex with cell shape protein RodZ. EMBO J. 29(6): p. 1081-90. 215. Takacs, C.N., et al., MreB drives de novo rod morphogenesis in Caulobacter crescentus via remodeling of the cell wall. J Bacteriol. 192(6): p. 1671-84. 216. Young, K.D., Bacterial shape. Mol Microbiol, 2003. 49(3): p. 571-80. 217. McGrath, P.T., et al., A dynamically localized protease complex and a polar specificity factor control a cell cycle master regulator. Cell, 2006. 124(3): p. 535- 47. 218. Werner, J.N., et al., Quantitative genome-scale analysis of protein localization in an asymmetric bacterium. Proc Natl Acad Sci U S A, 2009. 106(19): p. 7858-63. 219. Shebelut, C.W., R.B. Jensen, and Z. Gitai, Growth conditions regulate the requirements for Caulobacter chromosome segregation. J Bacteriol, 2009. 191(3): p. 1097-100. 220. Bowman, G.R., et al., Caulobacter PopZ forms a polar subdomain dictating sequential changes in pole composition and function. Mol Microbiol. 221. Bowman, G.R., et al., A polymeric protein anchors the chromosomal origin/ParB complex at a bacterial cell pole. Cell, 2008. 134(6): p. 945-55. 222. Ebersbach, G., et al., A self-associating protein critical for chromosome attachment, division, and polar organization in caulobacter. Cell, 2008. 134(6): p. 956-68. 223. Alyahya, S.A., et al., RodZ, a component of the bacterial core morphogenic apparatus. Proc Natl Acad Sci U S A, 2009. 106(4): p. 1239-44.

P a g e | 259

224. Bendezu, F.O., et al., RodZ (YfgA) is required for proper assembly of the MreB actin cytoskeleton and cell shape in E. coli. Embo J, 2009. 28(3): p. 193-204. 225. Shiomi, D., M. Sakai, and H. Niki, Determination of bacterial rod shape by a novel cytoskeletal membrane protein. Embo J, 2008. 27(23): p. 3081-91. 226. Kabsch, W. and K.C. Holmes, The actin fold. FASEB J, 1995. 9(2): p. 167-74. 227. Defeu Soufo, H.J. and P.L. Graumann, Dynamic localization and interaction with other Bacillus subtilis actin-like proteins are important for the function of MreB. Mol Microbiol, 2006. 62(5): p. 1340-56. 228. Kim, S.Y., et al., Single molecules of the bacterial actin MreB undergo directed treadmilling motion in Caulobacter crescentus. Proc Natl Acad Sci U S A, 2006. 103(29): p. 10929-34. 229. Varma, A., M.A. de Pedro, and K.D. Young, FtsZ directs a second mode of peptidoglycan synthesis in Escherichia coli. J Bacteriol, 2007. 189(15): p. 5692- 704. 230. van den Ent, F., et al., Dimeric structure of the cell shape protein MreC and its functional implications. Mol Microbiol, 2006. 62(6): p. 1631-42. 231. Vats, P., Y.L. Shih, and L. Rothfield, Assembly of the MreB-associated cytoskeletal ring of Escherichia coli. Mol Microbiol, 2009. 72(1): p. 170-82. 232. Huang, K.C., et al., Cell shape and cell-wall organization in Gram-negative bacteria. Proc Natl Acad Sci U S A, 2008. 105(49): p. 19282-7. 233. Christen, B., et al., High-throughput identification of protein localization dependency networks. Proc Natl Acad Sci U S A. 107(10): p. 4681-6. 234. Thanbichler, M., A.A. Iniesta, and L. Shapiro, A comprehensive set of plasmids for vanillate- and xylose-inducible gene expression in Caulobacter crescentus. Nucleic Acids Res, 2007. 35(20): p. e137. 235. Rubner, Y., C. Tomasi, and L.J. Guibas, The Earth Mover's Distance as a metric for image retrieval. International Journal of Computer Vision, 2000. 40(2): p. 99- 121. 236. Cowles Commission for Research in Economics. and T.C. Koopmans, Activity analysis of production and allocation. Its Monograph. 1951, New York,: Wiley. 404 p. 237. Chernoff, H., Use of Faces to Represent Points in K-Dimensional Space Graphically. Journal of the American Statistical Association, 1973. 68(342): p. 361-368. 238. Cameron, L.A., et al., Motility of ActA protein-coated microspheres driven by actin polymerization. Proc Natl Acad Sci U S A, 1999. 96(9): p. 4908-13. 239. Loisel, T.P., et al., Reconstitution of actin-based motility of Listeria and Shigella using pure proteins. Nature, 1999. 401(6753): p. 613-6. 240. Theriot, J.A., et al., Involvement of in the actin-based motility of L. monocytogenes in cells and in cell-free extracts. Cell, 1994. 76(3): p. 505-17. 241. Bean, G.J. and K.J. Amann, Polymerization properties of the Thermotoga maritima actin MreB: roles of temperature, nucleotides, and ions. Biochemistry, 2008. 47(2): p. 826-35. 242. Esue, O., et al., The assembly of MreB, a prokaryotic homolog of actin. J Biol Chem, 2005. 280(4): p. 2628-35. 243. Esue, O., D. Wirtz, and Y. Tseng, GTPase activity, structure, and mechanical

