Electrostatics and Binding Properties of Phosphatidylinositol- 4,5-Bisphosphate in Model Membranes

ELECTROSTATICS AND BINDING PROPERTIES OF PHOSPHATIDYLINOSITOL- 4,5-BISPHOSPHATE IN MODEL MEMBRANES

A dissertation submitted to

Kent State University in partial

fulfillment of the requirements for the

degree of Doctor of Philosophy

Zachary T. Graber

December, 2014

Except for previously published materials Dissertation written by

Zachary T. Graber

B.S., Grantham University, 2008

Ph.D., Kent State University, 2014

Approved by

Arne Gericke, Department Head, Dr. rer. nat., WPI Department of Chemistry and Biochemistry, Doctoral Advisor

Edgar Kooijman, Associate Professor, Ph.D., Department of Biological Sciences, Doctoral Co-advisor

Roger Gregory, Professor, Ph.D., Department of Chemistry and Biochemistry

Anatoly Khitrin, Professor, Ph.D., Department of Chemistry and Biochemistry

Derek Damron, Professor, Ph.D., Department of Biological Sciences

Elizabeth Mann, Professor, Ph.D., Department of Physics

Accepted by

Michael Tubergen, Professor, Ph.D., Chair, Department of Chemistry and Biochemistry

James Blank, Interim Dean, Ph.D., College of Arts and Sciences TABLE OF CONTENTS

LIST OF FIGURES ...... VIII

LIST OF TABLES ...... XII

DEDICATION...... XIII

ACKNOWLEDGEMENTS ...... XIV

1. INTRODUCTION

1.1 Phosphatidylinositol 4,5-bisphosphate location and importance ...... 17

1.2 The Structure of PI(4,5)P2...... 20

1.3 PI(4,5)P2 Charge ...... 21

1.4 Phosphatidic acid electrostatic behavior ...... 22

1.5 PI(4,5)P2 electrostatic behavior ...... 23

1.6 PI(4,5)P2 domain formation and rafts ...... 24

1.7 PI(4,5)P2 interactions with cations ...... 26

1.8 PI(4,5)P2 interactions with PTEN ...... 28

1.9 Objectives and Methodology ...... 29

2. METHODS

2.1 31P NMR Titration...... 32

2.2 Preparation of Lipid Films ...... 35

2.3 Preparation of MLV dispersions ...... 36

2.4 NMR Experimental Procedure ...... 38

2.5 31P NMR Data Analysis ...... 38

2.6 Error Determination and ANOVA Analysis ...... 40

2.7 31P Spin-lattice Relaxation Experiments ...... 41

III

2.8 T1 NMR Experimental Procedure ...... 42

2.9 Solution NMR Experiments ...... 43

2.10 Solution NMR Procedure ...... 44

2.11 Surface sensitive X-ray experiments ...... 44

2.12 X-ray Data Analysis ...... 48

2.13 Giant Unilamellar Vesicle Preparation ...... 52

3. MODELING COMPLEX IONIZATION OF POLYPHOSPHOINOSITIDES

3.1 Introduction ...... 54

3.2 The Fitting Model ...... 55

3.3 Fitting of phosphatidylinositol bisphosphate ionization behavior: Theory ...... 56

3.4 Fitting of phosphatidylinositol trisphosphate ionization behavior: Theory ...... 61

3.5 Fitting of PI(3,4)P2 ionization behavior: Results ...... 65

3.6 Fitting of PI(4,5)P2 ionization behavior: Results ...... 68

3.7 Fitting of PI(3,5)P2 ionization behavior: Results ...... 70

3.8 Fitting of PI(3,4,5)P3 ionization behavior: Results ...... 73

3.9 Discussion ...... 76

4. IONIZATION OF PI(4,5)P2 IN MODEL MEMBRANES

4.1 Introduction ...... 78

4.2 Results ...... 80

4.2.1 PE promotes increased deprotonation of PI(4,5)P2 ...... 81

4.2.2 Phosphatidylserine has little effect on the ionization state of PI(4,5)P2 in a PC

matrix ...... 84

4.2.3 Phosphatidylinositol induces PI(4,5)P2 domain formation and differentially

effects the ionization state of PI(4,5)P2 ...... 90

4.3 Discussion ...... 94

4.4 Summary ...... 98

5. NMR STUDY OF PI(4,5)P2 INTERACTIONS WITH POTENTIAL CLUSTERING

AGENTS

5.1 Introduction ...... 101

5.2 Results

2+ 2+ 5.2.1 Interaction of PI(4,5)P2 with the Divalent Cations Mg and Ca ...... 104

5.2.2 The varying effect of divalent cations on PI(4,5)P2 ...... 109

5.2.3 Interaction of cholesterol with PI(4,5)P2 ...... 110

5.2.4 Comparison of the interaction of cholesterol and Phosphatidylinositol with

PI(4,5)P2 ...... 114

5.2.5 Cholesterol and the divalent cations Ca2+ and Mg2+ have a cumulative effect on

PI(4,5)P2 ...... 117

2+ 5.2.6 PE and PI influence Ca binding to PI(4,5)P2 ...... 120

5.3 Discussion

2+ 2+ 5.3.1 Interaction of PI(4,5)P2 with Ca and Mg ...... 122

5.3.2 Interactions at the 5-phosphate of PI(4,5)P2 ...... 125

5.3.3 Effect of cholesterol and interaction with PE and PI ...... 126

2+ 5.3.4 Effect of PE and PI on the PI(4,5)P2 interaction with Ca ...... 129

5.4 Conclusions ...... 131

6. CATION BINDING TO PI(4,5)P2 X-RAY STUDY

6.1 Introduction ...... 135

6.2 Results

6.2.1 Calcium reorients PI(4,5)P2 in model lipid membranes ...... 137

6.3 X-Ray Fluorescence ...... 143

6.4 Discussion ...... 147

6.5 Conclusion ...... 151

7. SPECIFIC INTERACTION BETWEEN PI(4,5)P2 AND PTEN

7.1 Introduction ...... 153

31 7.2 P NMR reveals an interaction between PTEN peptides and PI(4,5)P2 ...... 154

7.3 Dynamics of the PI(4,5)P2 headgroup ...... 159

7.4 PTEN peptide NMR assignment ...... 162

7.5 PI(4,5)P2-amine NMR assignment...... 167

7.6 NMR investigation of PTEN/PI(4,5)P2 interaction ...... 168

7.7 Discussion

7.7.1 The PI(4,5)P2 5-phosphate is important for the PTEN/PI(4,5)P2 interaction .. 171

2+ 7.7.2 Ca and PTEN compete for PI(4,5)P2 ...... 171

7.7.3 PTEN binds tightly to PI(4,5)P2 and alters its dynamics ...... 172

7.7.4 The tyrosine, lysine and arginine residues of PTEN’s n-terminal end interact

with PI(4,5)P2 ...... 172

7.8 Conclusion ...... 173

8. DISCUSSION AND CONCLUSIONS

8.1 Project summation ...... 174

8.2 Future work...... 177

REFERENCES ...... 180

APPENDIX + LIST OF MATERIALS ...... 194

VII

LIST OF FIGURES

Figure 1-1. Chemical structures of the seven naturally occurring phosphoinositides ...... 18

Figure 1-2. Regulation of the AKT signaling pathway by PTEN ...... 19

Figure 1-3. Chemical structure of Phosphatidylinositol-4,5-bisphosphate ...... 20

Figure 2-1. 31P NMR of a MLV dispersion ...... 33

31 Figure 2-2. P NMR titration curve for bPI(4,5)P2 in a DOPC matrix ...... 34

Figure 2-3. Degree of protonation for bPI(4,5)P2...... 40

Figure 2-4. Relaxation of the nucleus over time...... 41

Figure 2-5. X-ray experimental setup...... 46

Figure 2-6. Bulk Fluorescence Spectra...... 50

Figure 2-7: Fluorescence Qz Dependence...... 51

Figure 3-1. The ionization model for PI(4,5)P2...... 57

Figure 3-2. The ionization model for PI(3,4,5)P3...... 62

Figure 3-3. Fitting results for PI(3,4)P2...... 66

Figure 3-4. Fitting results for PI(4,5)P2...... 69

Figure 3-5. Fitting results for PI(3,5)P2...... 71

Figure 3-6. Fitting results for PI(3,4,5)P3...... 74

Figure 4-1. Chemical structures of the lipids ...... 80

31 Figure 4-2. P MAS NMR spectra and pH titration curves for PC / PE / PI(4,5)P2 (47.5% /

47.5% / 5%)...... 82

Figure 4-3. Ionization behavior of PA in PC/PS/PA (75% / 20% / 5%) vesicles as determined by

solid state 31P NMR ...... 86

Figure 4-4. GUVs composed of PC/PS/PI(4,5)P2 ...... 88

VIII

31 Figure 4-5: P MAS NMR spectra and pH titration curves for PC / PS / PI(4,5)P2 (78% / 20% /

2%) ...... 89

Figure 4-6. 31P MAS NMR spectra and pH titration curves for PC / PI / PI(4,5)P2 (88% / 10% /

2%) ...... 91

Figure 4-7. Cooperative domain formation of PI and PI(4,5)P2 in POPC GUVs ...... 93

Figure 5-1. Chemical structures of the lipids ...... 103

2+ Figure 5-2. Effect of Ca on the ionization properties of PI(4,5)P2 ...... 106

2+ Figure 5-3. Effect of Mg on the ionization properties of PI(4,5)P2 ...... 107

Figure 5-4. Effect of cations on the ionization properties of PI(4,5)P2 ...... 110

Figure 5-5. Effect of cholesterol on the ionization properties of PI(4,5)P2 ...... 112

31 Figure 5-6. P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with varying

amounts of PE and cholesterol ...... 114

31 Figure 5-7. P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with varying

amounts of PI and cholesterol ...... 116

31 Figure 5-8. P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with divalent cations

and cholesterol ...... 118

31 Figure 5-9. P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with divalent cations

and the phospholipids PI and PE ...... 121

2+ Figure 6-1. X-ray reflectivity of the PI(4,5)P2 monolayer in the presence of varying [Ca ] .... 137

2+ Figure 6-2. X-ray reflectivity of the PI(4,5)P2 monolayer in the presence of 1 mM Mg and

varying [Ca2+] ...... 140

Figure 6-3. X-ray fluorescence reveals cation binding to the PI(4,5)P2 monolayer ...... 144

+ 2+ Figure 6-4. Calculated ion density for K and Ca at the PI(4,5)P2 monolayer ...... 146

31 Figure 7-1. MAS P NMR chemical shift of PI(4,5)P2 with PTEN10-16 ...... 155

31 Figure 7-2. P NMR spectra of PI(4,5)P2-amine in the presence of varying PTEN10-16 ...... 157

31 2+ Figure 7-3. MAS P NMR chemical shift of PI(4,5)P2 with PTEN10-16 and Ca ...... 158

31 Figure 7-4. MAS P NMR inversion recovery experiment with PC:PI(4,5)P2 MLVs ...... 160

Figure 7-5. T1 spin-lattice relaxation vales for PI(4,5)P2 and PC in PC / PI(4,5)P2 MLVs with

PTEN10-16 ...... 161

Figure 7-6. Chemical structures of the tested peptides and soluble PI(4,5)P2-amine ...... 163

Figure 7-7. COSY assignment of PTEN10-16...... 164

Figure 7-8. NOESY assignment of PTEN10-16 ...... 165

Figure 7-9. COSY assignment of PI(4,5)P2-amine ...... 167

Figure 7-10. Comparison of COSY spectra of PTEN10-16 in the presence and absence of

PI(4,5)P2-amine ...... 168

Figure 7-11. NOESY spectra of a PTEN10-16:PI(4,5)P2-amine mixture ...... 170

Figure A1. MAS 31P NMR spectra for a 1:1 mixture of PC and PE, at pH ~7 ...... 211

31 Figure A2. Solid state static P NMR spectra for ternary mixtures of PC, PI(4,5)P2, and PE, PS,

or PI at three pH covering the pH range investigated in the pH titration curves ...... 212

Figure A3. Ionization behavior of PS as determined by solid state 31P NMR ...... 213

Figure A4. MAS 31P NMR spectra for MLVs containing varying ratios of PC and PI,

at pH ~7 ...... 214

31 Figure A5. P MAS NMR spectra and pH titration curves for PC / PI / PI(4,5)P2 (78% / 20% /

2%) ...... 215

Figure A6. Comparison of 31P MAS NMR titration curves for two concentrations of PI ...... 216

2+ Figure A7. pH dependent effect of Ca on ionization of PI(4,5)P2 ...... 218

31 Figure A8. Solid-state Static P NMR spectra for PC / PI(4,5)P2 vesicles with varying

composition ...... 219

31 Figure A9. P MAS NMR chemical shift bar graph for PC / PI(4,5)P2 vesicles with varying

amounts of PE and cholesterol at pH 5 ...... 226

31 Figure A10. P NMR spectra of PI(4,5)P2-amine in the presence of varying PTEN10-16 ...... 227

Figure A11. PTEN10-16 TOCSY in the presence of PI(4,5)P2-amine ...... 228

Figure A12. PTEN12-18 TOCSY in the presence of PI(4,5)P2-amine ...... 229

Figure A13. DSC of PC:Cholesterol:PI:PI(4,5)P2 vesicles ...... 230

LIST OF TABLES

Table 3-1. Comparison of pKa values for PI(4,5)P2 and Ins(4,5)P2 ...... 77

2+ 2+ Table 5-1. PI(4,5)P2 phosphomonoester chemical shifts in the presence of Ca and Mg ...... 109

Table 5-2. Comparison of PI(4,5)P2 phosphomonoester chemical shifts in the presence of cholesterol or cholesterol and divalent cations (PC/PI(4,5)P2/Cholesterol (98%-x/2%/x%) 119

Table 6-1. PI(4,5)P2 monolayer structural parameters derived from a 2 box fit to the reflectivity curves from Figure 6-1A ...... 138

Table 6-2. PI(4,5)P2 monolayer structural parameters derived from a 2 box fit to the reflectivity curves from Figure 6-2A ...... 141

Table 6-3. PI(4,5)P2 monolayer structural parameters derived from a 2 box fit to the reflectivity curves from Figure 6-2C ...... 142

Table 7-1. Relaxation times for PI(4,5)P2 in the presence and absence of PTEN10-16 ...... 161

Table 7-2. Standard amino acid chemical shift values ...... 166

Table 7-3. Assigned PTEN10-16 amino acid chemical shift values ...... 166

XII

Dedication

This work is dedicated to my wonderful wife Melody and our beautiful little baby. Thank you for loving me and believing in me.

XIII

Acknowledgements

I am extremely grateful to the many people who have helped to make this dissertation possible. Without the support of so many wonderful people I could never have completed this work.

First, I would like to thank my outstanding advisors, Dr. Arne Gericke and Dr. Edgar

Kooijman. Thank you for allowing me to work in your labs, to make mistakes and to learn through experience. You have encouraged me to think for myself and have shown me how to conduct good science. Your advice, criticism, and encouragement have helped me develop from a student into a scientist. I hope that one day I might be able to be as good a mentor to my own students as you both have been to me.

I have had many wonderful lab mates over the years. Katie King, Dr. Zhiping Jiang, Dr.

Yasmin Blaih, and Dr. Stephen Woods, thank you for welcoming me into the Gericke lab and helping me to get settled as a new graduate student. Zhiping, your patient training and kindness helped me greatly as I started my first experiments. Katie, I appreciated your friendliness and your generosity. I have also really appreciated my Kooijman lab coworkers, Sewwandi

Rathnayake, Dr. Anne Froyd-Rankenberg, Priya Putta, and Mona Mirheydari. Sewwandi has been my coworker in the Kooijman lab for four wonderful years now. I am grateful for the many ways she has helped me and made work in the lab more enjoyable. Anne has offered much- appreciated scientific insight, and has pushed me to learn more cooking. Priya has been a great friend over the past couple of years in the lab. Her cheerful spirit and encouragement have helped to make the lab a great place to work. I am also grateful to Mona for her friendship.

This work is not mine alone but also that of the many wonderful colleagues and teachers who have assisted or taught me along the way. First of all, without Dr. Mahinda Gangoda this

XIV work would not be possible. His training enabled me to perform the NMR experiments that are recorded in this work. He has kept the NMR machines running over the years and has helped me out with many problems. I have greatly appreciated his hard work and cheerful personality. Dr.

Mike Model trained me in the use of the fluorescent microscope for the GUV experiments.

Zhiping Jiang’s skill with GUVs is amazing, and it is thanks to her training that I was able to complete the GUV experiments that I performed. Some of her own incredible GUV work is included in this thesis. Dr. Elizabeth Mann has been a great mentor. Her excellent advice and criticism have helped to improve my writing, speaking, and scientific reasoning. I would also like to thank Dr. Mann and her student, Piotr Popov, for allowing me to come into their lab and for training me to run Langmuir isotherms and use the Brewster angle microscope. My x-ray experiments were performed in collaboration with Dr. Wenjie Wang, Dr. Gautam Singh, Dr.

Ivan Kuzmenko, and Dr. David Vaknin at the advanced photon source (APS) of the Argonne

National Lab. Ivan Kuzmenko, the beamlime scientist at the 9-IDC beamline where we performed our experiments, helped to get us started and stepped in to save us on multiple occasions. Without his help we would not have been able to get any usable data from our experiments at the APS. Wenjie Wang was outstanding in his efforts at the APS and later at

Ames lab. I am very grateful for his patient training and his advice. David Vaknin acted as my advisor for the data analysis that I performed at Ames lab. Wenjie Wang and David Vaknin assisted in writing up the x-ray work in chapter 6, as well as the x-ray methodology discussed in chapter 2. Gautam Singh willingly came with us to the APS and worked very long hours with us to help us get our data. I am grateful for his assistance.

I would also like to thank my committee members, Dr. Roger Gregory, Dr. Anatoly

Khitrin, and Dr. Derek Damron. Your advice and helpful comments have greatly improved this work. Thank you for taking the time to help make this possible.

I am grateful to the agencies that have funded this work, the majority of which was supported by a Farris Family Fellowship to Dr. Edgar E. Kooijman, as well as an NSF chemistry grant. The work at the Ames Laboratory was supported by the Office of Basic Energy Sciences,

U.S. Department of Energy under Contract No. DE-AC02-07CH11358. Use of the Advanced

Photon Source was supported by the U. S. Department of Energy, Office of Science, Office of

Basic Energy Sciences, under Contract No. DE-AC02-06CH11357.

I would like to thank my friends and families who have encouraged me, inspired me, and put up with me over the years as I have worked on my dissertation. My parents’ incredible love and support has made this entire journey possible, I can never thank them enough. My siblings are amazing, I am so grateful that I have such wonderful family members. I really appreciate their support. My wife, Melody, has been an incredible blessing to me. She is my cheerleader and companion, as well as my editor at times. My in-laws, the Lyons family, have been incredibly helpful and supportive as well. I am grateful for their encouragement and support. I have so many friends that have helped me in many ways, with kind words, good advice, or simply a listening ear. I want to especially thank Dr. Chris Carmichael and Dr. Al Gotch, who both encouraged me and instructed me as I began my career in the sciences. I have been very grateful for their advice over the years. Finally, I would like to thank God for His incredible blessings. This work is due solely to Him.

Zachary Graber

December, 2014, Kent, Ohio

XVI

Chapter 1

Introduction

Phosphatidylinositol 4,5-bisphosphate location and importance

The cell plasma membrane, composed of a bilayer of phospholipids with embedded membrane proteins, is a critical component of the cell. It acts as a barrier for the cell, keeping cellular components and extracellular components apart. It also acts as a signaling platform, and many signaling events within the cell occur on the surface of the cell plasma membrane.

Signaling phospholipids are crucial to these signaling events. Many proteins bind to signaling lipids on the surface of the plasma membrane in order to associate with the membrane and organize into functional complexes (Cho 2006). Phosphoinositides are a relatively rare and yet extremely important group of signaling phospholipids that reside on the inner leaflet of the cell plasma membrane, as well as in many intracellular membranes. The phosphoinositide headgroup consists of an inositol ring with one to three phosphates on the 3, 4, and 5 positions of the inositol ring. The seven members of the phosphoinositide family are distinguished by the number and the varied position of the phosphate group on the inositol ring (see Figure 1-1).

Phosphatidylinositol-4,5-bisphosphate [PI(4,5)P2] has phosphate groups on the 4- and 5- positions of the inositol ring, and is the most common phosphoinositide, despite only comprising about 1 mol% of all lipid within the cell plasma membrane. Within the nuclear envelope, significantly higher levels of PI(4,5)P2 and other phosphoinositides have been found, reaching levels above 7 mol% (Zhendre, Grelard et al. 2011). PI(4,5)P2 is noted among the

Figure 1-1. Chemical structures of the seven naturally occuring phosphoinositides. Phosphatidylinositol-3-phosphate (PI3P), Phosphatidylinositol-4-phosphate (PI4P), phosphatidylinositol-

5-phosphate (PI5P), phosphatidylinositol-3,4-bisphosphate [PI(3,4)P2], phosphatidylinositol-3,5- bisphosphate [PI(3,5)P2], phosphatidylinositol-4,5-bisphosphate [PI(4,5)P2], phosphatidylinositol-3,4,5- trisphosphate [PI(3,4,5)P3].

phosphoinositides for its great importance in many different signaling roles (Catimel,

Schieber et al. 2008).

Among the most well-known signaling role is PI(4,5)P2’s involvement in calcium signaling.

Phospholipase C (PLC) catalyzes the cleavage of the PI(4,5)P2 headgroup to produce diacyl glycerol and inositol-1,4,5-trisphosphate (IP3). The produced IP3 activates calcium channels in the endoplasmic reticulum to release calcium into the cell cytosol. The calcium, along with

DAG, activates protein kinase C (PKC) at the plasma membrane to trigger downstream signaling events (Lodish 2013). PI(4,5)P2 is also known to affect the function of several other ion channels

(Suh and Hille 2008; Kooijman, Kuzenko et al. 2011). In some cases function of these channels also requires the presence of cholesterol. PI(4,5)P2 also plays a critical role in cell growth and

Figure 1-2. Regulation of the AKT signaling pathway by PTEN.

motility (see Figure 1-2). PI3K phosphorylates PI(4,5)P2 to form Phosphatidylinositol-3,4,5- trisphosphate [PI(3,4,5)P3]. PI(3,4,5)P3 is a target for AKT (also known as protein kinase B), which binds to PI(3,4,5)P3 at the membrane. This interaction allows for phosphorylation of AKT by phosphoinositide dependent kinase 1 (PDK1), which activates AKT (Stokoe, Stephens et al.

1997). Activated AKT then triggers cell growth as well as cell motility. In chemotaxis, a

PI(3,4,5)P3 gradient is formed at the leading edge of the cell and is critical for the cell movement

(Kolsch, Charest et al. 2008; Falke and Ziemba 2014). In order to regulate these AKT activated processes, the tumor suppressor PTEN binds to PI(3,4,5)P3 and dephosphorylates it to reform

PI(4,5)P2, which prevents AKT activation (Sun, Lesche et al. 1999). PTEN requires membrane association in order to activate its lipid phosphatase function (Das, Dixon et al. 2003). The N– terminus of PTEN binds to PI(4,5)P2 in the membrane through a specific interaction that is required for PTEN activation (Campbell, Liu et al. 2003; Iijima, Huang et al. 2004; Redfern,

Redfern et al. 2008; Singh, Odriozola et al. 2011).

The importance of PI(4,5)P2 is demonstrated by the fact that the concentration of PI(4,5)P2

19 binding proteins outnumber the actual PI(4,5)P2 concentration in the membrane (Catimel,

Schieber et al. 2008). Therefore, there is a high rate of turnover in PI(4,5)P2 binding.

The Structure of PI(4,5)P2

The diverse signaling functions of PI(4,5)P2 are made possible by the rich functionality of the

PI(4,5)P2 molecule (see Figure 1-3). The predominant acyl chains of PI(4,5)P2 are stearoyl and arachidonoyl. The saturated stearoyl chain is 18 carbons long while the highly unsaturated arachidonoyl chain has four double bonds in its 20 carbon chain. The highly unsaturated arachidonoyl chain ensures that PI(4,5)P2 prefers a disordered environment (Liu and Fletcher

2006; Shaw, Epand et al. 2006; Tong, Nguyen et al. 2008). The headgroup of PI(4,5)P2, like all phosphoinositides, consists of the six carbon inositol ring with hydroxyl groups at each carbon.

Three of these hydroxyl groups are bound to phosphates, with two phosphomonoesters at the 4- and 5- position and a phosphodiester in the first position securing the headgroup to the glycerol backbone. The hydroxyl groups and phosphomonoesters of PI(4,5)P2 give it a rich capacity for hydrogen bond formation with other molecules. The three phosphates of PI(4,5)P2 can each carry negative charge, allowing PI(4,5)P2 to interact electrostatically with cations within the cytosol.

Figure 1-3. Chemical structure of Phosphatidylinositol-4,5-bisphosphate. The chemical structure

of natural porcine brain phosphatidylinositol-4,5-bisphosphate [PI(4,5)P2]. The predominant acyl chain composition of stearoyl and arachidonoyl chains is depicted. The phosphomonoesters at the 4- and 5-position of the inositol ring are shown in blue and red respectively.

PI(4,5)P2 Charge

The ionization behavior of PI(4,5)P2 has been studied extensively. Van Paridon et al. used

31 P-NMR of PI(4,5)P2 in micelles and small unilamellar vesicles (SUVs) to determine the charge of PI(4,5)P2 (van Paridon, de Kruijff et al. 1986). The charge was determined based on the change in phosphorus chemical shift as the pH was varied. The pKa of the 4-phosphate was determined as 6.7, while the pKa of the 5-phosphate was 7.7, giving an overall (including the phosphodiester group) charge of roughly -4 at pH 7.2. However, it should be noted that the high curvature of the micelles and SUVs used in this study may have a significant effect on the charge of PI(4,5)P2 (Swairjo, Seaton et al. 1994; Kooijman, Carter et al. 2005). Other studies have used electrophoresis of PI(4,5)P2 in multilamellar vesicle suspensions containing phosphatidylcholine

(PC) (Toner, Vaio et al. 1988). The charge of PI(4,5)P2 was determined based on the mobility of the vesicles in the electrophoresis chamber. Based on these results and calculations using the

Gouy-Chapman theory, the charge of PI(4,5)P2 was determined to be -3 at pH 7, with one proton and one potassium ion bound to PI(4,5)P2 (Toner, Vaio et al. 1988). The charge of the D-myo- inositol 4,5-bisphosphate [Ins(4,5)P2] headgroup of PI(4,5)P2 has also been determined separately using potentiometric and 31P-NMR measurements (Schmitt, Bortmann et al. 1993). In contrast with the work of Paridon et al., a biphasic behavior was observed for the deprotonation of the 4- and 5-phosphates. The authors suggested that the biphasic behavior is likely due to intramolecular interactions between the two phosphates as they form a hydrogen-bond with each other which alters the ionization properties (Schmitt, Bortmann et al. 1993). Based on these results, the charge at pH 7.4 is calculated as -1.47 for the 4-phosphate and -1.34 for the 5- phosphate charge (total charge -2.81). Ohki et al. measured the zeta potential for PI(4,5)P2 vesicles at a variety of pH values, and estimated the charge at pH 7 as -4 (Ohki, Muller et al.

31 2010). In addition, the charge on PI(4,5)P2 has also been measured using P NMR of multilamellar vesicles, which will be discussed in further detail in the next section.

Phosphatidic acid electrostatic behavior

31P NMR is an incredibly useful technique for evaluating lipid charge. While many studies have measured 31P NMR of small vesicles or micelles, these model systems are not good models for the cell membrane due to their high degree of curvature (Swairjo, Seaton et al. 1994;

Kooijman, Carter et al. 2005). This problem can be resolved by using micrometer-sized multilamellar vesicles. However, when using such large vesicles, chemical shift anisotropy

(CSA) becomes a problem due to the slow molecular tumbling of the MLVs. Therefore, magic angle spinning (MAS) NMR must be used to eliminate the CSA to determine the chemical shift values for the individual phosphates as a function of pH. Kooijman et al., have used this MAS technique to measure the charge of phosphatidic acid (PA) (Kooijman, Carter et al. 2005;

Kooijman, Tieleman et al. 2007). In addition to more accurately modeling the cellular curvature,

MLVs may also be used with diverse lipid compositions in order to test the effect of other lipids on the ionization behavior of acidic phospholipids. For the signaling lipid PA, Kooijman et al. found that the presence of other lipids had a striking effect on PA charge (Kooijman, Carter et al.

2005). In the presence of the hydrogen-bond donor lipid phosphatidylethanolamine (PE), there is a distinct increase in deprotonation of PA. This increased charge was thought to be caused by the formation of a hydrogen-bond between the headgroups of PE and PA. This interaction would stabilize the deprotonated state of PA, thus promoting further deprotonation of PA. The ionization of PA was also investigated in the presence of the peptide KALP23, which acts as a simple model for a membrane protein. KALP23 is composed of a single transmembrane helix of

23 amino acids with two flanking lysine groups on either side of the helix extending into the

22 headgroup/aqueous interface region. In the presence of KALP23, PA was again deprotonated by the interaction with the peptide. These observations lead to the proposed ‘Electrostatic hydrogen- bond switch model’ for PA binding to peripheral proteins (Kooijman, Tieleman et al. 2007).

Initially, the PA binding protein binds loosely to the membrane via electrostatic interaction.

Once the protein binding domain finds PA, it is able to form hydrogen-bonds which promotes further deprotonation of PA and an increase of its charge. The combination of increased charge and hydrogen-bond formation leads to a much tighter interaction of the protein with the membrane. Mean field calculations show a dramatic increase in binding affinity due to the presence of the hydrogen bond interaction over electrostatic interaction alone (Loew, Kooijman et al. 2013).

PI(4,5)P2 electrostatic behavior

31 PI(4,5)P2 has been examined via this P NMR MAS technique as well (Kooijman, King et al.

2009). MLVs were made with 95 mol% Phosphatidylcholine and 5 mol% PI(4,5)P2. As with the

NMR measurements of inositol phosphates, the pH dependent chemical shift variation of the phosphates was found to follow a biphasic pattern. The biphasic ionization pattern indicates that in increasingly basic conditions the final proton of PI(4,5)P2 can be shared between the two adjacent phosphate groups. For the phosphoinositide PI(3,5)P2, which has no adjacent phosphate groups, this biphasic behavior is not observed as the final proton in a pH titration cannot be shared between the phosphomonoesters. The ionization of PI(4,5)P2 was determined over the pH range from 4 to 10, and the charge at pH 7 was determined to be -1.58 for the 4-phosphate and -

1.41 for the 5-phosphate, for a total charge of -4 (including the phosphodiester group).

Surprisingly, Kooijman et al. found a higher charge for PI(4,5)P2 within a natural membrane than for dissolved Ins(4,5)P2 (-2.99 for the phosphomonoesters of PI(4,5)P2 vs. -2.81 for Ins(4,5)P2

(Schlewer, Guedat et al. 1998)). For anionic lipids in a bilayer, the negative surface charge of the bilayer leads to a lower pH at the interface compared to the bulk pH. This would lead to a proportionately lower charge of the charged groups at the membrane, while the opposite was observed in this case. The increased charge for PI(4,5)P2 in the membrane could be due to hydrogen-bond formation. However, in this case the only other lipid present is PC, which has no potential for hydrogen-bond formation. Therefore, the PI(4,5)P2 molecules must form intermolecular hydrogen-bonds among themselves. This suggests that PI(4,5)P2 can overcome the mutual electrostatic repulsion and form intermolecular hydrogen-bonds. These interactions could potentially lead to PI(4,5)P2 clustering and the formation of heterogeneous domains within the membrane. The concept of PI(4,5)P2 domains and the evidence for their existence will be discussed below.

PI(4,5)P2 domain formation and rafts

The Singer and Nicholson “Fluid mosaic model” describes the cell plasma membrane as a fluid bilayer of phospholipids with a mosaic of integral membrane proteins floating in the lipid bilayer (Singer and Nicolson 1972). However, since Singer and Nicholson developed their model, there have been many suggestions that the lipid bilayer may have some level of organization, with distinct lipid domains within the membrane (Simons and van Meer 1988;

Simons and Ikonen 1997). Domains were first proposed based on experiments that showed a certain fraction of the membrane was resistant to the detergent Triton X-100 (London and Brown

2000). This detergent-resistant fraction of the membrane has come to be known as “lipid rafts”, and are now commonly thought to occur in the cell in some form or another. Raft domains are rich in cholesterol and saturated sphingolipids, making them more ordered than the bulk lipids around them. Raft domains are proposed to act as signaling platforms, with GPI-anchored

24 proteins and some integral membrane proteins suggested as being enriched within these liquid- ordered domains (Levental, Lingwood et al. 2010). Macroscopic ordered domains, or rafts, have been observed in many model membranes, including giant plasma membrane vesicles (GPMVs, also known as blebs) which are formed from the natural cell plasma membrane during apotosis and share many of the characteristics of the natural membrane (Baumgart, Hammond et al.

2007). However, macroscopic lipid rafts have not been observed in vivo, although some studies have found evidence of nanometer-sized raft domains (Owen, Williamson et al. 2012; Truong-

Quang and Lenne 2014).

PI(4,5)P2 has been proposed to also enrich in domains, either in raft domains or in separate

PI(4,5)P2 based domains (Koreh and Monaco 1986; Hope and Pike 1996; Pike and Casey 1996;

Pike and Miller 1998; Redfern and Gericke 2004; Redfern and Gericke 2005; Levental, Christian et al. 2009; Kwiatkowska 2010; Gao, Lowry et al. 2011; Wang, Collins et al. 2012; Jiang,

Redfern et al. 2014; Salvemini, Gau et al. 2014). In many cases, PI(4,5)P2 signaling has been found to be linked to the presence of cholesterol. Cholesterol depletion studies show impaired hormone-stimulated phosphatidylinositol turnover due to PI(4,5)P2 delocalization (Pike and

Miller 1998). These studies led to the suggestion that PI(4,5)P2 is enriched within raft domains.

However, raft domains are generally considered to be ordered, while the predominant PI(4,5)P2 acyl chains are highly disordered. Thus, PI(4,5)P2 is unlikely to partition into raft domains, unless there is some other strong driving factor. Other studies have also questioned the presence of PI(4,5)P2 in rafts, pointing out that the majority of studies suggesting otherwise use cholesterol depletion, which may have other effects that disrupt PI(4,5)P2 signaling (van

Rheenen, Achame et al. 2005).

An alternative hypothesis is that PI(4,5)P2 forms clusters independently, based on

25 intermolecular hydrogen bond formation between PI(4,5)P2 molecules. In 2004, FRET studies suggested that PI(4,5)P2 formed clusters in model membranes (Redfern and Gericke 2004;

Redfern and Gericke 2005). These results were strongly questioned by other researchers, who insisted that electrostatic repulsion between the highly charged PI(4,5)P2 molecules would be too strong for clusters to form (Fernandes, Loura et al. 2006; Gamper and Shapiro 2007; Blin,

31 Margeat et al. 2008). P NMR studies also suggested formation of PI(4,5)P2 clusters (Kooijman,

King et al. 2009). More recently, macroscopic PI(4,5)P2 clusters were observed at low temperatures by fluorescence in micrometer-sized giant unilamellar vesicles (GUVs) composed of PC and PI(4,5)P2 (Jiang, Redfern et al. 2014). These macroscopic clusters may be formed as the result of nanometer PI(4,5)P2 clusters coalescing at low temperatures. In addition to these studies suggesting PI(4,5)P2 clustering by itself, there have been many studies that have found

PI(4,5)P2 cluster formation to be induced by cellular components interacting with PI(4,5)P2.

Many of these induced clusters involve cellular cations interacting with PI(4,5)P2 via electrostatic interactions, which will be discussed below (Wang, McLaughlin et al. 2003;

Gambhir, Hangyas-Mihalyne et al. 2004; Wang, Collins et al. 2012; Sarmento, Coutinho et al.

2013; Wang, Slochower et al. 2014). In addition to cations, cholesterol has been suggested to promote PI(4,5)P2 cluster formation (Dasgupta, Bamba et al. 2009; Jiang, Redfern et al. 2014).

PI(4,5)P2 interactions with cations

The negative charge of the inositol headgroup gives PI(4,5)P2 a strong capability for electrostatic interactions with positively charged protein segments or cations. The interaction

2+ 2+ between PI(4,5)P2 and the cytosolic divalent cations Ca and Mg may play a critical role in

PI(4,5)P2 mediated signaling events. In early studies, PI(4,5)P2-divalent cation interaction was

2+ 2+ noted when it was observed that the addition of Ca or Mg led to PI(4,5)P2 precipitation

(Folch, Lees et al. 1957). In partition studies, Ca2+ was found to be more effective than Mg2+, suggesting a higher binding affinity for Ca2+ (Dawson 1965; Hauser and Dawson 1967). Toner et al. further examined this interaction using PI(4,5)P2 vesicles and determined the change in electrophoresis mobility with varying cation concentration (Toner, Vaio et al. 1988). Ca2+ can directly affect PI(4,5)P2 signaling, as it has been found to mediate interactions between protein

2+ 2+ C2 domains and PI(4,5)P2 (Evans, Gerber et al. 2004). In other cases, Ca and Mg simply

2+ 2+ affect the conformation of PI(4,5)P2. Levental at al., found that Ca and Mg bind to, and cause condensation of, PI(4,5)P2 monolayers (Levental, Cebers et al. 2008; Levental, Christian et al.

2+ 2009). More recently, Ca has been suggested to cause clustering of PI(4,5)P2 in the cell plasma membrane (Wang, Collins et al. 2012; Sarmento, Coutinho et al. 2013). This was observed by

AFM and FCS measurements of membrane models containing PI(4,5)P2. The positive charge of

2+ Ca is proposed to shield the negative charge of PI(4,5)P2, which then promotes formation of transient PI(4,5)P2 clusters through intermolecular hydrogen bond formation between the

PI(4,5)P2 headgroups. These clusters could have important signaling implications.

In addition to divalent cations, cellular polyamines also interact with PI(4,5)P2 (Wang,

Slochower et al. 2014). Spermine, a small molecule with several charged amine groups, can be found at roughly millimolar levels in the cytosol and has been shown to interact strongly with

PI(4,5)P2 (Toner, Vaio et al. 1988). However, McLaughlin et al. suggest that this smaller

2+ polyamine cannot promote PI(4,5)P2 clusters, as can Ca (McLaughlin, Wang et al. 2002). The large and highly charged MARCKS protein does seem to bind multiple PI(4,5)P2 molecules and sequester them (Wang, Gambhir et al. 2002; Wang, McLaughlin et al. 2003). The sequestering of

PI(4,5)P2 is supported by EPR measurements of spin-labeled PI(4,5)P2, which show multiple

PI(4,5)P2 molecules in close proximity in the presence of MARCKS (Rauch, Ferguson et al.

2002). Release of MARCKS sequestered PI(4,5)P2 molecules is proposed to be mediated by

Calmodulin and Ca2+ (McLaughlin and Murray 2005).

PI(4,5)P2 interactions with PTEN

In addition to general electrostatic interactions, PI(4,5)P2 also interacts with proteins through highly specific protein binding domains. Many of these domains are able to differentiate between the various phosphoinositides that differ only in the numbers and positioning of the phosphate groups (Stahelin, Scott et al. 2014). The pleckstrin homology (PH) lipid binding domain is a well-known protein domain that binds to PI(4,5)P2 (Lemmon 2007). The interaction between

PTEN and PI(4,5)P2 discussed earlier does not rely on a common protein domain, which is remarkable considering its importance in controlling cell growth and its high specificity for

PI(4,5)P2 (Campbell, Liu et al. 2003). The PTEN-PI(4,5)P2 association has been studied in model membranes using fluorescence quenching techniques, and it was found that PTEN preferentially binds PI(4,5)P2 over other phosphoinositides (Redfern, Redfern et al. 2008). Besides PI(4,5)P2, only phosphatidylinositol-5-phosphate [PI(5)P] showed moderate PTEN binding. PI(4,5)P2, and to a lesser extent PI(5)P, have been shown to activate PTEN (Campbell, Liu et al. 2003). This same binding preference was also observed for PTEN1-21, a peptide that was derived from the N- terminal end of PTEN, indicating that the N-terminus is primarily responsible for this specific interaction. Since phosphatidylinositol-3,5-bisphosphate [PI(3,5)P2], PI(3,4)P2 (which has the same charge as PI(4,5)P2), and PI(3,4,5)P3 (which has a higher charge than PI(4,5)P2) do not bind to PTEN’s N-terminus, this suggests that PTEN binding to PI(4,5)P2 is driven by hydrogen- bond formation rather than non-specific electrostatic interactions. FTIR measurements show that the N-terminal remains in an extended structure rather than forming a distinct secondary structure like an α-helix or β-sheet (Redfern, Redfern et al. 2008). Within the PTEN N-terminus,

28 the K13E amino acid substitution is a cancer relevant mutation, which suggests that it may be important for the PI(4,5)P2 interaction (Duerr, Rollbrocker et al. 1998). Indeed, N12K/K13N and

K13R/R14K amino acid switch mutants showed a loss or reduction of binding to PI(4,5)P2.