260 | P a g e

properties of filaments assembled from bacterial cytoskeleton protein MreB. J Bacteriol, 2006. 188(3): p. 968-76. 244. Formstone, A. and J. Errington, A magnesium-dependent mreB null mutant: implications for the role of mreB in Bacillus subtilis. Mol Microbiol, 2005. 55(6): p. 1646-57. 245. Hu, B., et al., MreB is important for cell shape but not for chromosome segregation of the filamentous cyanobacterium Anabaena sp. PCC 7120. Mol Microbiol, 2007. 63(6): p. 1640-52. 246. Karczmarek, A., et al., DNA and origin region segregation are not affected by the transition from rod to sphere after inhibition of Escherichia coli MreB by A22. Mol Microbiol, 2007. 65(1): p. 51-63. 247. Mazza, P., et al., MreB of Streptomyces coelicolor is not essential for vegetative growth but is required for the integrity of aerial hyphae and spores. Mol Microbiol, 2006. 60(4): p. 838-52. 248. Jensen, G.J. and A. Briegel, How electron cryotomography is opening a new window onto prokaryotic ultrastructure. Curr Opin Struct Biol, 2007. 17(2): p. 260-7. 249. Murphy, G.E. and G.J. Jensen, Electron cryotomography. Biotechniques, 2007. 43(4): p. 413, 415, 417 passim. 250. Biteen, J.S., et al., Super-resolution imaging in live Caulobacter crescentus cells using photoswitchable EYFP. Nat Methods, 2008. 5(11): p. 947-9. 251. Defeu Soufo, H.J., et al., Bacterial translation elongation factor EF-Tu interacts and colocalizes with actin-like MreB protein. Proc Natl Acad Sci U S A. 107(7): p. 3163-8. 252. Demma, M., et al., Isolation of an abundant 50,000-dalton actin filament bundling protein from Dictyostelium amoebae. J Biol Chem, 1990. 265(4): p. 2286-91. 253. Dharmawardhane, S., et al., Compartmentalization and actin binding properties of ABP-50: the elongation factor-1 alpha of Dictyostelium. Cell Motil Cytoskeleton, 1991. 20(4): p. 279-88. 254. Edmonds, B.T., J. Murray, and J. Condeelis, pH regulation of the F-actin binding properties of Dictyostelium elongation factor 1 alpha. J Biol Chem, 1995. 270(25): p. 15222-30. 255. Itano, N. and S. Hatano, F-actin bundling protein from Physarum polycephalum: purification and its capacity for co-bundling of actin filaments and microtubules. Cell Motil Cytoskeleton, 1991. 19(4): p. 244-54. 256. Kurasawa, Y., et al., F-actin bundling activity of Tetrahymena elongation factor 1 alpha is regulated by Ca2+/calmodulin. J Biochem, 1996. 119(4): p. 791-8. 257. Kurasawa, Y., Y. Watanabe, and O. Numata, Characterization of F-actin bundling activity of Tetrahymena elongation factor 1 alpha investigated with rabbit skeletal muscle actin. Zoolog Sci, 1996. 13(3): p. 371-5. 258. Yang, F., et al., Identification of an actin-binding protein from Dictyostelium as elongation factor 1a. Nature, 1990. 347(6292): p. 494-6. 259. Andreu, J.M., M.A. Oliva, and O. Monasterio, Reversible unfolding of FtsZ cell division proteins from archaea and bacteria. Comparison with eukaryotic tubulin folding and assembly. J Biol Chem, 2002. 277(45): p. 43262-70.