Surprisingly, the K13R/R14K switch mutant showed a stronger binding to PI(3,4)P2 than wt

PTEN (Redfern, Ross, Gericke, personal communication). These results suggest that PTEN’s N- terminal forms a rigid structure around the PI(4,5)P2 headgroup in which several amino acid side chains make contact with the headgroup.

Objectives and Methodology

While the charge for PI(4,5)P2 has been successfully characterized in simple binary lipid systems, the physiological membrane environment is considerably more complicated. In the inner leaflet of the cell plasma membrane there are multiple phospholipid species, including phosphatidylethanolamine (PE), phosphatidylinositol (PI), and phosphatidylserine (PS), each of which could potentially interact with PI(4,5)P2 (discussed in chapter 4). The potential for

PI(4,5)P2 domain formation further complicates the situation. Calcium and magnesium ions in the cytosol can both interact electrostatically with PI(4,5)P2, and potentially cause clustering of the lipid (Wang, Collins et al. 2012; Sarmento, Coutinho et al. 2013). Cholesterol within the membrane has been shown to affect PI(4,5)P2 signaling, and may also be involved in PI(4,5)P2 domain formation (Dasgupta, Bamba et al. 2009; Jiang, Redfern et al. 2014). Each of these components have a strong potential to interact with PI(4,5)P2 and may influence its involvement in signaling events (discussed in chapter 5). In this study, I have used solid-state MAS 31P NMR to examine MLVs containing each of these components in combination with PC and PI(4,5)P2.

31 P NMR enables a determination of the pH dependent change in PI(4,5)P2 charge, as well as an

29 indication of the strength of the interaction. The pH effect on each of the phosphates can be examined individually, as each peak is well resolved in the 31P NMR spectra.

The interaction between divalent cations and PI(4,5)P2 can be further studied using x-ray

2+ 2+ techniques. Many previous studies have investigated Ca and Mg binding to PI(4,5)P2 (Wang,

Collins et al. 2012; Sarmento, Coutinho et al. 2013). However, many of these studies have relied on indirect detection of cations, such as changes in monolayer condensation or altered diffusion of PI(4,5)P2 due to calcium binding. These studies are also often done in the absence of physiological salt levels. In this study, I have used x-ray fluorescence spectroscopy near total x-

2+ + ray reflection to measure Ca and K binding to a natural PI(4,5)P2 monolayer. X-ray fluorescence has been used extensively to study cation binding to charged lipid layers (Bu, Ryan et al. 2006; Bu, Flores et al. 2009; Kooijman, Vaknin et al. 2009). The incident x-ray beam excites cations within the solution, leading to fluorescence via their Kα and Kβ emission bands.

By using an x-ray angle that is less than the critical angle (αc), only a select portion of cations near the surface are excited, and thus the cation levels bound to the PI(4,5)P2 headgroup may be determined. In addition to determining cation binding, I have also evaluated the effect of cation binding on PI(4,5)P2 structure using x-ray reflectivity (discussed in chapter 6).

The interaction between PTEN and PI(4,5)P2 is critical for PTEN activity (Campbell, Liu et al. 2003). As PTEN is one of the most frequently mutated tumor suppressor proteins in humans, it is essential that we learn more about this interaction. At this point, the exact stoichiometry of the interaction between PTEN’s N-terminus and PI(4,5)P2 is unknown. There is a wealth of thermodynamic and kinetic data detailing the PI(4,5)P2/PTEN interaction (Gericke, Munson et al. 2006; Ross and Gericke 2009; Shenoy, Shekhar et al. 2012), however there is a lack of information about the structural details of the interaction. I have attempted to investigate the

30 structure of this interaction using a variety of NMR techniques, including solid-state and solution techniques. Large MLVs containing POPC and brain PI(4,5)P2 were used to simulate the membrane environment. For solution NMR experiments, a water soluble PI(4,5)P2 analog was used (kindly provided by Michael Best, University of Tennessee). Peptides were derived from the first 21 amino acids of the PTEN N-terminus (PTEN1-21: MTAIIKEIVSRNKRRYQEDGF), as this sequence is known to have the same binding preference for PI(4,5)P2 as the full length protein (Redfern, Redfern et al. 2008). The effect of this interaction on the lipid was examined using 31P NMR and NMR relaxation measurements. Solution NMR experiments were used to determine which residues interact with PI(4,5)P2 and where the most significant interactions occur (discussed in chapter 7).

Chapter 2

Methods

31P NMR Titration

Previously, 31P NMR was used to determine the ionization behavior of phosphomonoester containing lipids in micellar or small unilamellar vesicle (SUV) dispersions (Hauser 1989;

Swairjo, Seaton et al. 1994). In these particular cases, 31P NMR experiments result in isotropic chemical shifts due to the rapid reorientation of the lipids with respect to the external magnetic field. pH titration curves can readily be prepared for such systems. The problem with these experiments is that they do not represent the native packing environment of a biological membrane. Micelles and SUVs are systems in which the lipids experience a high degree of membrane curvature (+ for the micelles, and both + and – for the SUV (van Dijck, de Kruijff et al. 1978)). A better model system are multilamellar vesicle (MLV) suspensions, where the lipids reside in an essentially flat membrane. However, 31P NMR of MLV dispersions results in a chemical shift anisotropy (CSA) that is representative of the organization and orientation of the lipids but yields little information on the ionization of the individual lipids making up the lipid membrane. An example of the CSA for a mixture of phosphatidylcholine (PC) and lysophosphatidic acid (LPA) is shown in Figure 2-1A. Note the different CSA profiles for PC and LPA representing differences in their motion and orientation in the membrane. Chemical shift values observed in a solid state magic angle spinning (MAS) experiment, where we average

A DOPC

LPA

40 20 0 -20 -40 ppm

A DOPC B DOPC

LPA

40 20 0 -20 -40 20 15 10 5 0 -5 -10 -15 -20 ppm ppm

Figure 2-1. 31P NMR of an MLV dispersion. A, static 31P NMR spectrum of a MLV dispersion of B 20mol% LPA and 80 mol% DOPC. The CSA for DOPC is ~ 41 ppm, and the CSA for LPA is ~10 DOPC ppm. B, MAS 31P NMR spectrum of the same lipid mixture. Peaks are relative to an external 85 wt%

H3PO4 standard. Taken from (Graber and Kooijman 2013).

out most orientation dependent interactions, enable us to study lipid ionization properties (Watts

1998; Kooijman, Carter et al. 2005). These MAS experiments form the foundation of the LPA method to construct pH titration curves described here. Figure 2-1B shows the MAS spectrum for the20 same15 mixture10 5 of PC/LPA0 -5 as- 10shown-15 in 2-20-1A. The chemical shift of LPA ppm (phosphomonoester) is very sensitive to pH and thus allows a pH titration curve to be determined by preparing many samples with different pH values (Kooijman, Carter et al. 2005; Kooijman,

Tieleman et al. 2007; Kooijman and Burger 2009). An example of such a pH titration curve for brain PI(4,5)P2 (bPI(4,5)P2) in dioleoyl-phosphatidylcholine (DOPC) is shown in Figure 2-2A together with the raw NMR spectra (Fig. 2-2B) from which the data points for the curves are taken (data taken from (Kooijman, King et al. 2009)).

31 31 Figure 2-2. P NMR titration curve for bPI(4,5)P2 in a DOPC matrix. A, P MAS NMR data for

MLVs composed of 95% / 5% DOPC:bPI(4,5)P2. The 4, and 5-phosphate peaks of bPI(4,5)P2 are indicated. B, pH titration curves for the 4, and 5-phosphate of bPI(4,5)P2 based on the observed peak values from A. Phosphorus chemical shift is relative to an 85 wt% H3PO4 standard. Data taken from Kooijman et al., Biochemistry 2009.

Lipid samples for NMR are made using concentrated lipid stock solutions. Lipid stock solutions are prepared from lipid powder. Unsaturated lipids should be purchased in powder form as the shelf life of prepared stock solutions in organic solvent is considerably less than the powder form (due to the slight acidity of chloroform). Both lipid powder and prepared stock solution should be stored under N2 (g) at or below -20°C. Polar lipid powders were dissolved in a

2:1 (by volume) chloroform/methanol solution. Lipids were carefully weighed on a semi-micro balance and dissolved in an exact volume of organic solvent using a volumetric flask. For highly charged lipids such as bPI(4,5)P2, a 20:9:1, by volume, mixture of chloroform, methanol, and water was used to dissolve the lipid. For some lipids (e.g. bPI(4,5)P2), some precipitation occurred, especially when warming it up after storage at -20°C. To fully dissolve the lipid the precipitated stock was heated in a warm water bath (~40°C) and sonicated briefly as required.

The concentration of the resulting lipid stock solutions was generally tested via phosphate assay

(Rouser, Fkeischer et al. 1970), except for PI(4,5)P2. Phospholipid stock solutions were tested for

34 purity via thin layer chromatography using a 65:25:4 (chloroform, methanol, water) running solvent on silica gel plates with a thickness of 250 µm. Lipids were deemed pure if only a single spot was observed, and no additional band in the front appeared (indicative of acyl-chain hydrolysis).

Preparation of Lipid Films

Vesicles for the NMR experiments were formed from dried lipid films which were obtained by mixing appropriate amounts of the respective lipids in organic solvents. Lipid films were prepared in 15 mm borosilicate glass tubes (i.d. = 1.5 cm) with a joint fitting enabling the tubes to be placed on the rotary evaporator. Previous work using bPI(4,5)P2 showed that a special sample prep procedure was required to avoid the creation of metastable, presumably non-bilayer phases (see supplemental material of (Kooijman, King et al. 2009)). The rotovap/test tube procedure described herein was found to yield the best results. Appropriate amounts of lipid stock were mixed in the test tubes to form an organic lipid solution. 4-10 μmol of lipid were used for each film. The amount of PI(4,5)P2 in each mixture was kept to a minimum due to the high expense of the lipid. Typically each film contained 0.2 μmol of PI(4,5)P2, as this was the minimum amount we could use and obtain good spectra within a reasonable experiment time.

For cheaper target lipids (e.g. PA), we used higher amounts of lipid (typically 0.5 µmol) in order to decrease the experimental time (doubling of the # of µmols of lipid results in a 4 times decrease in experimental time). Roughly ~400-600 μL of CHCl3 were added to each film to increase total solution volume to ~800-1200 μL. This extra volume of chloroform was found to help eliminate the appearance of metastable phases when films were prepared with the rotary evaporator (Kooijman, King et al. 2009). Lipid films were dried by using a rotary evaporator to remove the organic solvent (roughly 3-10 min). The rotary evaporator water bath temperature

35 was set to 40-45°C. This temperature was selected as it is well above the highest Tm for the lipids used (lipids should all be in the fluid phase during sample preparation), while it is not hot enough to boil the solvent away too quickly (rapid boiling will lead to poor quality lipid films) or promote lipid breakdown of the sensitive polyunsaturated fatty acids of bPI(4,5)P2. The dry lipid films were placed in a vacuum oven under vacuum (at ≥ 25 in Hg) for ≥ 4 hrs to remove any residual organic solvent. Dried lipid films were stored under N2 (g) at -20°C.

Preparation of MLV dispersions

Lipid films need to be hydrated to form MLV dispersions. The dried lipid films were hydrated with 2 mL of buffer and vortexed to create the MLV dispersion. The following buffering agents were used for the indicated pH ranges: 20 mM citric acid/ 30 mM MES for pH 4-6.5, 50mM

HEPES for pH 6.5-8.5, and 50 mM glycine for pH 8.5-10. The buffer also contained 100 mM

NaCl to mimic cellular ionic strength, and 2 mM EDTA in order to complex trace amounts of divalent cations. For MLVs that compared different mixtures at a constant pH value, a pH 7.20

(± 0.05) buffer with 100 mM HEPES and 100 mM NaCl was used. EDTA was excluded from all samples that included divalent cations. The resulting MLVs were flash frozen in an ethanol/dry ice mixture and then gently thawed in warm water while occasionally vortexing the sample. This

“freeze-thaw” technique helps to remove metastable lipid phases and provides a more homogeneous size distribution of the MLVs. The freeze-thaw cycle was repeated 1-2 times for most lipid mixtures. However, for the highly anionic PI(4,5)P2, no more than two freeze- thaw cycles were done, as freeze-thawing more than this can create small enough vesicles to interfere with the MAS NMR experiment (Brownian motion of small vesicles opposes the averaging accomplished by the MAS NMR technique). For samples containing divalent cations, A23187 ionophore was added to the lipid vesicles in a 1:1000 mol% ratio to equilibrate divalent cations

36 across the MLV lipid membranes (typically 1-10 µL of ionophore solution). The A23187 ionophore is known to have an affinity for each of the divalent cations that we tested (Pfeiffer and Lardy 1976). Addition of ionophore at these concentrations has no effect on the 31P NMR spectra of model lipid membranes (Kooijman, Carter et al. 2005). Appropriate volumes of a 100 mM MeCl2 (Me: divalent cation) solution were added to set the concentration of divalent cations in the sample. The lipid mixture was vortexed thoroughly and left to sit for 15 min to allow equilibration of the ionophore and divalent cations throughout the vesicle (multilamellar vesicles,

MLVs) suspension. The pH of the lipid suspension was measured using a Sentron Intelli

CupFET pH probe (Sentron, Roden, The Netherlands). A regular glass electrode is incompatible with these concentrated liposome dispersions. The measured pH is used to construct the pH titration curve. The MLVs are spun down in a table top, temperature controlled, Eppendorf type centrifuge for 45-60 min at 15,000 rpm (or highest possible rpm) using a 2 mL centrifuge tube compatible rotor. The centrifuge is refrigerated at 4°C in order to reduce the risk of lipid degradation. The supernatant was removed and the lipid pellets were collected and transferred into a 4 mm zirconia (ZrO2) MAS NMR tube. When filling the NMR tube, air bubbles can easily form in the viscous lipid dispersion. These bubbles must be eliminated; otherwise the NMR rotor may not spin properly. Any air bubbles were removed by “stirring” the solution with a thin stirring rod to pop any bubbles that were present. It is extremely important not to overfill the

NMR tube. If the tube is too full, the cap will not stay on, and may come off during the NMR experiment, leading to loss of sample and a contamination of the MAS probe. The lipid level was carefully adjusted within the sample tube and the cap was checked to make sure it sealed tightly.

NMR Experimental Procedure

31 The P NMR chemical shift values were all referenced to an external 85% H3PO4 standard. A

31 single dedicated NMR tube was filled with 85% H3PO4 and used as the standard for all P NMR experiments. The standard was run with a MAS spin rate of 1,000-2,000 Hz. A low spin rate was used to reduce the strain on the cap of the standard tube in order to prevent possible instrument damage due to spilled H3PO4. Several scans are acquired and the location of the single H3PO4 peak is used to set the spectral reference value. The lipid sample is placed into the

NMR spectrometer and set to an initial MAS speed of 2,000 Hz. Once the sample spins without difficulty, the spin rate is increased to 5,000 Hz. When the spinning stabilizes, the spectrometer is tuned to the 31P resonance (161.97 MHz). Generally, 30,000 to 50,000 scans were recorded

(when using 0.2 µmol of target lipid). All experiments were conducted at 22.5 ± 1.5 °C. The pulse program and pulse parameters generally used for this experiment are shown in the appendix.

After running the MAS experiment, a static solid state 31P NMR experiment was used to examine the phase of the lipid solution. A WALTZ16 pulse program was used for low-power proton decoupling. For static experiments the spectrometer must be tuned to the 31P NMR frequency and the 1H NMR frequency. The static experiment is fairly noisy, and takes many scans to create a spectrum with a good signal to noise ratio. 20,000 scans will give a rough spectrum and can indicate the primary phase, but 100,000 or more scans may be required to create a nice smooth spectrum (after 50 Hz exponential line broadening). Pulse program and pulse parameters for the static experiment are shown in the appendix.

31P NMR Data Analysis

A titration curve may be established by plotting the chemical shifts of the phosphate peaks vs. the pH of the samples (see Figure 2-2A for an example for 5 mol% PI(4,5)P2 in 95 mol%

DOPC). These peaks are picked by the TopSpin software provided with the NMR spectrometer.

Higher chemical shift values for the phosphomonoester peak indicate deshielding of the 31P nucleus and a corresponding increase in deprotonation. The degree of protonation can be calculated from the chemical shift according to,

표푏푠 훿푖 − 훿푖,푑 푓푖,푝 = , (2-1) 훿푖,푝 − 훿푖,푑

표푏푠 where fi,p is the degree of protonation for phosphomonoester group i, 훿푖 is the pH-dependent chemical shift such as shown in Figure 2-2A, and δi,p and δi,d are the chemical shifts of the singly protonated and completely deprotonated form of phosphomonoester group i (for phosphoinositides, the phosphomonoester group is labeled according to its position on the inositol ring). The chemical shift for the singly deprotonated and doubly deprotonated states can be estimated from the low pH (~4) and high pH (~10) data. Alternatively the data can be fit with a Henderson-Hasselbalch type equation detailed in reference (Kooijman, Carter et al. 2005). The chemical shift values for these two states are variables in this equation and can thus be determined exactly. This works well for sigmoidal titration curves as observed for PA, ceramide-

1-phosphate, and other lipids containing a single phosphomonoester. For bPI(4,5)P2 the titration behavior over the pH range of 4

Figure 2-3 shows the result of this 1.2 calculation which is used to determine 1.0 the charge on each of the 0.8 0.6

i,p phosphomonoesters of bPI(4,5)P . Note f 2 0.4 that we use the assumption that the 0.2 4-phosphate 0.0 chemical shift at any pH can be 5-phosphate considered as a weighted average of the 4 5 6 7 8 9 10 pH concentration of the protonation states Figure 2-3. Degree of protonation for bPI(4,5)P2. multiplied by the chemical shift for that Degree of protonation data for the 4, and 5-phosphate of bPI(4,5)P2 calculated using equation 1 from the state, i.e.: data shown in Figure 2-2A.

[퐴]훿 +[퐵]훿 훿 = 퐴 퐵, (2-2) [퐴]+ [퐵] where δ is the chemical shift, [A] is the concentration of the protonated form, and [B] is the concentration of the deprotonated form. δA and δB are the chemical shifts of the protonated and deprotonated forms respectively, which are determined from the data shown in Figure 2-2A.

Error Determination and ANOVA Analysis

For MLVs that compared different mixtures at a constant pH value, the chemical shift values and errors are the average and standard deviation for at least 3 independent samples at each of the conditions investigated. The exact number of repeats is listed in the legend of each figure

(see chapter 5). The standard deviation calculated for the charge of each phosphomonoester of

PI(4,5)P2 is derived from the charge calculated for each repeat experiment and then follows from

40 the average determination of this charge value. ANOVA analysis for statistical significance was performed using a Holm-Sidak pairwise comparison of the sample sets. A difference was considered significant for P < 0.005.

31P Spin-lattice relaxation experiments

The spin-lattice relaxation time, T1, indicates how long a given nucleus takes to relax upon excitation by an NMR pulse. After a 90° pulse, the magnetization will be along the x axis normal. The nucleus will then slowly relax from this excited state to its original equilibrium M0.

This process is governed by the constant T1, as follows:

−푡/푇1 푀0 − 푀푧 = 퐴푒 , (2-3)

where M0 is the original magnetization along the z axis, Mz is the current magnetization along the z axis, and t is the time elapsed since the original NMR pulse. Thus T1 is the time at which

-1 the magnetization along the z axis is equal to (1 – e )M0 or 0.63M0. The T1 constant can be measured using an inverse recovery experiment, in which the nucleus is first pulsed with a 180° pulse, transferring the magnetization to Mz =

-M0. The system is then allowed to relax for a variable delay time before being pulsed with a 90° pulse.

Depending on how quickly the system Figure 2-4. Relaxation of the nucleus over time. Plot of the magnetization of the z axis (Mz) as a function of time relaxes, you may measure a negative (t) immediately after a 90° pulse. The magnetization eventually returns to the initial value, M0. T1 governs the relaxation time.

41 peak, a positive peak, or no peak at all (which occurs at ln(0.5)T1). By using a variety of delay times we can determine the relaxation constant T1.

The spin-lattice relaxation time, T1, depends on the dynamics of the studied system but can vary based on a large number of factors such as dipole-dipole interactions, spin-rotation, chemical shift anisotropy, or interactions with unpaired electrons. By measuring the T1 in the presence and absence of a ligand and its binding partner, we can determine the change in dynamics associated with the binding.

For our T1 measurements we used MLVs containing DOPC and PI(4,5)P2 at a 9:1 ratio.