P a g e | 261

260. De La Cruz, E.M., et al., Polymerization and structure of nucleotide-free actin filaments. J Mol Biol, 2000. 295(3): p. 517-26. 261. De La Cruz, E.M. and T.D. Pollard, Nucleotide-free actin: stabilization by sucrose and nucleotide binding kinetics. Biochemistry, 1995. 34(16): p. 5452-61. 262. Taoka, A., et al., Polymerization of the actin-like protein MamK, which is associated with magnetosomes. J Bacteriol, 2007. 189(23): p. 8737-40. 263. Mayer, J.A. and K.J. Amann, Assembly properties of the Bacillus subtilis actin, MreB. Cell Motil Cytoskeleton, 2009. 66(2): p. 109-18. 264. Stricker, J., et al., Rapid assembly dynamics of the Escherichia coli FtsZ-ring demonstrated by fluorescence recovery after photobleaching. Proc Natl Acad Sci U S A, 2002. 99(5): p. 3171-5. 265. Michie, K.A., et al., Trapping of a spiral-like intermediate of the bacterial cytokinetic protein FtsZ. J Bacteriol, 2006. 188(5): p. 1680-90. 266. Stricker, J. and H.P. Erickson, In vivo characterization of Escherichia coli ftsZ mutants: effects on Z-ring structure and function. J Bacteriol, 2003. 185(16): p. 4796-805. 267. Grantcharova, N., U. Lustig, and K. Flardh, Dynamics of FtsZ assembly during sporulation in Streptomyces coelicolor A3(2). J Bacteriol, 2005. 187(9): p. 3227- 37. 268. Vats, P. and L. Rothfield, Duplication and segregation of the actin (MreB) cytoskeleton during the prokaryotic cell cycle. Proc Natl Acad Sci U S A, 2007. 104(45): p. 17795-800. 269. Srinivasan, R., et al., Filament formation of the Escherichia coli actin-related protein, MreB, in fission yeast. Curr Biol, 2007. 17(3): p. 266-72. 270. Gerdes, K., RodZ, a new player in bacterial cell morphogenesis. EMBO J, 2009. 28(3): p. 171-2. 271. O'Reilly, G. and F. Clarke, Identification of an actin binding region in aldolase. FEBS Lett, 1993. 321(1): p. 69-72. 272. Walsh, T.P., F.M. Clarke, and C.J. Masters, Modification of the kinetic parameters of aldolase on binding to the actin-containing filaments of skeletal muscle. Biochem J, 1977. 165(1): p. 165-7. 273. Walsh, T.P., et al., Binding of aldolase to actin-containing filaments. Evidence of interaction with the regulatory proteins of skeletal muscle. Biochem J, 1980. 186(1): p. 89-98. 274. Wang, J., et al., The molecular nature of the F-actin binding activity of aldolase revealed with site-directed mutants. J Biol Chem, 1996. 271(12): p. 6861-5. 275. Wang, J., D.R. Tolan, and L. Pagliaro, Metabolic compartmentation in living cells: structural association of aldolase. Exp Cell Res, 1997. 237(2): p. 445-51. 276. Bosch, J., et al., Aldolase provides an unusual binding site for thrombospondin- related anonymous protein in the invasion machinery of the malaria parasite. Proc Natl Acad Sci U S A, 2007. 104(17): p. 7015-20. 277. Buscaglia, C.A., et al., Characterization of an aldolase-binding site in the Wiskott-Aldrich syndrome protein. J Biol Chem, 2006. 281(3): p. 1324-31. 278. Jewett, T.J. and L.D. Sibley, Aldolase forms a bridge between cell surface adhesins and the actin cytoskeleton in apicomplexan parasites. Mol Cell, 2003. 11(4): p. 885-94.

262 | P a g e

279. Thorell, S., et al., Crystal structure of decameric fructose-6-phosphate aldolase from Escherichia coli reveals inter-subunit helix swapping as a structural basis for assembly differences in the transaldolase family. J Mol Biol, 2002. 319(1): p. 161-71. 280. Kruse, T., et al., Actin homolog MreB and RNA polymerase interact and are both required for chromosome segregation in Escherichia coli. Genes Dev, 2006. 20(1): p. 113-24. 281. Brieher, W.M., et al., Rapid actin monomer-insensitive depolymerization of Listeria actin comet tails by cofilin, coronin, and Aip1. J Cell Biol, 2006. 175(2): p. 315-24. 282. Alberts, B.M., et al., Studies with DNA-cellulose chromatography. I. DNA- binding proteins from Escherichia coli. Cold Spring Harb Symp Quant Biol, 1968. 33: p. 289-305. 283. Palter, K.B. and B.M. Alberts, The use of DNA-cellulose for analyzing histone- DNA interactions. Discovery of nucleosome-like histone binding to single- stranded DNA. J Biol Chem, 1979. 254(21): p. 11160-9. 284. Tsolas, O. and L. Joris, Transaldolase. Methods Enzymol, 1975. 42: p. 290-7. 285. Wachi, M. and M. Matsuhashi, Negative control of cell division by mreB, a gene that functions in determining the rod shape of Escherichia coli cells. J Bacteriol, 1989. 171(6): p. 3123-7. 286. Garner, E.C., et al., Reconstitution of DNA segregation driven by assembly of a prokaryotic actin homolog. Science, 2007. 315(5816): p. 1270-4. 287. Salje, J., B. Zuber, and J. Lowe, Electron cryomicroscopy of E. coli reveals filament bundles involved in plasmid DNA segregation. Science, 2009. 323(5913): p. 509-12. 288. Egelman, E.H., Actin's prokaryotic homologs. Curr Opin Struct Biol, 2003. 13(2): p. 244-8. 289. Galkin, V.E., et al., A new internal mode in F-actin helps explain the remarkable evolutionary conservation of actin's sequence and structure. Curr Biol, 2002. 12(7): p. 570-5. 290. Chiu, S.W., S.Y. Chen, and H.C. Wong, Dynamic localization of MreB in Vibrio parahaemolyticus and in the ectopic host bacterium Escherichia coli. Appl Environ Microbiol, 2008. 74(21): p. 6739-45. 291. Slovak, P.M., G.H. Wadhams, and J.P. Armitage, Localization of MreB in Rhodobacter sphaeroides under conditions causing changes in cell shape and membrane structure. J Bacteriol, 2005. 187(1): p. 54-64. 292. Andrews, S.S. and A.P. Arkin, A mechanical explanation for cytoskeletal rings and helices in bacteria. Biophys J, 2007. 93(6): p. 1872-84. 293. Allard, J.F. and A.D. Rutenberg, Steady-state helices of the actin homolog MreB inside bacteria: dynamics without motors. Phys Rev E Stat Nonlin Soft Matter Phys, 2007. 76(3 Pt 1): p. 031916. 294. Shiomi, D., et al., Helical distribution of the bacterial chemoreceptor via colocalization with the Sec protein translocation machinery. Mol Microbiol, 2006. 60(4): p. 894-906. 295. Campo, N., et al., Subcellular sites for bacterial protein export. Mol Microbiol, 2004. 53(6): p. 1583-99.