MLVs were formed as described above for the MAS NMR experiments. In order to reduce the acquisition times, 1 µmol of PI(4,5)P2 was used for each sample. The T1 value was measured first for the PC:PI(4,5)P2 vesicles alone. Next we added the peptide PTEN10-16 at a 1:1 molar ratio with PI(4,5)P2. The peptide was added from a concentrated stock, resulting in minimal dilution. The MLV mixture was then vortexed and subjected to two freeze/thaw cycles to help incorporate the peptide throughout the layers of the MLVs.

T1 NMR Experimental Procedure

The T1 experiments were performed using our solid-state MAS equipped 400 MHz NMR spectrometer. As with the 31P NMR MAS experiments above, the chemical shift values were all referenced to an external 85% H3PO4 standard and the spin rate was set to 5,000 Hz. A standard inversion recovery pulse program was used (see appendix), with a single 180° pulse followed by a variable delay before a final 90° pulse. Sixteen delay times were tested for each sample, from 1 ms to 10 s (see appendix for full delay list). The sample was pulsed for ~1,000 scans for each

42 delay time. The T1 values for each phosphate were calculated from the resulting spectra using the Topspin T1 relaxation module.

Solution NMR Experiments

Solution NMR experiments are well developed and have been used extensively for determining the structure of organic molecules in solution. Here I have used basic solution NMR

1D, COSY, NOESY, and HSQC experiments to examine the structure of my target molecules in solution. I have used each of the nuclei, 1H, 13C, 31P, and 15N. COSY experiments measure through bond coupling between nuclei allowing us to determine the structural sequence of the target molecule. NOESY experiments measure through space coupling between nuclei, which allows us to determine the spatial arrangement of the molecule (or molecules). Finally HSQC (or

HMBC) experiments provide us with a correlation between the 1H spectra and the 13C or 15N spectra.

For solution NMR experiments we used aqueous solutions containing the soluble peptide

PTEN10-16 and either micelles of 1,2-dioctanoyl-sn-glycero-3-phosphatidylinositol-4,5- bisphosphate or the fully soluble derivative PI(4,5)P2-amine. All aqueous solutions contained

10% D2O to enable us to lock onto the deuterium signal in the spectrometer. Generally 15 mM phosphate buffer and 100 mM NaCl were included to prevent precipitation of the peptide-lipid mixture and to provide a more physiologically relevant mixture. The pH was set to either ~5.2 or

7. The lower pH was used for some 1H experiments in order to reduce the exchange rate of the amine and amide protons to allow them to be observed. pH values were measured using our

Sentron Intelli CupFET pH probe (Sentron, Roden, The Netherlands). In D2O solutions the pH

43 probe actually measures the pD of the target solution. This must be adjusted to pH using the relation below (Pentz and Thornton 1967):

푝퐻 = 푝퐷 + 0.3139훼 + 0.0854훼2, (2-4)

where α is the fraction of D2O in solution, in our case 0.1. Thus, using Eqn. 2-4 we find that the pH ≈ pD + 0.03. Generally, the lipid/peptide was dissolved in 0.6 mL of buffer to yield a final concentration of ~5 mM. A 1:1 peptide/lipid ratio was used for the majority of experiments.

Solution NMR Procedure

Solution NMR experiments were carried out in a 500 MHz spectrometer. Standard pulse sequences were used for gradient COSY, NOESY, ROESY, HSQC and HMBC experiments. In order to suppress the water peak, a two second presaturation pulse with a 5 dB power level was centered on the water resonance (4.65 ppm). Generally NOESY or ROESY experiments were run overnight to obtain the best possible signal to noise ratio.

Surface sensitive X-ray experiments

The surface sensitive x-ray experiments of X-ray reflectivity (XRR) and near-total-reflection fluorescence (XNTRF) are very useful for examining lipid monolayers. X-ray reflectivity experiments allow us to examine the structure of the lipid monolayer in the z-direction, enabling us to determine the thickness of the monolayer as well as the electron density. X-ray fluorescence excites ions near the interface of the lipid monolayer, leading to a fluorescence signal from those ions. In this way, the binding of ions to negatively charged lipid interfaces can be determined (Bu, Flores et al. 2009).

To study cation binding to PI(4,5)P2 we used x-ray fluorescence on a monolayer composed of pure bPI(4,5)P2. X-ray reflectivity (XRR) and near-total-reflection fluorescence (XNTRF) measurements were conducted on the liquid surface spectrometer (LSS) at beamline 9ID-C,

Advanced Photon Source (APS), Argonne National Laboratory. Fig. 2-5 provides a sketch of the experimental setup. The monolayer is formed by drop-wise deposit of PI(4,5)P2 stock solution

(see above regarding stock preparation) onto the aqueous surface of the subphase solutions contained in a Langmuir trough that is maintained at constant temperature of 20 °C. Buffer subphase was made by dissolving appropriate amounts of chemicals in MilliQ water. The buffer subphase contained 10 mM Tris and 100 mM KCl, along with varying amounts of CaCl2, MgCl2, and EDTA. Buffer pH was set to pH 7.2 using high purity HCl, 10 mM Tris was used to ensure that the pH did not vary. In order to replicate physiological salt levels, 100 mM KCl was included in the subphase. For buffers with no divalent cations, 0.1 mM EDTA was used to chelate any trace divalent cation impurities. The Langmuir trough is encapsulated in an enclosure with Kapton windows and purged with water-saturated helium gas to minimize background scattering from air and potential X-ray radiation damage to the samples. An oxygen sensor

(S101, Qubit System Inc.) monitors the oxygen level in the enclosure. The monolayer is compressed with a Teflon barrier to our desired surface pressure of 30 ± 2 mN/m. Surface pressure is monitored by a Wilhelmy microbalance and filter paper plate. The X-ray measurements are started after the oxygen-helium exchange reaches equilibrium, where the oxygen content inside the enclosure is reduced by a factor of at least 100 with respect to that of the ambient atmosphere. The monolayer is maintained at π = 30 ± 2 mN/m during the course of the experiments. Radiation damage is examined by systematically moving the trough laterally to probe fresh surfaces and compare reproducibility. All reflectivity measurements were found to

Figure 2-5: X-ray experimental Setup. Illustration of X-ray fluorescence and reflectivity 2+ measurement’s setup on the PI(4,5)P monolayer over a subphase containing divalent ions M , 2 where M represents Ca or Mg. The box enclosed by the dashed lines is a side view of the interfaces of a vapor phase (water-saturated helium), monolayer and aqueous bulk. The X-ray beam impinges on the surface at an incident angle α and reflects at an exit angle α and α = α in a i f i f specular reflectivity measurement. The probe of the EDD subtends the center of the sample surface with a collimator in front of it to define the EDD footprint that is smaller than the X-ray beam footprint and only accept X-ray photons of the fluorescence signals and background scattering (emanated from bulk water and helium) restricted in the surface normal direction.

46 be reproducible.

The highly monochromated and collimated X-ray beam (E = 8.0 keV; wavelength λ = 1.5497

Å) passes through an ionization chamber (incident beam intensity monitor) and variable beam attenuator (multilayers of nickel foil, for attenuation of the primary incident beam), and is steered onto the aqueous surface by a second monochromator to a desired incident angle αi with respect to the surface. The specular reflection is collected by a Bicron scintillation detector at an X-ray exit angle αi =αf. The angular-dependence of both XRR and XNTRF are expressed as functions of Qz, the z-component (surface normal) of the scattering (wave vector transfer) vector Q. Qz is related to the incident angle αi via Qz = (4π/λ) sin αi. An incident angle of αi ≈ 0.154° (at E = 8 keV) corresponds to the critical angle (αc) where there is total reflection. Qc is used to represent the Qz corresponding to the critical angle, αc.

The fluorescence data for a sample surface are collected at a series of αi near the critical angle

αc by a Vortex energy dispersive detector (EDD, silicon-drift, Vortex-90EX) with a collimator in the front end of the EDD probe which only accepts the X-ray fluorescence and scattering photons emitted in the direction along the surface normal (~1° angular resolution).

Fluorescence signals from a pure water subphase are measured as a background to be subtracted from the samples. The EDD spectra also contain the primary beam signals at E = 8.0 keV mainly from Thomson elastic scattering of bulk water and Compton inelastic scattering

(from bulk water and water-saturated helium, at a lower energy shifted from E= 8.0 keV by ≤ 0.1 keV). Both fluorescence and reflectivity data are normalized to incident beam intensity (more details can be found elsewhere (Vaknin 2012; Wang, Murthy et al. 2013)).

X-ray Data Analysis

X-ray reflectivity data, R(Qz), are normalized to the calculated Fresnel reflectivity (RF), for an ideally smooth, flat air-water interface (Als-Nielsen and McMorrow 2011; Pershan and

Schlossman 2012). The R/RF data for a surface monolayer on an aqueous surface can be accounted for in terms of a simplistic, two-box structural model, where one box contains the hydrophilic head group, and the other contains the hydrocarbon tail (Kjaer 1994; Als-Nielsen and McMorrow 2011; Vaknin 2012). Each box is characterized by a uniform electron density (ρ) and vertical height (l) relative to the interface. Given (ρH, lH) and (ρT,lT), in which subscripts ``H'' and ``T'' represent the headgroup and the hydrocarbon tail respectively, a step-like, discrete ED profile across the interfaces, ρ0(z), is constructed and its corresponding reflectivity R0(Qz) can be calculated (Als-Nielsen and McMorrow 2011; Vaknin 2012). The realistic ED profile, ρ(z) is obtained by smearing ρ0(z) with an interfacial roughness, σ, mainly arising from capillary waves

(Pershan and Schlossman 2012), which lowers the R0(Qz) by a ``Debye-Waller-like'' factor, exp(-

2 2 Q z σ ) as follows (Als-Nielsen and McMorrow 2011; Pershan and Schlossman 2012; Vaknin

2012):

2 2 −푄푧 휎 푅⁄푅퐹 = (푅0/푅퐹) e , (2-5)

The Parratt's exact recursive method (Als-Nielsen and McMorrow 2011), along with Eqn. 2-5, is applied to calculate the R/RF based on the structural parameters (ρH,T, lH,T and σ). The structural refinements are carried out through a least squares optimization method (Vaknin 2012).

Surface fluorescence studies have been performed in the past (Yun and Bloch 1990; Daillant,

Bosio et al. 1991) and analytic routines to quantify the surface enrichment of ions have been established (Bu, Flores et al. 2009; Wang, Anderson et al. 2012) and are discussed below.

At αi < αc, the X-rays penetrate into the liquid by a very shallow depth (< 10 nm) in the direction of the surface normal (Als-Nielsen and McMorrow 2011; Vaknin 2012). At αi > αc, the

X-rays penetrate into the bulk by a few micrometers (within the X-ray energy of 8-16 keV typically used for liquid surface scattering). The combination of the shallow X-ray penetration depth and surface intensity enhancement arising from evanescence when αi < αc provide the surface sensitivity of the XNTRF method.

In this study, the fluorescence measurements were conducted over the range 0.5 αc < αi < 1.5

-1 -1 αc, corresponding to Qz = 0.01-0.035Å , with a spacing of 0.001Å between consecutive Qz points. The fluorescence data collected at Qz < Qc and Qz > Qc are referred to as surface fluorescence and bulk fluorescence signal respectively.

To calibrate the spectral intensity to known quantities of fluorescing ions, the fluorescence data are compared from two bulk solutions of CaCl2 (pre-acidified with a small amount of HCl to prevent potential formation and precipitation of Ca(OH)2) and KCl both prepared at 100 mM with a bare air/water interface. Figure 2-6 shows the individual bulk fluorescence spectra for

Ca2+ and K+. Each spectrum within the energy range shown in Figure 2-6 is characterized with two emission lines, Kα and Kβ, each of which is further profile-fit with a gaussian function centered on the emission line energy. The finite spread of the emission lines is a measure of the

EDD energy resolution at each emission line energy. The bulk fluorescence spectra shown in

Figure 2-6 provide the “standard spectral profile” for each element, which characterizes (1) the relative intensity ratio of Kα to Kβ and (2) the relative line(s) intensity ratio of calcium to potassium, given the same bulk concentrations. As 100 mM KCl is present in relevant samples and the potassium Kβ line partially overlaps with the calcium Kα line, the standard spectral profiles serve to discriminate the potassium and calcium contribution when both are present in

Figure 2-6: Bulk Fluorescence Spectra. The bulk fluorescence spectra for a pure KCl and a CaCl2

solution as indicated. Both solutions were prepared at 100 mM. CaCl2 solutions were acidified to -1 prevent calcium precipitation as Ca(OH)2. The spectra were integrated over Qz = 0.025-0.035 A to demonstrate the relative intensity of emission lines of calcium relative to potassium sensed by EDD.

2+ the spectra. The Mg ion emission lines are Kα = 1.25 keV and Kβ = 1.07 keV. Due to the low energy of these emission lines we were unable to excite the magnesium ions using our 8 keV x- ray beam. At each Qz, the fluorescence intensity of a specific emission line is integrated over its spread in the spectrum and expressed as a function of sin αi (i.e., proportional to Qz). For a solution of fluorescing ions with a bare surface the fluorescence intensity, Ib (αi), is proportional to its bulk concentration nb and the volume illuminated by X-ray, and can be expressed as follows:

2 퐼푏(훼푖) = 퐶|푡퐹(훼푖)| 퐷(훼푖)푛푏, (2-6)

Figure 2-7: Fluorescence Qz Dependence. Fluorescence intensities of calcium emission line(s) as a

function of Qz for a solution of CaCl2 at concentration of 10 mM covered with a PI(4,5)P2 monolayer

(red) and an equivalent 10 mM CaCl2 solution with a bare surface (black). Each data point represents

the intensity integrated exclusively over the Ca Kα emission line in the spectrum. The solid lines are best-fit profiles calculated in terms of Eqn. 2-6 for the bulk contribution and the sum of Eqn. 2-6 and

Eqn. 2-7 for the solution covered with the PI(4,5)P2 monolayer. The dashed line is solely the surface contribution in terms of Eqn. 2-7.

where C is an element-specific scale factor that accounts for elemental fluorescence yield and detector efficiency, tF(αi) is the x-ray Fresnel amplitude transmission coefficient at a sharp, flat vapor/liquid interface, and D(αi) is the x-ray penetration depth normal to the surface (Als-Nielsen and McMorrow 2011; Vaknin 2012). For a thin layer of surface excess ions, their fluorescence contribution, Is (αi), can be expressed as follows:

2 −|푧푖표푛|⁄퐷(훼푖) 퐼푠(훼푖) = 퐶|푡퐹(훼푖)| 퐷(훼푖)푛푠푒 , (2-7)

51 where zion is the z-coordinate of the ion-enriched layer with z = 0 at the monolayer/vapor interface.

For a cation solution with a bare surface, the fluorescence comes solely from the bulk contribution (see the black data points in Figure 2-7). In this case the ions are randomly distributed throughout the solution, and very few ions are observed in the 5 nm surface region.

Thus, for Qz < Qc we observe very little fluorescence. For Qz > Qc, the ions in the bulk solution are excited and we observe high fluorescence intensity. For a cation surface covered with a lipid monolayer, we observe fluorescence from the cation-enriched surface as well as from the bulk

(see Figure 2-7 in red). The fluorescence contribution from the ions at the surface can be determined by subtracting the bulk contribution (shown in Figure 2-7 with a blue dashed line).

The number of cations at the surface can be calculated for a cation surface covered with a lipid monolayer by subtracting the bulk contribution below the critical angle (Qc):

퐼 (훼 )−퐼 (훼 ) 푠 푖 푏 푖 푧푖표푛⁄퐷(훼푖) 푛푠 = 퐷(훼푖)푛푏푒 , (2-8) 퐼푏(훼푖) resulting in the ion density at the surface. If the mean molecular area of the lipids is known, then the number of ions per lipid can be calculated by simply multiplying the ion density by the area per lipid.

Giant Unilamellar Vesicle Preparation

Giant unilamellar vesicles (GUVs) are 10-100 µm sized lipid vesicles with a single lipid bilayer. GUVs are a good model for the cellular membrane due to their similar size. The large size of GUVs enables them to be observed via microscope. If we then add fluorescently labeled

52 lipids to the GUVs we can observe the localization of these fluorescent lipids. I will use these

GUVs to study the localization of PI(4,5)P2 in PC:PS:PI(4,5)P2 and PC:PI:PI(4,5)P2 systems.

With a few changes, we followed the method described by Akashi et al. (Akashi et al., 1996).

The lipid mixtures with the desired lipid composition were mixed with 0.1 mol% of the fluorescently labeled lipid, RhB DOPE, and diluted with 300 µL of organic solvent (chloroform: methanol =2:1 (by volume)). RhB DOPE is known to prefer disordered lipid environments. A lipid film was then formed using the procedure described above. The completely dried lipid film was then prehydrated at 50oC with water-saturated nitrogen gas for 30-45 minutes. Two mL of an N2 purged aqueous solution (pH 7 buffer made from 100 mM NaCl, 5 mM HEPES, and 0.1 mM EDTA) was added gently to the test tube. The tube was sealed under nitrogen and wrapped with aluminum foil, then incubated at 50°C overnight (>18 h). GUVs develop spontaneously but slowly during the incubation. After that, the GUV suspension was slowly cooled down to room temperature.

About 400 µL of the diluted GUV suspension (10 times dilution from the original GUV suspension) was transferred to an 8-well Lab-Tek chamber (Lab-Tek, 155411) (Rochester, NY).

An inverted microscope (Olympus IX 71) equipped with a 60x oil immersion objective and a

CCD camera (Retiga 1300, QIMAGING) (Burnaby, BC, Canada) was used. Images were post- processed to JPEG or TIFF format with Image J.

Chapter 3

Modeling the Complex Ionization Behavior of Polyphosphoinositides

Introduction

The polyphosphoinositides PI(3,4)P2, PI(4,5)P2, PI(3,5)P2 and PI(3,4,5)P3 each assume important signaling roles in the cell. Despite the similarities of their structures, these phosphoinositides often have highly distinct roles. The distinction between these phosphoinositides is partly determined by subcellular location. PI(3,5)P2 is primarily produced and acts within multivesicular bodies and late endosomes (Takasuga, Horie et al. 2013). PI4P is found in the Golgi and contributes to the PI(4,5)P2 population in the plasma membrane (Dickson,

Jensen et al. 2014). PI3P is found in endosomes (Schink, Raiborg et al. 2013). The other polyphosphoinositides are primarily found in the inner leaflet of the cell plasma membrane (van

Meer, Voelker et al. 2008). A substantial pool of phosphoinositides also exists in the nuclear envelope (Zhendre, Grelard et al. 2011). The phosphoinositides are also differentiated by subtle differences in their headgroup structure. The rearrangement of phosphate groups on the inositol ring can alter the geometry of the headgroup and thus alter the interaction of the phosphoinositide with protein binding domains. In the case of PI(4,5)P2, the interaction between

PTEN and PI(4,5)P2 has been shown to be highly specific for the arrangement of the phosphate groups, as PTEN does not bind to PI(3,4)P2 (Redfern, Redfern et al. 2008). The specific phosphate arrangement may also affect the charge of the headgroup. For PI(3,4,5)P3 the

54 additional phosphate group will obviously lead to a higher charge, and therefore, to potentially stronger electrostatic interactions. In 2009, Kooijman et al. studied the ionization behavior of each of the polyphosphoinositides in MLVs composed of PC along with the phosphoinositide of interest (Kooijman, King et al. 2009). The chemical shift of each phosphate group was monitored over a range of pH values from 4-10 to determine the charge. In this pH range the second proton from each phosphate is removed to achieve fully deprotonated phosphate groups. Since the 31P

NMR chemical shift for each phosphate can be observed distinctly from the others, the single deprotonation event for that phosphate can be observed clearly, and was expected to follow a simple Henderson-Hasselbalch relation. However, Kooijman et al. found that this was not the case for phosphoinositides with adjacent phosphate groups. In these cases, the phosphate group ionization showed a complex biphasic behavior. In this study I will investigate this complex ionization further and attempt to model it.

The Fitting Model

In Kooijman et al, the ionization of many of the phosphoinositides was found to have an

‘inflection point’ in the center of the curve, around pH 7. For PI(3,4,5)P3 the effect is strikingly strong, with the 4-phosphate chemical shift actually decreasing over several pH units before increasing again (see Figure 5 in (Kooijman, King et al. 2009)). Similarly complex ionization behavior was also observed for the corresponding inositol bisphosphates (Schmitt, Bortmann et al. 1993; Schlewer, Guedat et al. 1998). While this biphasic behavior could be modeled simply with a two pKa Henderson-Hasselbalch relation, this would not correspond to reality, as each phosphate only undergoes a single deprotonation event. In order to successfully model this behavior we must first understand the exact nature of it. When two phosphate groups are adjacent, these phosphates are not completely independent, but can interact with each other

55 through mutual hydrogen bond formation. Once one phosphate is deprotonated, this phosphate has a strong incentive to form hydrogen bonds to stabilize its charged state. In the case of adjacent phosphate groups, it can form a hydrogen bond with its protonated partner, and the resulting proton is then effectively shared between the two phosphates. This results in a more strongly attached proton after the adjacent phosphate is deprotonated. Thus, the phosphates are affected by the protonation state of the adjacent phosphate groups. This sharing of protons and complex ionization behavior was also observed in studies of inositol polyphosphates (Schmitt,

Bortmann et al. 1993; Schlewer, Guedat et al. 1998).

Fitting of phosphatidylinositol bisphosphate ionization behavior: Theory

In order to describe this complex ionization, I will use a separate pKa value to describe the deprotonation of a phosphate group when its partner(s) are protonated versus when they are deprotonated, in a similar manner as described by Schlewer et al. for the inositol polyphosphates

(Schlewer, Guedat et al. 1998). Thus, for a phosphatidylinositol bisphosphate we will have four pKa values, one for each phosphate when the other is protonated (pKa1 and pKa2), and one for each phosphate when the other is deprotonated (pKa4 and pKa5). In addition we can describe the proton sharing between the phosphates by a fifth equilibrium constant (K3). The ionization of

PI(4,5)P2 can thus be represented as shown in Figure 3-1.

In order to fit this theoretical model to the experimental data, we must first determine how the chemical shift of the phosphate group relates to its current ionization state. In previous studies the chemical shift at a specific pH was represented as a weighted average of the chemical shifts of the protonated (A) and deprotonated (B) states, as follows (Kooijman, Carter et al. 2005):

훿 [퐴]+훿 [퐵] 훿 = 퐴 퐵 , (3-1) [퐴]+[퐵]

56 where δ is the chemical shift at a specific pH, δA is the chemical shift of the protonated state, δB is the chemical shift of the deprotonated state, [A] is the concentration of the protonated form, and [B] is the concentration of the deprotonated form. Thus, each ionization state is attributed its own chemical shift and the chemical shift at a given pH is evaluated as a combination of the chemical shifts of each species present at that pH. For our model, we have four separate ionization states which must each be assigned its own chemical shift value. These chemical shift values will be labeled δ0 (fully protonated state), δ4 (deprotonated 4-phosphate state), δ5

(deprotonated 5-phosphate), and δ45 (fully deprotonated state), for our example case of PI(4,5)P2.

Figure 3-1. The ionization model for PI(4,5)P2. The model describes four distinct states for the

ionization of PI(4,5)P2: the initial protonated state at pH 4 where each phosphate group carries one charge, a state with the 4-phosphate deprotonated while the 5-phosphate remains protonated, a state with the 5-phosphate deprotonated and the 4-phosphate protonated, and the final state where both

phosphates are fully deprotonated (at pH 10). pKa1 describes the deprotonation event from state 1 to

state 2. pKa2 describes the deprotonation event from state 1 to state 3. pKa values 4 and 5 describe the

deprotonation events from states 2 and 3 respectively to the final deprotonated state. K3 describes the exchange of the proton between the two singly deprotonated states.

Here I will make a basic assumption: the chemical shift for a given state will be primarily affected by the charge state of that phosphate, and will only be minimally affected by the charge state of any adjacent phosphate. Thus, for the 4-phosphate, I assume that both δ0 and δ5 (that is, the state in which the 5-phosphate is deprotonated) will be equal to the observed chemical shift for the protonated 4-phosphate (δ4P, that is, the chemical shift observed for the 4-phosphate at pH

4). The chemical shifts δ4 and δ45, which both represent states in which the 4-phosphate is deprotonated, will be assumed to be equal to the observed chemical shift for the 4-phosphate when it is deprotonated (δ4d, that is, the chemical shift observed for the 4-phosphate at pH 10).

When fitting the 5-phosphate, the situation will of course be reversed, with δ0 and δ4 being equal to δ5P (the chemical shift observed for the 5-phosphate at pH 4), along with δ5 and δ45 being equal to δ5d (the chemical shift observed for the 5-phosphate at pH 10). The overall chemical shift at a given pH can be represented as follows:

훿 [0]+훿 [4]+훿 [5]+훿 [45] 훿 = 0 4 5 45 , (3-2) [0]+[4]+[5]+[45] where the chemical shifts are as described above and [0], [4], [5], and [45] are the concentrations of the respective ionization states (the number represents the deptrotonated phosphate in each state).To determine the pKa values that describe the ionization, we simply need to replace the concentrations in equation 3-2 based on the chemical equation for each equilibrium constant.

Each chemical equation provides us with a relation between the concentration ratio and the pKa:

[4] = 10푝퐻−푝퐾푎1, (3-3) [0]

[5] = 10푝퐻−푝퐾푎2, (3-4) [0]

[5] = 퐾 , (3-5) [4] 3

[45] = 10푝퐻−푝퐾푎4 , (3-6) [4]

[45] = 10푝퐻−푝퐾푎5 , (3-7) [5]

Now, by simply dividing the numerator and denominator of equation 3-2 by [0] we can substitute the concentration ratios to obtain an equation relating the chemical shift to the pKa values describing the ionization behavior:

훿 +훿 10(푝퐻−푝퐾푎1)+훿 10(푝퐻−푝퐾푎2)+훿 10(2푝퐻−푝퐾푎1−푝퐾푎4) 훿 = 0 4 5 45 , (3-8) 1+10(푝퐻−푝퐾푎1)+10(푝퐻−푝퐾푎2)+10(2푝퐻−푝퐾푎1−푝퐾푎4)

Equation 3-8 now gives us a relation between our two variables, pH and chemical shift, which we can use to fit our chemical shift data. In Eqn. 3-8 we have seven parameters which will be determined based on our fitting. Note that K3 and pKa5 are not in Eqn. 3-8 and will not be determined as part of the fit parameters. These values are related to the other pKa values by the following relations:

푝퐾푎1−푝퐾푎2 퐾3 = 10 , (3-9)

푝퐾푎5 = 푝퐾푎4 − 푝퐾푎2 + 푝퐾푎1, (3-10)

These two pKa values can be calculated from the other pKa values. Alternatively, equation 3-8 can be rearranged to include pKa5 and refit. The fit parameters are unchanged despite the rearrangement. Thus, from a single fit we can determine all of the parameters necessary to describe the ionization of the phosphatidylinositol bisphosphate.

The fitting was carried out using MATLAB R2014a student version. The m-files for the fitting function are supplied in the appendix. As noted in equation 3-8, we have seven parameters relating our independent variable, pH, to our dependent variable, δ. However, in this case we actually have two related sets of chemical shift data, one from the 4-phosphate and one from the

5-phosphate. While these two sets of data will have differing δ0, δ4, δ5, and δ45 values, they should have the same pKa values. Therefore I fit both sets of data simultaneously to provide the best pKa values to fit both curves. In order to account for the varying chemical shift parameters, I wrapped these eight parameters (one chemical shift value for each ionization state and each phosphate) into four parameters—δ4p, δ4d, δ5p, and δ5d. The two sets of chemical shift parameters

(δ0, δ4, δ5, and δ45) were then assigned to these parameters based on their protonation state. For the 4-phosphate, δ0 and δ5 were set to δ4p, while for the 5-phosphate δ0 and δ4 were set to δ5p (see above). Non-linear fitting was performed to minimize the sum of the chi squared values from the two sets of data. Each data point was equally weighted, as the experimental error was determined to be consistent across the pH range. The experimental error was determined directly by measuring the chemical shift of multiple samples at the same pH value. The error was determined to be ± 0.06 or less. The uncertainty in the resulting parameters was determined by varying each parameter and monitoring the chi squared value. Once the chi squared value changed by 1 (one sigma uncertainty), the change in the parameter was recorded as the uncertainty in that parameter. To expedite the fitting and prevent false results, initial parameters were set to reasonable values, and an upper and lower bound was placed on the parameters. δ4p,

δ4d, δ5p, and δ5d initial parameters were set to the minimum and maximum chemical shift value over the titration curve for the protonated and deprotonated chemical shift respectively. I assume that at pH 4 each phosphate is singly protonated, and at pH 10 the phosphate is nearly completely

60 deprotonated. These values are then bound to allow them to vary within a limited range of ± 0.1 to account for some variation (due to uncertainty in the chemical shift data), but to prevent deviation from the observed values. Initial pKa values were set to either 6 or 7, depending on whether it involves the first or the second proton disassociation. The pKa values were allowed to vary between 3 and 10, as values outside of this range would not be consistent with the observed ionization behavior.

Fitting of phosphatidylinositol trisphosphate ionization behavior: Theory

In order to fit the phosphatidylinositol trisphosphate, PI(3,4,5)P3, I will require a somewhat more complex model. I will use the same reasoning as before, using eight distinct ionization states to describe the ionization behavior—one fully protonated state, three singly deprotonated states, three doubly deprotonated states, and one fully deprotonated state. Each ionization state will be connected with a separate pKa to describe each deprotonation event. The model is illustrated below in Figure 3-2.

The fitting proceeds from this model in much the same way as for the phosphatidylinositol bisphosphates. The chemical shift is again expressed as a weighted average of the chemical shift for each ionization state.

훿 [0]+훿 [3]+훿 [4]+훿 [5]+훿 [34]+훿 [35]+훿 [45]+훿 [345] 훿 = 0 3 4 5 34 35 45 345 , (3-11) [0]+[3]+[4]+[5]+[34]+[35]+[45]+[345]

Here, as for the phosphatidylinositol bisphosphates, I have multiplied the chemical shift associated with each ionization state by the concentration of that state (the number represents the

Figure 3-2. The ionization model for PI(3,4,5)P3. The model describes eight distinct states for the ionization of PI(3,4,5)P3: the initial protonated state at pH 4 where each phosphate group carries one charge, three singly deprotonated states with the 3, 4, and 5-phosphate deprotonated, three doubly deprotonated states with the 3,4, the 3,5, and the 4,5-phosphates deprotonated, and the final state where all phosphates are fully deprotonated (at pH 11). pKa1, pKa2, and pKa3 describe the deprotonation events from the protonated to singly deprotonated states. pKa7, pKa8, pKa9, pKa10, pKa11, and pKa12 describe the deprotonation events from the singly deprotonated states to the doubly deprotonated states. pKa16, pKa17, and pKa18 describe the deprotonation events from the doubly deprotonated states to the fully deprotonated state. K4, K5, and K6, describe the exchange between singly deprotonated states. K13, K14, and K15 describe the exchange between doubly deprotonated states.

62 deprotonated phosphate group(s) in each state). The concentration ratios are then related to the pKa values via the chemical equations:

[3] [34] = 10푝퐻−푝퐾푎1, (3-12) = 10푝퐻−푝퐾푎9 , (3-20) [0] [4]

[4] [45] = 10푝퐻−푝퐾푎2, (3-13) = 10푝퐻−푝퐾푎10, (3-21) [0] [4]

[5] [35] = 10푝퐻−푝퐾푎3, (3-14) = 10푝퐻−푝퐾푎11, (3-22) [0] [5]

[4] [45] = 퐾 , (3-15) = 10푝퐻−푝퐾푎12, (3-23) [3] 4 [5]

[5] [35] = 퐾 , (3-16) = 퐾 , (3-24) [3] 5 [34] 13

[5] [45] = 퐾 , (3-17) = 퐾 , (3-25) [4] 6 [34] 14

[34] 푝퐻−푝퐾 [45] = 10 푎7 , (3-18) = 퐾15, (3-26) [3] [35]

[345] 푝퐻−푝퐾 [35] 푝퐻−푝퐾 = 10 푎16 , (3-27) = 10 푎8 , (3-19) [34] [3]

[345] = 10푝퐻−푝퐾푎17 , (3-28) [35]

[345] = 10푝퐻−푝퐾푎18 , (3-29) [45]

Now, we again divide the numerator and denominator of equation 3-11 by [0] and substitute for the pKa values to obtain a rather complex equation:

(푝퐻−푝퐾 ) (푝퐻−푝퐾 ) (푝퐻−푝퐾 ) 훿0+훿310 푎1 +훿410 푎2 +훿510 푎3 (2푝퐻−푝퐾 −푝퐾 ) (2푝퐻−푝퐾 −푝퐾 ) (2푝퐻−푝퐾 −푝퐾 ) +훿3410 푎1 푎7 +훿3510 푎1 푎8 +훿4510 푎2 푎10 +훿 10(3푝퐻−푝퐾푎1−푝퐾푎7−푝퐾푎16) 훿 = 345 , (3-30) 1+10(푝퐻−푝퐾푎1)+10(푝퐻−푝퐾푎2)+10(푝퐻−푝퐾푎3) +10(2푝퐻−푝퐾푎1−푝퐾푎7)+10(2푝퐻−푝퐾푎1−푝퐾푎8)+10(2푝퐻−푝퐾푎2−푝퐾푎10) +10(3푝퐻−푝퐾푎1−푝퐾푎7−푝퐾푎16)

Equation 3-30 is used as the fit equation between our pH and chemical shift variables. Eqn. 3-30 has 15 unique parameters relating pH to chemical shift. Eleven pKa values are not used in the equation and are calculated post-fit using the best parameters:

푝퐾푎2−푝퐾푎1 푝퐾푎7−푝퐾푎8 퐾4 = 10 , (3-31) 퐾13 = 10 , (3-37)

푝퐾푎1−푝퐾푎3 푝퐾푎9−푝퐾푎10 퐾5 = 10 , (3-32) 퐾14 = 10 , (3-38)

푝퐾푎2−푝퐾푎3 푝퐾푎11−푝퐾푎12 퐾6 = 10 , (3-33) 퐾15 = 10 , (3-39)

푝퐾푎9 = 푝퐾푎7 − 푝퐾푎2 + 푝퐾푎1, (3-34) 푝퐾푎5 = 푝퐾푎16 − 푝퐾푎8 + 푝퐾푎7, (3-40)

푝퐾푎11 = 푝퐾푎8 − 푝퐾푎3 + 푝퐾푎1, (3-35) 푝퐾푎5 = 푝퐾푎16 − 푝퐾푎10 + 푝퐾푎9, (3-41)

푝퐾푎12 = 푝퐾푎10 − 푝퐾푎3 + 푝퐾푎2 , (3-36)

As with the phosphatidylinositol bisphosphate fitting, eqn. 3-30 could also be rearranged to place these other pKa values in the fitting equation. The fitting was performed using MATLAB. The m-files for the fitting function are supplied in the appendix. As with the phosphatidylinositol

bisphosphate fitting, we can use a single fitting function to fit each of the phosphates. Each phosphate will again have differing chemical shift values, and thus the total number of fit parameters is 15 (7 pKa values and 8 chemical shift parameters for each phosphate). I will again cut down the number of parameters by wrapping the chemical shift parameters into six parameters—δ3p, δ3d, δ4p, δ4d, δ5p, and δ5d. The chemical shift parameters for each phosphate are assigned based on their protonation state, e.g., for the 3-phosphate, δ0, δ4, δ5, and δ45 were set to

δ3p. Non-linear fitting was done to minimize the sum of the chi squared values from each data set. Once again the uncertainty was determined within one standard deviation by examining the variance for each parameter that results in a one unit variation in chi square. The initial values for the chemical shift parameters were set based on the minimum and maximum chemical shift values over their respective titration curve and held to ± 0.1 of their original values. Initial pKa values were set to 6, 7, or 8, depending on whether it produces a singly deprotonated, doubly deprotonated, or fully deprotonated state. The pKa values were allowed to vary between 3 and 10 to provide values consistent with the observational data.

Fitting of PI(3,4)P2 ionization behavior: Results

The phosphatidylinositol bisphosphate PI(3,4)P2 has two adjacent phosphates, and the titration curve has a bimodal behavior that cannot be fit with a Henderson-Hasselbalch relation.

The new four-state model fits the data quite well. The complete fit is shown in Figure 3-3A against the original chemical shift data. The fitted pKa values and fitting model for PI(3,4)P2 is shown in Fig. 3-3B and C. Based on the fit pKa values the charge for each phosphate and relative fraction of each ionization state can be calculated. The results are shown in Fig. 3-3D and E. za

A B

D E

Figure 3-3. Fitting results for PI(3,4)P2. A) pH titration curve for PI(3,4)P2 showing chemical shift versus pH. Data from Kooijman et al. (Kooijman, King et al. 2009). Solid lines show the best fitting

from the phosphatidylinositol bisphosphate fitting model. B) Table of pKa values and charge at pH 7

based on the fit results. C) Model scheme showing the pKa values from the fit next to each deprotonation event. D) A plot of the charge versus the pH for each phosphate. E) Plot showing the

relative prevalence of each ionization state. f0 is the protonated fraction, f3 is the fraction with P3

deprotonated, f4 is the fraction with P4 deprotonated, and f34 is the fully deprotonated fraction.

The pKa values for the initial deprotonation event are nearly equivalent, with no significant difference between pKa1 and pKa2. The charge and relative fraction calculations suggest that the

P3 phosphate carries a slightly higher charge, but the difference is quite small and insignificant.

The two phosphates seem to be roughly equivalent, with no significant differences in their ionization. The similarity between these two phosphates is somewhat surprising when we consider that the 3-phosphate is adjacent to an axial hydroxyl group while the 4-phosphate is adjacent to an equatorial hydroxyl group. Apparently the positioning of this hydroxyl group has very little impact on the adjacent phosphate group’s ionization. As suggested by the bimodal shape of the curve, both phosphates are much harder to deprotonate after one proton has disassociated (pKa of 7.7, versus 6.4 for the first deprotonation). The last remaining proton is stabilized by exchange between the two phosphates. The exchange rate between the phosphates is close to one, indicating an equal sharing of the final proton between the two phosphates. The fitting results are completely consistent with the observations of Kooijman et al., within the uncertainty of the fit (Kooijman, King et al. 2009).

Fitting of PI(4,5)P2 ionization behavior: Results

The chemical structure of PI(4,5)P2 is quite similar to PI(3,4)P2, with two adjacent phosphates on the end of the inositol ring. Unlike PI(3,4)P2, these phosphates both have an equatorial hydroxyl group adjacent to them, although simulations have suggested that the 5-phosphate may actually interact with the axial hydroxyl in the second position (Slochower, Huwe et al. 2013).

The titration curve shows a complex bimodal behavior that fits quite well with the four-state model. The complete fit is shown in Figure 3-4A against the original chemical shift data. The fitted pKa values and fitting model for PI(4,5)P2 is shown in Fig. 3-4B and C, while the charge and relative fractions of the ionization states are calculated in Fig. 3-4D and E. The 4-phosphate shows a significantly higher preference for deprotonation, with its initial deprotonation event 0.2 units lower than the 5-phosphate’s pKa. When the 5-phosphate is deprotonated, the pKa for the 4- phosphate is still lower than the comparable pKa for the 5-phosphate (pKa5 vs. pKa4). This shows that the 4-phosphate can more readily carry charge as compared to the 5-phosphate and both phosphates for PI(3,4)P2. The higher charge on the 4-phosphate gives PI(4,5)P2 a slightly higher charge as compared to PI(3,4)P2 as well. The lower pKa values for the second deprotonation event means that PI(4,5)P2 will have an even greater charge than PI(3,4)P2 at pH values around

8. The exchange rate of 0.65 shows the strong preference for the proton to reside on the 5- phosphate as opposed to the 4-phosphate. The overall charge (including the phosphodiester) calculated from this fit model (-4.04) is slightly higher than the charge reported by Kooijman et al (-3.99), but still within the limits of the uncertainty (Kooijman, King et al. 2009).

A B

E D

Figure 3-4. Fitting results for PI(4,5)P2. A) pH titration curve for PI(4,5)P2 showing chemical shift versus pH. Data from Kooijman et al. (Kooijman, King et al. 2009). Solid lines show the best fitting

from the phosphatidylinositol bisphosphate fitting model. B) Table of pKa values and charge at pH 7

based on the fit results. C) Model scheme showing the pKa values from the fit next to each deprotonation event. D) A plot of the charge versus the pH for each phosphate. E) Plot showing the

relative prevalence of each ionization state. f0 is the protonated fraction, f4 is the fraction with P4

deprotonated, f5 is the fraction with P5 deprotonated, and f45 is the fully deprotonated fraction.