P a g e | 263

296. Den Blaauwen, T., et al., Penicillin-binding protein PBP2 of Escherichia coli localizes preferentially in the lateral wall and at mid-cell in comparison with the old cell pole. Mol Microbiol, 2003. 47(2): p. 539-47. 297. Gibbs, K.A., et al., Complex spatial distribution and dynamics of an abundant Escherichia coli outer membrane protein, LamB. Mol Microbiol, 2004. 53(6): p. 1771-83. 298. Taghbalout, A. and L. Rothfield, RNaseE and the other constituents of the RNA degradosome are components of the bacterial cytoskeleton. Proc Natl Acad Sci U S A, 2007. 104(5): p. 1667-72. 299. Munoz-, D., et al., The actin-like MreB cytoskeleton organizes viral DNA replication in bacteria. Proc Natl Acad Sci U S A, 2009. 300. Boeneman, K., et al., Escherichia coli DnaA forms helical structures along the longitudinal cell axis distinct from MreB filaments. Mol Microbiol, 2009. 301. Barak, I., et al., Lipid spirals in Bacillus subtilis and their role in cell division. Mol Microbiol, 2008. 68(5): p. 1315-27. 302. Lambrechts, A., et al., Listeria comet tails: the actin-based motility machinery at work. Trends Cell Biol, 2008. 18(5): p. 220-7. 303. Brown, P.J., et al., Complex regulatory pathways coordinate cell-cycle progression and development in Caulobacter crescentus. Adv Microb Physiol, 2009. 54: p. 1-101. 304. Wegner, A., Head to tail polymerization of actin. J Mol Biol, 1976. 108(1): p. 139-50. 305. Hill, T.L. and M.W. Kirschner, Subunit treadmilling of microtubules or actin in the presence of cellular barriers: possible conversion of chemical free energy into mechanical work. Proc Natl Acad Sci U S A, 1982. 79(2): p. 490-4. 306. Hill, T.L. and M.W. Kirschner, Bioenergetics and kinetics of microtubule and actin filament assembly-disassembly. Int Rev Cytol, 1982. 78: p. 1-125. 307. Mitchison, T.J. and E.D. Salmon, Poleward kinetochore fiber movement occurs during both metaphase and anaphase-A in newt lung cell mitosis. J Cell Biol, 1992. 119(3): p. 569-82. 308. Mitchison, T.J., Polewards microtubule flux in the mitotic spindle: evidence from photoactivation of fluorescence. J Cell Biol, 1989. 109(2): p. 637-52. 309. Sawin, K.E. and T.J. Mitchison, Poleward microtubule flux mitotic spindles assembled in vitro. J Cell Biol, 1991. 112(5): p. 941-54. 310. Waterman-Storer, C.M., et al., Fluorescent speckle microscopy, a method to visualize the dynamics of protein assemblies in living cells. Curr Biol, 1998. 8(22): p. 1227-30. 311. Desai, A., et al., Anaphase A chromosome movement and poleward spindle microtubule flux occur At similar rates in Xenopus extract spindles. J Cell Biol, 1998. 141(3): p. 703-13. 312. Maddox, P., et al., Poleward microtubule flux is a major component of spindle dynamics and anaphase a in mitotic Drosophila embryos. Curr Biol, 2002. 12(19): p. 1670-4. 313. Yang, G., et al., Architectural dynamics of the meiotic spindle revealed by single- fluorophore imaging. Nat Cell Biol, 2007. 9(11): p. 1233-42. 314. Rodionov, V.I. and G.G. Borisy, Microtubule treadmilling in vivo. Science, 1997.