Fitting of PI(3,5)P2 ionization behavior: Results

Unlike PI(4,5)P2 and PI(3,4)P2, the phosphates in PI(3,5)P2 are not adjacent, and the titration curves for the 3- and 5-phosphates can be fit independently using a Henderson-Hasselbalch relation. However, a Henderson-Hasselbalch fit assumes that these two ionization steps are completely independent. Even though the phosphates are not adjacent, they may still influence each other via their effect on the overall hydrogen bond network of the inositol ring, particularly through the 4’-OH group, which could form hydrogen bonds with either phosphate. When one phosphate is deprotonated, this will remove a hydrogen-bond donor from the ring, potentially breaking hydrogen bonds and causing new bonds to form, which could potentially stabilize the proton on the other phosphate. Therefore, I carried out the fitting on PI(3,5)P2 using the new model. The complete fit is shown in Figure 3-5A against the original chemical shift data. The fitted pKa values and fitting model are shown in Figure 3-5B and C. Figure 3-5D and E show the calculated charge and relative fractions of the ionization states. The complex model fits the chemical shift data quite well. The pKa values for the second deprotonation event are significantly different from the pKa values for the first deprotonation, which suggests that the phosphate groups are indeed affected by the ionization of the other phosphate. The difference is

A B

D E

Figure 3-5. Fitting results for PI(3,5)P2. A) pH titration curve for PI(3,5)P2 showing chemical shift versus pH. Data from Kooijman et al. (Kooijman, King et al. 2009). Solid lines show the best fitting

from the phosphatidylinositol bisphosphate fitting model. B) Table of pKa values and charge at pH 7

based on the fit results. C) Model scheme showing the pKa values from the fit next to each deprotonation event. D) A plot of the charge versus the pH for each phosphate. E) Plot showing the

relative prevalence of each ionization state. f0 is the protonated fraction, f3 is the fraction with P3

deprotonated, f5 is the fraction with P5 deprotonated, and f35 is the fully deprotonated fraction.

much smaller than for the other phosphatidylinositol bisphosphates as the phosphates cannot directly share the final proton and only indirectly affect each other. The 5-phosphate has two adjacent equatorial hydroxyl groups which results in a well stabilized deprotonated form, compared to the 3-phosphate which has adjacent axial and equatorial hydroxyl groups

(Kooijman, King et al. 2009). Thus the 5-phosphate carries a much higher charge. The exchange constant here represents the greater preference for the 5-phosphate to be more highly charged than the 3-phosphate and no direct exchange is likely to take place. Overall, PI(3,5)P2 carries a somewhat higher charge than the other phosphatidylinositol bisphosphates. The low pKa values for the second deprotonation event result in a higher fraction of fully deprotonated PI(3,5)P2 at lower pH values, leading to the higher charge. The separation between the two phosphate groups makes it much easier to fully deprotonate the phosphate groups as the last proton is not shared between the phosphates and there is less repulsion between the negative charges due to the increased separation.

The charge values measured with the new fit model agree within the margin of error with previous results. The pKa values based on a Henderson-Hasselbalch fit are 6.96 and 6.58 for the

3- and 5-phosphates respectively (Kooijman, King et al. 2009). The pKa values from this model

differ from the previous values, as the values from this model account for interaction between the phosphates (through the mutually shared 4’-OH). The average of the two pKa values for each phosphate (e.g. average of pKa1 and pKa5 for the 3-phosphate) is roughly equivalent to the pKa based on the Henderson-Hasselbalch fit (6.90 vs. 6.96, and 6.64 vs. 6.58 for the 3- and 5- phosphates respectively).

Fitting of PI(3,4,5)P3 ionization behavior: Results

PI(3,4,5)P3 has three phosphates at positions 3, 4, and 5 on the inositol ring. This makes it unique from the phosphatidylinositol bisphosphates. PI(3,4,5)P3 has two phosphates (P3 and P5) which are adjacent to a hydroxyl group and a phosphate group. The final phosphate, P4, is adjacent to two other phosphates. This results in some rather unusual ionization behavior. Figure

3-6A shows the chemical shift data and the fitting curves that were calculated using the model for phosphatidylinositol trisphosphate. The fitted pKa values and fitting model for PI(4,5)P2 is shown in Figure 3-4B and C, while the charge and relative fractions of the ionization states are calculated in Figure 3-4D and E. PI(3,4,5)P3 has three sets of pKa values, one for each deprotonation event. The first proton comes off at roughly pH 6.5 (pKa1, pKa2, pKa3). The second proton pKa lies between 7.08 and 8.19. The final deprotonation event has a pKa of 9.14 to 9.83.

For the first deprotonation, the 4-phosphate is deprotonated much more strongly than the other phosphates, with a low pKa of 6.10. The exchange constants show the preference for the deprotonated 4-phosphate, the transfer from deprotonated 3-phosphate to 4-phosphate has a constant of 2.42 (K4), and the transfer from the 4-phosphate to 5-phosphate has a constant of

0.38 (K6). The fraction of deprotonated 4-phosphate is more than twice the fraction of the other

A B

D E

Figure 3-6. Fitting results for PI(3,4,5)P3. A) pH titration curve for PI(3,4,5)P3 showing chemical shift versus pH. Data from Kooijman et al. (Kooijman, King et al. 2009). Solid lines show the best

fitting from the phosphatidylinositol trisphosphate fitting model. B) Table of pKa values and charge at

pH 7 based on the fit results. C) Model scheme showing the pKa values from the fit next to each deprotonation event. D) A plot of the charge versus the pH for each phosphate. E) Plot showing the

relative prevalence of each ionization state. f0 is the protonated fraction, f3 is the fraction with P3

deprotonated, f4 is the fraction with P4 deprotonated, f5 is the fraction with P5 deprotonated, f34 is the

fraction with P3 and P4 deprotonated, f35 is the fraction with P3 and P5 deprotonated, f45 is the fraction

with P4 and P5 deprotonated, and f345 is the fully deprotonated fraction.

deprotonated phosphates. The 4-phosphate can hydrogen bond with both of the adjacent phosphates which can themselves be stabilized by the hydroxyl groups next to them. This makes the 4-phosphate the preferred location for the first deprotonation event. In the second deprotonation step, pKa8 and pKa11 are the lowest, both leading to the deprotonated 3- and 5- phosphates. The 3- and 5-phosphate deprotonated state is much more favored than the other ionization states, with exchange constants to the 3,5- deprotonated state of 4.89 (K13) and 3.85

(1/K15). The 3,5-phosphate deprotonated state is almost four times as prevalent as the other doubly deprotonated states. The 3- and 5-phosphate deprotonated state allows the charge to be spread out, with opportunities to hydrogen bond with the 2’-OH, the 4-phosphate, and the 6’-OH.

Thus, this deprotonation state is much more stable than the others, which would force the charges to be clumped together and have less opportunity for hydrogen bonding. The final deprotonation step is the most difficult one with pKa values above 9. From the stable 3,5- deprotonated state the pKa is as high as 9.83. The fully deprotonated state carries a high charge and the 4-phosphosphate has no nearby hydrogen bond donors to help stabilize its negative charge. At pH 7, the overall charge (including the phosphodiester) is -5.14, quite a bit higher than the phosphatidylinositol bisphosphates, although the individual charge on the phosphates is

actually lower than for the bisphosphates (charge is reduced due to the proximity of multiple charged groups). At this pH the deprotonated 4-phosphate is the most prominent form, although the 3,5-deprotonated state increases rapidly from this point as the pH increases. The charge is similar to the calculations by Kooijman et al (-5.05). However, according to this model, the charge on the 4-phosphate is higher, leading to a slightly higher overall charge.

Discussion

The proposed fitting model appears to be very effective at modeling the ionization data for these polyphosphoinositides. Using the fitting, the charge and the relative fraction of each ionization state was determined for each pH value, as well as pKa values governing the conversion between the various states. This gives us insight into the ionization behavior, as well as giving us quantitative values for calculations involving the ionization of these important signaling lipids. The results of this ionization model illustrate the importance of intramolecular interactions to the ionization of these lipids. The formation of intramolecular hydrogen-bonds can stabilize a charged group, while the presence of another charged group within the molecule can increase the resistance to further deprotonation. In addition to these intramolecular effects,

Kooijman et al. suggested that intermolecular interactions occur between phosphoinositides in the bilayer environment, leading to a higher charge at the membrane than in solution (Kooijman,

King et al. 2009). This conclusion was based on a direct comparison of the charge (at pH 7) for

PI(4,5)P2 in the membrane versus Ins(4,5)P2 in solution (-2.99 for the phosphomonoesters of

PI(4,5)P2 vs. -2.81 for Ins(4,5)P2 (Schlewer, Guedat et al. 1998)). Since we have calculated pKa values to describe the ionization of PI(4,5)P2 within the membrane, we can now compare these

directly with the pKa values of Ins(4,5)P2 Table 3-1. Comparison of pKa values for PI(4,5)P2 and Ins(4,5)P2. to determine the relative impact of these a pKa Ins(4,5)P2 PI(4,5)P2 Deprotonation - intermolecular interactions. The pKa1 5.82 ± 0.05 6.25 ± 0.04 P4,P5 → P4 ,P5 - pKa2 6.09 ± 0.05 6.44 ± 0.05 P4,P5 → P4,P5 - - - Ins(4,5)P2 values from Schmitt et al. are pKa4 8.15 ± 0.05 7.60 ± 0.04 P4 ,P5 → P4 ,P5 - - - pKa5 8.01 ± 0.07 7.41 ± 0.04 P4,P5 → P4 ,P5 a compared against the values we Values for Ins(4,5)P2 are from Schmitt et al. (Schmitt, Bortmann et al. 1993). determined for PI(4,5)P2 in table 3-1

(Schmitt, Bortmann et al. 1993). The initial deprotonation event for Ins(4,5)P2 is

actually more favorable than for PI(4,5)P2 (compare pKa1 and pKa2 for the two compounds). This is likely due to the low dielectric constant and negative surface charge at the membrane, which make it more difficult to deprotonate groups at the membrane. This would seem to indicate a higher charge for Ins(4,5)P2 than for PI(4,5)P2. However, the second deprotonation step is much more favorable for PI(4,5)P2 than for Ins(4,5)P2 (compare pKa4 and pKa5 for the two compounds). Because of these low pKa values for the second deprotonation step, the fully deprotonated fraction is already present at pH 7 and the charge is higher for PI(4,5)P2 then for

Ins(4,5)P2. Why would the fully deprotonated state PI(4,5)P2 be more favorable than the fully deprotonated state of Ins(4,5)P2? Most likely this is due to intermolecular interactions, as suggested by Kooijman et al. (Kooijman, King et al. 2009). When Ins(4,5)P2 is fully deprotonated, the charged phosphates can only be stabilized by interacting with the hydroxyl groups on the inositol ring. For PI(4,5)P2 in the membrane, the fully deprotonated PI(4,5)P2 molecules can be additionally stabilized by hydrogen bond formation with other PI(4,5)P2 molecules within the membrane. These reduced pKa values for PI(4,5)P2 within the membrane illustrate the strength and importance of intermolecular interactions between phosphoinositides. 77

Chapter 4*

Phosphatidylinositol-4,5-bisphosphate ionization and domain formation in the presence of

lipids with hydrogen bond donor capabilities

Introduction

In a previous study (Kooijman, King et al. 2009), we found that the ionization states of phosphatidylinositol bisphosphates and phosphatidylinositol-3,4,5-trisphosphate are strongly affected by intra- and intermolecular hydrogen bond formation. We have found that hydroxyl groups vicinal to the respective phosphomonoester group engage in hydrogen bond formation that leads to an increased deprotonation (higher charge) of the respective group. For PI(4,5)P2 and phosphatidylinositol-3,4-bisphosphate (PI(3,4)P2) we found that for pH >7 the last remaining proton is shared between the two phosphate groups. In addition to intramolecular hydrogen bond formation between the hydroxyl and phosphomonoester groups as well as between vicinal phosphomonoester groups, we found strong evidence for intermolecular hydrogen bond formation between neighboring phosphoinositide molecules. We rationalized this surprising behavior with the formation of an intricate inter- and intramolecular hydrogen bond network that leads to a dissipation of the negative charge of the phosphoinositide headgroup and therefore, reduced repulsive forces.

* This chapter is adapted from Graber, Z. T., Z. Jiang, et al. (2012). "Phosphatidylinositol-4,5-bisphosphate ionization and domain formation in the presence of lipids with hydrogen bond donor capabilities." Chem Phys Lipids 165(6): 696-704. 78

In the work described above, we investigated the ionization behavior of PI(4,5)P2 in mixed vesicles with phosphatidylcholine (PC). PC functions in this case as a “matrix lipid” that allows for the fabrication of vesicles (PI(4,5)P2 by itself does not form stable vesicles). It is assumed that the zwitterionic PC lipid has little or no effect on the ionization properties of PI(4,5)P2.

This study expands on our earlier work by examining the effect of inner leaflet membrane lipids on the ionization behavior of PI(4,5)P2 (see Figure 4-1 for lipid structures).

Phosphatidylethanolamine (PE) is a zwitterionic phospholipid that is expected to engage in hydrogen bond formation with PI(4,5)P2, leading to enhanced deprotonation of the phosphomonoester groups as it was observed for phosphatidic acid (PA) (Kooijman, Carter et al.

2005). Phosphatidylserine (PS) is, at pH 7, an anionic lipid that can also engage to some extent in hydrogen bond formation. As a result, two opposing effects might act on the PI(4,5)P2 headgroup: The negative charge of the PS headgroup will lead to a reduction of the interfacial pH and hence an increased protonation of the PI(4,5)P2 headgroup. If hydrogen bond formation occurs between PS and PI(4,5)P2, this is expected to lead to an increased deprotonation of the

PI(4,5)P2 phosphomonoester groups. The overall PI(4,5)P2 charge will be determined by the balance between these two opposing effects. This holds even more for phosphatidylinositol (PI), which is also an anionic lipid that is capable of engaging in hydrogen bond formation with

PI(4,5)P2. What sets PI apart from PS is the richer hydrogen bond capability of the PI headgroup in comparison to PS. For either of these anionic lipids, hydrogen bond formation will only occur if the attractive PS/PI(4,5)P2 or PI/PI(4,5)P2 force due to hydrogen bond formation is stronger than the repulsive force due to negative charges on the headgroups.

PI(4,5)P PE PI PS 2

Figure 4-1: Chemical structures of the lipids. From left to right: Phosphatidylinositol-4,5-bisphosphate, phosphatidylethanolamine, phosphatidylinositol, and phosphatidylserine.

Results

In our previous study we found that the ionization state of the PI(4,5)P2 phosphomonoester groups in mixed PI(4,5)P2/PC vesicles is strongly affected by intermolecular hydrogen bond formation between neighboring PI(4,5)P2 molecules and intramolecular hydrogen bonds between the two phosphomonoester groups as well as the hydroxyl and phosphomonoester groups

(Kooijman, King et al. 2009). Phosphatidylethanolamine (PE) is a zwitterionic lipid that in contrast to phosphatidylcholine (PC) is able to function as a hydrogen bond donor and the question arises to what extent PE, which is found in high concentrations in the inner leaflet of the plasma membrane, affects PI(4,5)P2 ionization state.

PE promotes increased deprotonation of PI(4,5)P2

To investigate the ionization state of PI(4,5)P2 in the presence of PE, we conducted pH dependent solid state MAS 31P-NMR experiments of mixed multilamellar vesicles composed of brain PI(4,5)P2 (5%)/DOPE (47.5%)/DOPC (47.5%) (Figure 4-2A). The strong peak at -0.029 ppm is a superposition of the peaks associated with the PE and PI(4,5)P2 phosphodiester groups, while the peak found at -0.616 ppm is associated with the phosphodiester group of the PC component (note the greater height for the PE peak is due to overlap with the PI(4,5)P2 phosphodiester as well as the asymmetric tail of the PC peak, see Figure A1 in the appendix).

The small peak furthest downfield has been linked to the phosphomonoester group at the 4- position of the inositol ring, while the peak slightly upfield from that peak is due to the phosphomonoester group in the 5-position (Kooijman, King et al. 2009). Static spectra confirming a bilayer configuration are shown in Figure A2 in the appendix. Upon increasing the pH, the two peaks associated with the PI(4,5)P2 phosphomonoester groups shift downfield, while the positions of the peaks linked to the phosphodiester groups remain unchanged within this pH interval (which is expected since the protonation state of the phosphodiester groups does not change for the investigated pH range). The shift of the phosphomonoester-associated peaks is due to enhanced deshielding of the respective phosphorus atom as a result of the increasing deprotonation of the phosphomonoester group as the pH is increased. Figure 4-2B shows the pH dependent chemical shift variation of the phosphomonoester groups peaks for PI(4,5)P2 in the absence and presence of PE (the data for PI(4,5)P2/PC mixed vesicles have been reproduced from Kooijman et al. (Kooijman, King et al. 2009)). Like it was found for the PC/PI(4,5)P2 vesicles, the pH dependent chemical shift variation for the PI(4,5)P2/PE/PC vesicles shows a 81

A B

D E

31 Figure 4-2: P MAS NMR spectra and pH titration curves for PC / PE / PI(4,5)P2 (47.5% / 31 47.5% / 5%). a). P MAS NMR spectra as a function of pH for 5 mol% PI(4,5)P2 in 47.5% PC / 47.5 % PE vesicles. Peak assignments were discussed in Kooijman et al. (Kooijman, King et al. 2009). b).

Peak positions of the 4- and 5-phosphate of PI(4,5)P2 as a function of pH. c). Ionization model and

pKa values for deprotonation of PI(4,5)P2 with PE. d) Charge of the 4- and 5-phosphate of PI(4,5)P2 as

a function of pH. e) Plot showing the relative prevalence of each ionization state as a function of pH. f0

is the protonated fraction, f4 is the fraction with P4 deprotonated, f5 is the fraction with P5

deprotonated, and f45 is the fully deprotonated fraction.

bimodal behavior. For pH values of about 4 – 5, both phosphomonoester groups carry one proton

(single protonated state). Upon increasing the pH, one of the two remaining protons dissociates

(first “step” in the titration curve), while the second proton is shared between the two phosphomonoester groups (Felemez, Bernard et al. 2000; Kooijman, King et al. 2009). Upon further increase of the pH, the proton shared between the two phosphomonoester groups dissociates, which is reflected in the pH titration curve as the second “step”. The chemical shift behavior was fitted using a four-state phosphatidylinositol bisphosphate ionization model (see chapter 3). The fit results are shown in Figure 4-2C. For the first ionization step (pH ~ 4.5 – 6.5), the comparison of the data obtained for the PI(4,5)P2/PE/PC vesicles with those obtained previously for PI(4,5)P2/PC vesicles reveals a downfield shift of the data for the vesicles that contain PE. Such a downfield shift indicates an increased deprotonation of the phosphomonoester groups for a given pH (lowering of the pKa), which is due to hydrogen bond formation between the PE and PI(4,5)P2 headgroups. The 4-phosphate pKa is reduced from 6.25 to 5.87, while the 5-phosphate (pKa2) is reduced from 6.44 to 6.01. Surprisingly, this is much less significant for the dissociation of the second proton (second dissociation step pH ~ 7 – 9.5), with pKa values changing from 7.60 to 7.48 for the 5-phosphate (pKa4), and from 7.41 to 7.34 for the

4-phosphate (pKa5). For the PI(4,5)P2 4-phosphate the data in the presence and absence of PE are essentially superimposable, while the data for the 5-phosphate group reveal a slight

downfield shift of the corresponding peak in comparison to the data obtained in the absence of

PE. This suggests that the interaction between the PE headgroup and the 5-phosphate is stronger than the respective interaction between the PE headgroup and the 4-phosphate group. As a result the charge for the 5-phosphate group is much closer to the corresponding charge of the 4- phosphate in the presence of PE than in its absence (Charge with PE: -1.50 vs. -1.62, charge without PE: -1.43 vs -1.60).

Overall, the effect of PE on the ionization state of the PI(4,5)P2 phosphomonoester groups is significantly less pronounced than it was observed for phosphatidic acid in the presence of PE

(Kooijman, Carter et al. 2005). In that case the pKa2 of PA dropped from 7.92 to 7.02 in the presence of PE, resulting in a charge increase at pH 7.0 from -1.11 to -1.50. In the case of the 4- and 5-phosphate these changes are as follows: 4-phosphate from -1.60 to -1.62 (which is within the error limits of the method) and 5-phosphate from -1.43 to -1.50. The charge for the PI(4,5)P2 phosphate groups are shown in figure 4-2D. It should be noted that this increase in charge is highly pH dependent as can be seen from Figure 4-2. These data thus indicate that small changes in intracellular pH (particularly to more acidic pH) will affect the charge carried by the phosphomonoester groups of PI(4,5)P2 significantly. Additionally this change in ionization will be highly dependent on the concentration of PE present in the membrane (Kooijman, Carter et al.

2005).

Phosphatidylserine has little effect on the ionization state of PI(4,5)P2 in a PC matrix

Phosphatidylserine (PS) is a major constituent of the plasma membrane inner leaflet and it is therefore important to study the effect of PS on PI(4,5)P2 ionization behavior. Due to the anionic

nature of the PS headgroup, the interfacial region becomes more negatively charged in the presence of PS, which leads to an enhanced proton concentration and hence a lower pH in the interfacial region in comparison to the bulk. This is expected to lead to an enhanced protonation of the PI(4,5)P2 phosphomonoester groups and hence the phosphomonoester group peaks are expected to shift upfield in comparison to the data obtained for the PC/PI(4,5)P2 vesicles. On the other hand, the serine group might engage in hydrogen bond formation with the PI(4,5)P2 phosphomonoester groups, which would lead to the opposite effect, i.e., enhanced deprotonation.

The hydrogen bond donor capability of PS is expected to change in the investigated pH range because of the pKa of the carboxyl and amine groups of the serine residue. The intrinsic pKa of the carboxyl group of serine is about 3.6 (Tsui, Ojcius et al. 1986) (this value is consistent with our own 31P NMR data, see supplementary data Figure A3). Below pH 3.6 the carboxyl group is protonated and therefore, PS is a zwitterionic lipid. Above this pH (in the range of 5.5

PS is an anionic lipid. Above pH 8.5 (see Figure A3) the amine group becomes deprotonated and

PS carries two negative charges.

31 We investigated by MAS P-NMR the ionization behavior of PI(4,5)P2 in mixed multilamellar PC/PS/PI(4,5)P2 vesicles (2% PI(4,5)P2, 20% PS, 78% PC). Figure A1 in the supplementary material section displays static spectra indicating that PC/PS/PI(4,5)P2 mixtures form bilayers over the entire pH range investigated. Figure 4-3A shows the corresponding MAS

31P-NMR spectra for different pH values between 4 and 10. The strong peak located at ~ - 0.75 ppm is associated with the phosphodiester group of PC, while the smaller peak downfield of the main peak is due to the phosphodiester group of PS. The PI(4,5)P2 phosphodiester peak cannot be seen because it is superimposed by the stronger PC phosphodiester peak. The pH dependent 85

A B

D E

31 Figure 4-3: P MAS NMR spectra and pH titration curves for PC / PS / PI(4,5)P2 (78% / 20% / 31 2%). a). P MAS NMR spectra as a function of pH for 2 mol% PI(4,5)P2 in 78% PC / 20% PS

vesicles.b). Peak positions of the 4- and 5-phosphate of PI(4,5)P2 as a function of pH. c). Ionization

model and pKa values for deprotonation of PI(4,5)P2 with PS. d) Charge of the 4- and 5-phosphate of

PI(4,5)P2 as a function of pH. e) Plot showing the relative prevalence of each ionization state as a

function of pH. f0 is the protonated fraction, f4 is the fraction with P4 deprotonated, f5 is the fraction

with P5 deprotonated, and f45 is the fully deprotonated fraction. chemical shift of the phosphodiester of PS is shown in figure A3 and demonstrates the pH dependent ionization of the carboxyl and amine group of PS.

The PI(4,5)P2 phosphomonoester group peaks are clearly visible despite the fact that the

PI(4,5)P2 concentration is lower than for the PC/PE/PI(4,5)P2 lipid mixture described above (we lowered the concentration to be closer to the global physiological PI(4,5)P2 plasma membrane concentration, which is about 1% of the total inner leaflet plasma membrane concentration

(McLaughlin and Murray 2005). Figure 4-3B shows the pH dependent chemical shift variation of the P-4 and P-5 phosphomonoester groups of PI(4,5)P2, and Figure 4-3C-E show the fitting model and resulting charge calculations. Surprisingly, the data for PC/PS/PI(4,5)P2 and

PC/PI(4,5)P2 are very similar. For pH values < 7, the chemical shift values (Figure 4-3B) for the

4-phosphate group are essentially superimposable for the two systems, while the 5-phosphate data show a very minor downfield shift that is close to the error limits of the method. The pKa values for this first deprotonation step are virtually indistinguishable, resulting in a total charge of the PI(4,5)P2 headgroup that is nearly the same as the charge in PC/PI(4,5)P2 vesicles. Above pH 7, the data for both phosphate groups indicate a slightly higher protonation in the presence of

PS than in its absence, resulting in slightly higher pKa values for the second deprotonation step

(increase of 0.17 for the 5-phosphate and 0.19 for the 4-phosphate). The relatively small change

of the PI(4,5)P2 ionization state in the presence of PS is surprising. To explore further the effect of PS on the ionization state of lipids with phosphomonoester groups, we compared the ionization behavior of PA in the presence and absence of PS (see Figure 4-4). In this case the two data sets are also very similar. This is in contrast to what was observed by Kooijman et al.

(Kooijman, Carter et al. 2005) who investigated the PA ionization state in PC/PE/PS mixed vesicles. In the earlier study, variation of the PS concentration resulted in changes of the PA ionization state resulting from the increased negative electrostatic potential in the interfacial region, leading to a higher protonation of PA. However, it should be noted that in this earlier study the additional PS in these mixtures replaced PE and PC within the vesicles, so this increased protonation of PA may simply have been as a result of decreased hydrogen bonding with PE. In summary, these results suggest that as the lipid mixtures become more complex, the

Figure 4-4. Ionization behavior of PA in PC/PS/PA (75% / 20% / 5%) vesicles as determined by solid state 31P NMR. The chemical shift of PA is recorded as a function of pH for PC/PS/PA (black) vesicles compared against PC/PA (95% / 5%, grey). The solid line is a fit to an equation derived from the Henderson-Hasselbalch equation as described previously. Chemical shift is recorded relative to an

external 85% H3PO4 standard.

prediction of the phosphomonoester group ionization state becomes more challenging. The absence of an apparently strong impact of PS on PI(4,5)P2 or PA ionization state is unexpected and will be discussed further below.

To explore the potential demixing of PI(4,5)P2 with PS, we investigated the morphology of giant unilamellar vesicles composed of PC/PS/PI(4,5)P2 at a 70:20:10 ratio using fluorescence microscopy (Figure 4-4). The 70:20:10 ratio was chosen due to its similarity to the ratio used for the NMR experiments, however the PI(4,5)P2 content was increased in order to allow visualization of the PI(4,5)P2 in the event of macroscopic domain formation. The GUVs do not show any macroscopic domain formation in these mixtures. This suggests that PS does not induce macroscopic PI(4,5)P2 domain formation, however, we cannot conclusively say that there are no domains, as there may be microscopic PI(4,5)P2 domains below the diffraction limit

(Redfern and Gericke 2004; Kooijman, King et al. 2009) or the rhodamine PE fluorescent probe may not be excluded from the domains.

Figure 4-5: GUVs composed of PC/PS/PI(4,5)P2. GUVs were composed of POPC (70%), POPS

(20%), and brain PI(4,5)P2 (10%). GUVs were developed in a pH 7 buffer with 5 mM HEPES, 0.1 mM EDTA and 100 mM NaCl. All images were taken by fluorescence microscopy at room temperature.

Phosphatidylinositol induces PI(4,5)P2 domain formation and differentially effects the ionization state of PI(4,5)P2

Phosphatidylinositol (PI) is found in the inner leaflet of the plasma membrane in concentrations between 6 – 10% (van Meer, Voelker et al. 2008). Like it was noted for PS, the negative charge of the phosphatidylinositol headgroup is expected to lead to an enhanced proton concentration in the interfacial region. What sets PI apart from PS is the significantly richer hydrogen bonding capability of the inositol headgroup as compared to the serine group. As a result, PI(4,5)P2 may experience two opposing effects in the presence of PI: The negative charge of the PI headgroup will attract protons to the interfacial region which is expected to result in enhanced PI(4,5)P2 protonation and hence a lower charge of the phosphomonoester groups for a given pH. In addition, PI might interact with PI(4,5)P2 via hydrogen bond formation, which would result in enhanced deprotonation of the phosphomonoester groups. In Figure 4-6A the

31 solid state P-NMR spectra for PC/PI/PI(4,5)P2 (88%/10%/2%) vesicles are shown for pH values between ~ 4 and ~10. Figure A2 (C) shows that the PC/PI/PI(4,5)P2 mixtures formed bilayers over the entire pH range investigated. As for the previously described lipid mixtures, the strong upfield peak (Figure 4-6A) centered at about -0.7 ppm is associated with the phosphodiester groups of the three lipid species, while the two small downfield peaks are due to the 4- and 5-phosphate respectively (note the shoulder on the edge of the PC peak is from the PI, see Figure A4). Figure 4-6B shows the pH dependent chemical shift variation of the P-4 and P-5 phosphomonoester groups of PI(4,5)P2. The data show the bimodal behavior observed previously. Figure 4-6C-E show the results of the phosphatidylinositol bisphosphate fitting model and calculated charge. For the dissociation of the first proton (pH < 7), the data obtained 90

A B

D E

31 Figure 4-6: P MAS NMR spectra and pH titration curves for PC / PI / PI(4,5)P2 (88% / 10% / 31 2%). A) P MAS NMR spectra as a function of pH for 2 mol% PI(4,5)P2 in 88% PC / 10% PI

vesicles. b). Peak positions of the 4- and 5-phosphate of PI(4,5)P2 as a function of pH. c). Ionization

model and pKa values for deprotonation of PI(4,5)P2 with PI. d) Charge of the 4- and 5-phosphate of

PI(4,5)P2 as a function of pH. e) Plot showing the relative prevalence of each ionization state as a

function of pH. f0 is the protonated fraction, f4 is the fraction with P4 deprotonated, f5 is the fraction

with P5 deprotonated, and f45 is the fully deprotonated fraction. in the presence and absence of PI are essentially superimposable. For pH values > 7 (dissociation of the second proton), the ionization of both PI(4,5)P2 phosphate groups is decreased in the presence of PI (upfield shift of the corresponding peaks). The pKa values show this clearly— pKa4 increases from 7.60 to 7.84, while pKa5 increases from 7.41 to 7.66. Figure 4-6D shows the charge is decreased at higher pH as well (as compared to the charge for PC/PI(4,5)P2). To explore whether increased PI concentrations and the associated charge increase result in a stronger effect on PI(4,5)P2 ionization behavior, we investigated the pH dependent chemical shift variation of PI(4,5)P2 in the presence of 20% PI (PC/PI/PI(4,5)P2 78%:20%:2%, see appendix

A5). The data obtained in the presence of 20% PI matched those obtained for 10% PI (see Figure

A6), i.e., the increased PI concentration had no effect on PI(4,5)P2 ionization behavior. There are two possible explanations for this behavior: Either PI and PI(4,5)P2 demix and as a result

PI(4,5)P2 is not impacted by the enhanced charge density due to the increased PI concentration or

PI and PI(4,5)P2 form a mixed phase that is enriched in PI and PI(4,5)P2. In the latter case, the

PI(4,5)P2 ionization state is not affected by the increased PI concentration because it is already in a PI rich environment regardless of the total percentage of PI.

To explore this aspect further, we investigated by confocal fluorescence microscopy the morphology of giant unilamellar vesicles (GUVs) composed of PC/PI/PI(4,5)P2 for varying

PI(4,5)P2 concentrations (Figure 4-7) (Jiang 2010). In the absence of PI(4,5)P2, an uniform vesicle is observed, suggesting that PC and PI are mixed. This observation is in line with DSC experiments carried out with different DPPI/DPPC mixtures that showed a good mixing behavior of the two lipids (Redfern & Gericke, unpublished results). However, already small amounts of

PI(4,5)P2 result in the formation of a small “bulge” that is indicative of the formation of

PI/PI(4,5)P2 domains (a PC/PI(4,5)P2 95%:5% mixed vesicle does not show such a bulge, not shown). This “bulge” increases as the PI(4,5)P2 concentration in the GUV is increased, clearly indicating that it is associated with PI and PI(4,5)P2. We have observed this “bulging” also in other GUV lipid model systems where phosphoinositide enriched phases were formed. We tested by confocal microscopy whether the “bulging” is a result of two vesicles lying on top of each

Figure 4-7: Cooperative domain formation of PI and PI(4,5)P2 in POPC GUVs. GUVs were

composed of POPC and liver PI (20%) with different concentrations of brain PI(4,5)P2, from left to

right, 0% brain PI(4,5)P2, 5% brain PI(4,5)P2, 10% brain PI(4,5)P2 and 20% PI(4,5)P2 respectively. All GUVs were developed in a pH 7 buffer with 5 mM PIPES, 0.1 mM EDTA and 100 mM NaCl. All images were taken by fluorescence microscopy at room temperature. Scale bars shown are 10 μm. These GUV experiments were performed by Zhiping Jiang (Jiang 2010).

other and found this not to be the case. The observation that PI and PI(4,5)P2 are able to form a separate phase is consistent with the fact that it is possible to make mixed large unilamellar vesicles from PI and PI(4,5)P2 (not shown). In contrast, for PC/PE/PI(4,5)P2 mixtures we did not find any evidence for PI(4,5)P2/PE domain formation (not shown).

Discussion

The effect of phosphatidylethanolamine (PE) on the ionization state of the PI(4,5)P2 phosphomonoester groups is significantly less pronounced than the corresponding impact of PE on phosphatidic acid ionization. Another striking observation is the fact that the ionization of the

4-phosphate is in comparison to the vesicles without PE slightly increased below pH 7, while at a higher pH the ionization state of the 4-phosphate group is unaffected by the presence of PE. In contrast, the ionization state of the 5-phosphate group is increased for the entire investigated pH range and the magnitude of the ionization increase for low pH values is significantly stronger than for the 4-phosphate.

Any hydrogen bond formation between PE and PI(4,5)P2 has to compete with intramolecular hydroxyl/phosphomonoester group hydrogen bond formation. In the case of phosphatidic acid the presence of PE caused a lowering of the pKa2 by approximately one pH unit (Kooijman,

Carter et al. 2005), while we observe a maximum decrease of 0.43 for PI(4,5)P2 (for the first deprotonation step). Lysophosphatidic acid (LPA) exhibits in PC bilayers a lower pKa2 than PA

(7.4 vs. 7.9), while the pKa2 values for PA and LPA in the presence of PE are almost identical

(~6.9), i.e., the pKa2 change is less pronounced for LPA than it is for PA (Kooijman, Carter et al.

2005). We have shown previously that PI(4,5)P2 forms an extensive intra- and intermolecular

hydrogen bond network (Kooijman, King et al. 2009). Following the argument made for the pKa changes of PA and LPA in the presence of PE, it is likely that the lower than expected effect of

PE on the ionization properties of PI(4,5)P2 is due to the existing intramolecular hydrogen bonds.

It is noteworthy, however, that the phosphate group in the 5-position is more strongly affected by the presence of PE than the 4-phosphate, which is impacted to a significantly lesser degree. This suggests that hydrogen bond formation between the 5-phosphate and PE is more favorable, which is probably due to a PI(4,5)P2 ring orientation that is unfavorable for an interaction of the

4-phosphate group with PE. An alternative explanation would be that the 4-phosphate forms stronger intramolecular hydrogen bonds than the 5-phosphate. The preferred interaction of PE with the 5-phosphate results in a stronger change in the negative charge for this group than for the 4-phosphate (-1.43 to -1.50 vs. -1.60 to -1.62, at pH 7.0). Another interesting feature is that the first deprotonation step is affected more strongly by the presence of PE than the deprotonation of the last remaining proton that is usually shared between the two phosphate groups (second ionization step). In the case of the second ionization step the last remaining proton might be more strongly associated with the 4-phosphate. Overall, the reduced impact of

PE on PI(4,5)P2 ionization at higher pH might suggest that the interaction between PE and

PI(4,5)P2 is weakened once only one shared proton is left.

The interaction behavior between PS and PI(4,5)P2 appears to be complex. The pKa of the PS carboxyl group is about 3.6, i.e., below the pKa PS is zwitterionic, while above the pKa the lipid is anionic (see figure A3). Even though the observed changes of the ionization state of the two

PI(4,5)P2 phosphate groups is small (and within the range of the error of the method), this would explain why the ionization of the phosphate groups is slightly lower for pH values above the 95

carboxyl pKa. An important finding is that the ionization state of PA seems to be also not strongly affected by the presence of PS (Figure 4-4). This suggests that PA and PI(4,5)P2 demix from PS or that the effect of the negative charge on the interfacial pH and hence the protonation state of the respective phosphomonoester groups is countered by hydrogen bond formation between the PS ammonium group and the phosphomonoester groups. Previous data (Kooijman,

Carter et al. 2005) for PA was based on PC/PE mixed membranes. The explanation for this observation might be that PE donates a hydrogen for hydrogen bond formation so that the additional effect of PS is more noticeable (PS counters the effect of PE to some extent), or PE facilitates PS/PA mixing in the PC/PA membrane. Additionally, the observed effect of PS may have been caused by a reduction of PE/PA hydrogen bond formation due to replacement of PE by PS, rather than a direct effect of PS. Based upon our current data set we cannot unequivocally answer this question.

The effect of phosphatidylinositol (PI) on PI(4,5)P2 is two-fold. The negative charge of the PI headgroup results in an increased interfacial proton concentration (lower interfacial pH), which should give rise to an increased protonation of the PI(4,5)P2 phosphomonoester groups (shift to lower ppm values). On the other hand, the hydroxyl groups of the PI headgroup can engage in hydrogen bond formation with the PI(4,5)P2 phosphomonoester groups, which is expected to lead to an enhanced deprotonation of the phosphomonoester groups, i.e., a higher charge and a downfield shift of the peaks observed in the NMR spectra. Apparently, the two effects largely compensate each other, i.e., the net change of the PI(4,5)P2 ionization state in the presence vs. absence of PI is small. While for the PI(4,5)P2/PE/PC mixed vesicle system the first ionization step was more strongly affected by the presence of PE than for the removal of the last remaining 96

proton, it is opposite for the PI(4,5)P2/PI/PC vesicle system. The data points for pH values between ~ 4 and ~ 6.5 are essentially superimposable for the PC/PI/PI(4,5)P2 and PC/PI(4,5)P2 bilayer systems. Above pH 6.5, the two data sets deviate from each other: In the presence of PI a lower degree of ionization is observed for both PI(4,5)P2 phosphomonoester groups, as reflected by the higher pKa values for the second deprotonation step. This suggests that the removal of the last proton is less affected by intermolecular PI/PI(4,5)P2 hydrogen bond formation than the dissociation of the first proton (please note that the charge of PI does not change in the investigated pH range). In contrast to the PI(4,5)P2 ionization in the presence of PE, both phosphomonoester groups are equally affected by the presence of PI. PI and PC form a largely mixed phase (see Figure 4-7). Even though some evidence exists that in mixed PC/PI(4,5)P2 vesicles PI(4,5)P2 may form domains that are too small to be observed by microscopy (Redfern and Gericke 2005; Kooijman, King et al. 2009), they certainly don’t show macroscopic demixing for the conditions used for the experiments described here (ionic strength, temperature). It is therefore extraordinary that macroscopic domains are formed when PI and PI(4,5)P2 are present together in the membrane. While for the GUV experiments shown in figure 4-7 the PI(4,5)P2 concentration is above the global physiological concentration of the lipid (about 1%), we also have evidence that this domain formation occurs at lower, physiologically relevant PI(4,5)P2 concentrations. For the NMR experiments we used 2 mol% of PI(4,5)P2, which is only slightly above the global physiological concentrations (please note that local PI(4,5)P2 concentrations might be well above 1 mol%). For varying PI concentrations (10 mol% vs. 20 mol%) we didn’t find any change in the pH dependent chemical shift variation. This suggests that PI(4,5)P2 experiences the same environment for 10 mol% PI as it experiences for 20 mol%, which can

only be explained based upon the assumption that PI and PI(4,5)P2 form mixed domains.

Strikingly, neither of these lipids form macroscopic domains in binary mixtures with PC, i.e., they both have to be present for the macroscopic domains to occur under these conditions. We also lowered the PI(4,5)P2 concentration in our GUV experiments and for 2 mol% PI(4,5)P2 we found a very small bulge (not shown). Taken together, we are confident that for physiological

PI(4,5)P2 concentrations these mixed PI/PI(4,5)P2 domains form. Why is the formation of a mixed PI/PI(4,5)P2 domain favored? While PI(4,5)P2 may show some weak mutual attractive interaction due to the formation of hydrogen bonds that leads to a “smearing” of the headgroup charge (Redfern and Gericke 2005; Kooijman, King et al. 2009), it appears that the presence of

PI strongly enhances the local accumulation of PI(4,5)P2. The presence of PI in a PI(4,5)P2 rich domain “dilutes” the highly negative charge of PI(4,5)P2 and at the same time, PI and PI(4,5)P2 can engage in hydrogen bond formation that leads to a further “smearing” of the charges, i.e., in a reduction of the charge density. Since this PI/PI(4,5)P2 interaction occurs at concentrations that are close to the global physiological concentrations for the two lipids, it is likely that such an interaction is also relevant for cellular systems. The observation that the presence of PI affects the ionization state of both PI(4,5)P2 phosphate groups while PE seems to interact primarily with the 5-phosphate might suggest that the PI(4,5)P2 ring re-orients in the presence of PI.