264 | P a g e

275(5297): p. 215-8. 315. Selve, N. and A. Wegner, Rate of treadmilling of actin filaments in vitro. J Mol Biol, 1986. 187(4): p. 627-31. 316. Pollard, T.D. and M.S. Mooseker, Direct measurement of actin polymerization rate constants by electron microscopy of actin filaments nucleated by isolated microvillus cores. J Cell Biol, 1981. 88(3): p. 654-9. 317. Fujiwara, I., et al., Microscopic analysis of polymerization dynamics with individual actin filaments. Nat Cell Biol, 2002. 4(9): p. 666-73. 318. Carlier, M.F., et al., Actin depolymerizing factor (ADF/cofilin) enhances the rate of filament turnover: implication in actin-based motility. J Cell Biol, 1997. 136(6): p. 1307-22. 319. Waterman-Storer, C.M., W.C. Salmon, and E.D. Salmon, Feedback interactions between cell-cell adherens junctions and cytoskeletal dynamics in newt lung epithelial cells. Mol Biol Cell, 2000. 11(7): p. 2471-83. 320. Watanabe, N. and T.J. Mitchison, Single-molecule speckle analysis of actin filament turnover in lamellipodia. Science, 2002. 295(5557): p. 1083-6. 321. Rzadzinska, A.K., et al., An actin molecular treadmill and maintain stereocilia functional architecture and self-renewal. J Cell Biol, 2004. 164(6): p. 887-97. 322. Niu, L. and J. Yu, Investigating intracellular dynamics of FtsZ cytoskeleton with photoactivation single-molecule tracking. Biophys J, 2008. 95(4): p. 2009-16. 323. Jensen, R.B., S.C. Wang, and L. Shapiro, A moving DNA replication factory in Caulobacter crescentus. Embo J, 2001. 20(17): p. 4952-63. 324. Vollmer, W., M. von Rechenberg, and J.V. Holtje, Demonstration of molecular interactions between the murein polymerase PBP1B, the lytic transglycosylase MltA, and the scaffolding protein MipA of Escherichia coli. J Biol Chem, 1999. 274(10): p. 6726-34. 325. Yao, X., et al., Thickness and elasticity of gram-negative murein sacculi measured by atomic force microscopy. J Bacteriol, 1999. 181(22): p. 6865-75. 326. Glauner, B. and J.V. Holtje, Growth pattern of the murein sacculus of Escherichia coli. J Biol Chem, 1990. 265(31): p. 18988-96. 327. Glauner, B., J.V. Holtje, and U. Schwarz, The composition of the murein of Escherichia coli. J Biol Chem, 1988. 263(21): p. 10088-95. 328. Harz, H., K. Burgdorf, and J.V. Holtje, Isolation and separation of the glycan strands from murein of Escherichia coli by reversed-phase high-performance liquid chromatography. Anal Biochem, 1990. 190(1): p. 120-8. 329. Barrett, D., et al., Analysis of glycan polymers produced by peptidoglycan glycosyltransferases. J Biol Chem, 2007. 282(44): p. 31964-71. 330. Wang, T.S., et al., Isolated peptidoglycan glycosyltransferases from different organisms produce different glycan chain lengths. J Am Chem Soc, 2008. 130(43): p. 14068-9. 331. Luijsterburg, M.S., et al., The architectural role of nucleoid-associated proteins in the organization of bacterial chromatin: a molecular perspective. J Struct Biol, 2006. 156(2): p. 262-72. 332. Bouffartigues, E., et al., H-NS cooperative binding to high-affinity sites in a regulatory element results in transcriptional silencing. Nat Struct Mol Biol, 2007.

P a g e | 265

14(5): p. 441-8. 333. Lang, B., et al., High-affinity DNA binding sites for H-NS provide a molecular basis for selective silencing within proteobacterial genomes. Nucleic Acids Res, 2007. 35(18): p. 6330-7. 334. Chen, C.C., et al., A cis-spreading nucleoprotein filament is responsible for the gene silencing activity found in the promoter relay mechanism. J Biol Chem, 2005. 280(6): p. 5101-12. 335. Toro, E., et al., Caulobacter requires a dedicated mechanism to initiate chromosome segregation. Proc Natl Acad Sci U S A, 2008. 105(40): p. 15435-40.

266 | P a g e