Summary

The finding that PI and PI(4,5)P2 form cooperatively macroscopic domains in ternary mixtures with PC (as seen in fluorescence microscopy measurements and inferred from the NMR experiments) has potentially far reaching implications. These results suggest that PI might be

involved in the spatial organization of phosphoinositide signaling events by promoting clustering of phosphoinositide pools, however, it remains to be seen whether such pools can be found in vivo.

The finding that PS had only a minimal effect on the ionization state of PI(4,5)P2 was unexpected. Previous work on the effect of PS on the ionization state of PA showed a pronounced impact, i.e. increasing molar concentrations of PS (anionic charge) diminished the degree of ionization of PA by decreasing the interfacial pH (Kooijman, Carter et al. 2005).

However those data were obtained in mixtures with PE. Our new data for PA in the presence of only PS (in a PC matrix) replicates our observations for PI(4,5)P2. This suggests that PA, like

PI(4,5)P2 does not mix well with PS when present in a PC matrix. It should be noted that the acyl chain component for PC, PS and PA is identical, namely dioleoyl, i.e., the putative demixing does not occur due to acyl chain mismatch. As PA does not form domains on its own in a PC matrix at neutral pH (Kooijman, unpublished results) this only leaves the conclusion that PS and

PA do not mix in ternary PC/PS/ PA mixtures. The aminophospholipid PE on the other hand appears to be able to reverse this demixing as evidenced by previous data (Kooijman, Carter et al. 2005). Thus, the striking observation that PS has little effect on the charge of PI(4,5)P2 and

PA suggests that under suitable conditions PS is able to segregate from other membrane components without the help of proteins or divalent cations. Cholesterol mediated phase separation of PS from PC has been observed in Monte Carlo simulations, using interaction parameters derived from experiment (Almeida, Best et al. 2011). The ability of PS to form domains in the plasma membrane might have important implications, and act to separate the

unique physiological functions of these lipids (Fairn, Hermansson et al. 2011; Fairn, Schieber et al. 2011).

Taken together our pH titration and GUV fluorescence data reveal an intriguing ionization and mixing behavior for PI(4,5)P2 that is considerably more complex than initially assumed. The fluid-fluid mixing of PI and PI(4,5)P2 is likely to have important implications for PI(4,5)P2 mediated signaling events, and the complex behavior of PS deserves more attention in the future.

100

Chapter 5†

Phosphatidylinositol-4,5-bisphosphate ionization in the presence of cholesterol, calcium or

magnesium ions

Introduction

The distribution and turnover of PI(4,5)P2 in activated cells changes more than the total

PI(4,5)P2 levels (Varnai and Balla 2006; Varnai, Thyagarajan et al. 2006). In this context the question arises how PI(4,5)P2 mediated signaling events are fine tuned in space and time. Local

PI(4,5)P2 accumulation has been demonstrated to occur in vivo (Franca-Koh, Kamimura et al.

2007; James, Khodthong et al. 2008; Johnson, Chichili et al. 2008; Gao, Lowry et al. 2011). This accumulation may play a large role in the control of PI(4,5)P2 signaling, and there are many cellular components that have been suggested to promote this clustering of PI(4,5)P2, including cationic proteins such as MARCKS (Wang, McLaughlin et al. 2003; Kwiatkowska 2010), divalent cations (Wang, Collins et al. 2012), cholesterol (Dasgupta, Bamba et al. 2009; Jiang,

Redfern et al. 2014), and phosphatidylinositol (PI) (Graber, Jiang et al. 2012).

2+ Calcium (Ca ) has been identified as an important factor in many PI(4,5)P2 signaling events.

2+ Ca is thought to promote local accumulation, or clustering, of PI(4,5)P2 by shielding its negative charge which shifts the balance between repulsive charge interactions (the PI(4,5)P2

† Adapted from Graber, Z. T., A. Gericke, et al. (2014). "Phosphatidylinositol-4,5-bisphosphate ionization in the presence of cholesterol, calcium or magnesium ions." Ibid. 182: 62-72. 101

headgroup charge is about -4 at physiological pH) and attractive forces like intermolecular hydrogen bond formation towards a net force that is attractive. Ca2+ is also important for mediating C2 domain binding to PI(4,5)P2 (Evans, Gerber et al. 2004; Lyakhova and Knight

2014; Stahelin, Scott et al. 2014). Ca2+ has been suggested to alleviate MARCKS based sequestering of PI(4,5)P2 by forming a complex with Calmodulin and binding to the MARCKS

2+ protein (McLaughlin and Murray 2005). PI(4,5)P2 and Ca are also linked through the classical

PLC signaling pathway, whereby PLC cleaves the headgroup of PI(4,5)P2 to form Ins(1,4,5)P3, which causes the release of Ca2+ from the endoplasmic reticulum. This Ca2+ release may have a strong impact on PI(4,5)P2 signaling, raising the possibility that it acts as a feedback loop. While

2+ 2+ Ca has been shown to have an important role in PI(4,5)P2 signaling, magnesium (Mg ) shares calcium’s divalent charge and is found in much higher concentrations in the cytosol. It is

2+ therefore important to consider the effect of Mg on PI(4,5)P2 as well. Previous work has shown

2+ 2+ 2+ that the binding constant of both Ca and Mg for PI(4,5)P2 are similar but that only Ca is able to induce local PI(4,5)P2 clustering (Wang, Collins et al. 2012).

Cholesterol (see Figure 5-1) is a major component of the cell membrane, playing an important role in the formation of so-called ‘raft’ domains. Cholesterol is also of relevance in PI(4,5)P2 signaling since several PI(4,5)P2 signaling events were found to be affected by cholesterol levels

(Elhyany, Assa-Kunik et al. 2004; Cinar, Mukhopadhyay et al. 2007; Lasserre, Guo et al. 2008;

Chun, Shin et al. 2010; Murray and Tamm 2011; Koushik, Powell et al. 2013). Based upon the cholesterol dependence of some PI(4,5)P2 mediated signaling events, some have suggested that

PI(4,5)P2 partitions into so-called lipid rafts, while others have challenged this notion (van

Rheenen, Achame et al. 2005). The acyl chain composition of PI(4,5)P2 is stearoyl-arachidonoyl

102

and therefore, PI(4,5)P2 is unlikely to partition into ordered lipid domains. For outer leaflet raft compositions (enriched in sphingolipid and cholesterol) it has been found that raft resident

PI(4,5)P2 binding proteins or peptides are required for PI(4,5)P2 partitioning into lipid rafts

(Tong, Nguyen et al. 2008). Despite this, in vitro experiments have shown cholesterol to promote

PI(4,5)P2 cluster formation (Dasgupta, Bamba et al. 2009; Jiang, Redfern et al. 2014). The nature of the interaction that leads to cholesterol induced PI(4,5)P2 cluster formation is only beginning to emerge.

Electrostatics are obviously important for protein-PI(4,5)P2 interactions due to PI(4,5)P2’s high negative charge. 31P NMR is uniquely able to monitor independently the ionization state of the respective phosphomonoester groups based upon the observed chemical shifts. In this study,

31 we have utilized MAS P NMR spectroscopy to investigate the interaction of PI(4,5)P2 with

Ca2+, Mg2+, and cholesterol, which allows us to characterize the effect of these chemical species on the ionization state of the phosphoinositide headgroup.

Figure 5-1: Chemical structures of the lipids. From left to right: cholesterol, phosphatidylinositol

(PI), phosphatidylethanolamine (PE), and phosphatidylinositol-4,5-bisphosphate [PI(4,5)P2].

103

Results

2+ 2+ Interaction of PI(4,5)P2 with the Divalent Cations Mg and Ca

PI(4,5)P2, as an anionic lipid with a high negative charge, can interact with many other membrane components via hydrogen-bond or non-specific electrostatic interactions. This enables

PI(4,5)P2 to interact with both positively charged peptides or protein domains, as well as divalent cations within the cell. Ca2+ and Mg2+ are both important cations in the cell, and interact with

2+ 2+ PI(4,5)P2 through their positive charge. We seek to understand to what extent Mg and Ca interact with each of the phosphomonoester groups of PI(4,5)P2, and how this interaction affects the ionization state of PI(4,5)P2.

In order to distinguish experimentally between the 4- and 5-phosphate we used MAS 31P-

NMR experiments of mixed, DOPC/PI(4,5)P2, multilamellar vesicles (MLVs), with a concentration of 3.5-5 mol% brain PI(4,5)P2 and the remainder being DOPC. Varying amounts

2+ of CaCl2 and MgCl2 solutions were added to obtain bulk concentrations of 0.1, 1, or 2 mM Mg or Ca2+. The Ca2+ and Mg2+ ionophore A23187 was used to ensure that the cations were able to penetrate into the interior of the MLVs. The cations were judged to be equilibrated evenly throughout the MLVs when single peaks for the 4- and 5-phosphate of PI(4,5)P2 were observed.

Figure 5-2 shows two clearly resolved peaks for the phosphomonoester groups, indicating that they are both interacting with similar concentrations of cations, i.e. sensing a similar environment (Kooijman, Carter et al. 2005). Static NMR experiments were performed to show that for all samples the lipids formed MLVs as indicated by spectra showing a chemical shift anisotropy (CSA) with a low field shoulder and a high field peak, indicative of a bilayer

104

configuration (Cullis and Dekruijff 1979) (see appendix Figure A8 (A) and (B)). The CSA observed around 0 ppm is a combination of smaller vesicles present in the sample and the CSA of the phosphates of PI(4,5)P2 (not clearly resolved in these spectra). Peak assignments were made according to Kooijman et al. (Kooijman, King et al. 2009): the most downfield peak is the

4-phosphate of PI(4,5)P2, followed by the 5-phosphate of PI(4,5)P2 slightly upfield of the 4- phosphate peak. The PC and PI(4,5)P2 phosphodiesters are furthest upfield in one overlapping peak. In accordance with our previous observations (Kooijman, King et al. 2009), in this low concentration regime, the molar ratio of PI(4,5)P2 did not influence the chemical shift of the phosphomonoesters. Hence, most experiments were conducted at the more physiological molar ratio of ~3.5% PI(4,5)P2 (McLaughlin and Murray 2005).

Fig. 5-2A shows the MAS NMR spectra for PI(4,5)P2 in the presence of 0, 0.1, 1.0, and 2.0 mM Ca2+. The addition of 0.1 mM Ca2+ leads to a distinct downfield shift of both phosphomonoester peaks. This downfield shift becomes more significant as the Ca2+ concentration is increased. The phosphodiester peaks show no significant Ca2+ dependence at any Ca2+ concentration, as expected since the phosphodiester groups are completely deprotonated at pH 7.2 (the pKa value is expected to be in the pH 2 – 3 range) (Marsh 2013). The observed downfield shift of the phosphomonoester peaks is due to enhanced deshielding upon increased deprotonation of the respective phosphate group. Ca2+ binding to the bilayer and

PI(4,5)P2 headgroup leads to a local decrease in the negative bilayer surface potential and hence an increased interfacial pH, which subsequently leads to an increased average charge for the phosphomonoesters. Average chemical shift values at each of the tested Ca2+ concentrations are shown in Fig. 5-2B. The chemical shift increase to downfield values can be clearly seen from the 105

bar graph. In the presence of 2 mM Ca2+ the 4-phosphate peak chemical shift value increases from 3.52 to 3.84 ppm, while the 5-phosphate peak shifts from 2.38 to 3.01 ppm. The 5- phosphate is thus more strongly affected by the addition of Ca2+, increasing by 0.63 ppm, while the 4-phosphate chemical shift increases only by 0.32 ppm.

2+ 31 Figure 5-2: Effect of Ca on the ionization properties of PI(4,5)P2. a).Representative P MAS NMR 2+ spectra for 3.5 mol% PI(4,5)P2 in 96.5% PC vesicles in the presence of increasing amounts of Ca . From 2+ bottom to top, 0, 0.1, 1, and 2 mM Ca b) Peak values for the 4- and 5-phosphates of PI(4,5)P2 as a function of Ca2+ concentration. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). For the 4-phosphate: * significantly different from 0 mM Ca2+. ** significantly different from all other mixtures. For the 5-phosphate, each concentration of Ca2+ results in a statistically significant difference. For each sample mixture the following number of trials were run: 2+ 2+ 2+ 2+ PC:PI(4,5)P2 0 mM Ca : 4 trials, 0.1 mM Ca 3 trials, 1 mM Ca : 4 trials, 2 mM Ca : 3 trials. The pH for all sample mixtures was 7.1 ± 0.1.

Figure 5-3A shows the MAS NMR spectra for PI(4,5)P2 in the presence of 0, 1, and 2 mM bulk

Mg2+ concentration. Since the addition of 1 mM Mg2+ leads to only a slight downfield shift of the 4- and 5-phosphate peaks of PI(4,5)P2, we did not investigate the lower 0.1 mM concentration as we did for Ca2+ . Additionally, the Mg2+ concentration is significantly higher 106

inside cells than the Ca2+ concentration and hence the relevant concentrations to investigate the

2+ PI(4,5)P2 ionization in the presence of Mg are in the millimolar range. The minor shift observed for 1 mM Mg2+ becomes more significant when the Mg2+ concentration is increased to

2 mM. Again, the phosphodiester peak of PI(4,5)P2 and PC does not shift, as expected. As with

2+ Ca , the increased deshielding of the 4- and 5-phosphate groups of PI(4,5)P2 is due to increased deprotonation caused by the decrease in the negative electrostatic potential at the bilayer interface, a lower interfacial proton concentration, and hence a higher interfacial pH. The chemical shift values for the 4-, and 5-phosphates were averaged for each condition and are plotted in the bar graph in Figure 5-3B. While the chemical shift of the 5-phosphate increases with increasing Mg2+ concentration, the chemical shift of the 4-phosphate is essentially

2+ 31 Figure 5-3: Effect of Mg on the ionization properties of PI(4,5)P2. a). Representative P MAS

NMR spectra for 3.5 mol% PI(4,5)P2 in 96.5% PC vesicles in the presence of increasing amounts of 2+ Mg . From bottom to top, 0, 1, and 2 mM b) Peak values for the 4- and 5-phosphate of PI(4,5)P2 as a function of Mg2+ concentration. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). For the 4-phosphate, there were no significant differences. For the 5-phosphate, each concentration of Mg2+ results in a statistically significant

difference. For each sample mixture the following number of trials were run: PC:PI(4,5)P2 0 mM Mg2+: 4 trials, 1 mM Mg2+ 4 trials, 2 mM Mg2+: 3 trials. The pH for all sample mixtures was 7.1 ± 0.1. 107

unaffected by the addition of Mg2+. The shift (0.02 ppm) observed for the 4-phosphate peak, upon an increase in Mg2+ concentration from 0 to 2 mM, is statistically not significant. In contrast, the 5-phosphate peak shifts from 2.38 ppm to 2.71 ppm, a shift of 0.33 ppm, which is statistically relevant.

The average chemical shift for the 4- and 5-phosphates of PI(4,5)P2 in the presence of 0 – 2 mM Ca2+ and Mg2+ are summarized in Table 5-1. At a cation concentration of 2 mM, both phosphomonoester peaks are found about 0.3 ppm further downfield in the presence of Ca2+ than they are observed for Mg2+. At the lowest Ca2+ concentration tested (0.1 mM), the shift of the 5- phosphate peak was as large as the shift for Mg2+ at a 10 times higher concentration, while the shift of the 4-phosphate was at best marginal for 0.1 mM Ca2+. Overall, these results indicate that

2+ 2+ Ca has a larger interaction with PI(4,5)P2 compared to Mg . For both cations we observe a much greater change in the chemical shift of the 5-phosphate as compared to the 4-phosphate.

Using equation 2-1 we can estimate the charge at each phosphate in the presence of Ca2+ or

Mg2+. The estimated charges are summarized in the last column of Table 5-1. The negative charge increases by as much as 0.15 for the 5-phosphate in the presence of 2 mM Ca2+, while the change is about half (0.08) in the presence of 2 mM Mg2+. The negative charge of the 4- phosphate increases only slightly by about 0.1 in the presence of 2 mM Ca2+, while no charge increase is observed in the presence of 2 mM Mg2+.

108

2+ 2+ Table 5-1. PI(4,5)P2 phosphomonoester chemical shifts in the presence of Ca and Mg Concentration Chemical Shift (ppm) Charge 2+ 2+ 2+ 2+ (mM) Ca Mg Ca Mg 0 3.52 ± 0.01 -1.60 ± 0.01

PI(4,5)P2 0.1 3.59 ± 0.04 -- -1.62 ± 0.01 -- 4-phosphate 1 3.61 ± 0.03 3.50 ± 0.02 -1.62 ± 0.01 -1.59 ± 0.01 2 3.84 ± 0.05 3.54 ± 0.06 -1.69 ± 0.02 -1.60 ± 0.02 0 2.38 ± 0.02 -1.42 ± 0.01

PI(4,5)P2 0.1 2.52 ± 0.06 -- -1.45 ± 0.01 -- 5-phosphate 1 2.63 ± 0.03 2.55 ± 0.04 -1.48 ± 0.01 -1.46 ± 0.01 2 3.0 ± 0.1 2.71 ± 0.06 -1.57 ± 0.02 -1.50 ± 0.01

The varying effect of divalent cations on PI(4,5)P2

The cellular divalent cations Mg2+ and Ca2+ have widely different levels of interaction with

2+ 2+ PI(4,5)P2. Ca and Mg are quite distinct ions, with significantly different hydration energies and ionic radius. To gain insight into the mechanism of this differential binding, we have

2+ 2+ examined the interaction between PI(4,5)P2 and the cations Ba and Ni . Fig. 5-4A shows the

2+ 2+ MAS NMR spectra for PI(4,5)P2 in the presence of no divalent cations, 1 mM Ba , 1mM Ca , 1

2+ 2+ mM Mg , and 1 mM Ni . The average chemical shift values for each of the PI(4,5)P2/cation mixtures are shown in Figure 5-4B The addition of 1 mM Ba2+ leads to a distinct downfield shift of both phosphomonoester peaks. Unlike for Ca2+ and Mg2+, Ba2+ affects both phosphomonoesters nearly equally. Ni2+ also causes a downfield shift of the two phosphomonoester peaks. Ni2+ is more similar to Ca2+ and Mg2+, causing a larger shift for the 5- phosphate than for the 4-phosphate (0.19 vs. 0.10). When we compare the various cations, we see that for the 4-phosphate Ca2+ and Ni2+ have the largest effect, while Mg2+ has the smallest

109

effect. For the 5-phosphate, Ca2+ has the largest effect while Ni2+ and Ba2+ have the smallest effect.

31 Figure 5-4: Effect of cations on the ionization properties of PI(4,5)P2. a). Representative P MAS 2+ NMR spectra for 3.5 mol% PI(4,5)P2 in 96.5% PC vesicles in the presence of no divalent cations, Ba , 2+ 2+ 2+ Ca , Mg , or Ni (1 mM each). b) Peak values for the 4- and 5-phosphate of PI(4,5)P2 as a function of cation species. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). The pH for all sample mixtures was 7.1 ± 0.1.

Interaction of cholesterol with PI(4,5)P2

The interaction between PI(4,5)P2 and cholesterol is somewhat enigmatic and poorly understood. Some evidence suggests that PI(4,5)P2 signaling is cholesterol dependent and that

PI(4,5)P2 may be resident in lipid “rafts”. However, the natural chain composition of PI(4,5)P2 is highly unsaturated and would favor more disordered, more fluid environments than is found in lipid raft domains. Considering the rigid sterol ring structure and the highly unsaturated and therefore, disordered, PI(4,5)P2 acyl chain structure, one might expect that cholesterol/PI(4,5)P2 110

interaction is not favorable. Since the opposite is observed, we hypothesize that the cholesterol/PI(4,5)P2 interaction is linked to hydrogen-bond formation between the cholesterol hydroxyl group and functional groups in the PI(4,5)P2 headgroup region. In support of this notion, Jiang et al. have shown that sterols lacking the hydroxyl group do not promote domain formation (Jiang, Redfern et al. 2014). We seek to shed some light on the origin of the interaction between cholesterol and PI(4,5)P2 by investigating the effect of cholesterol incorporation on the PI(4,5)P2 4- and 5-phosphate ionization.

To investigate the interaction of PI(4,5)P2 and cholesterol we formed MLVs from mixtures of

DOPC, Cholesterol, and PI(4,5)P2. Several cholesterol concentrations from 0 to 40 mol% were tested. We used either 5 or 2 mol% PI(4,5)P2 in order to provide a convenient comparison against our previous PC:PI(4,5)P2 data with 5 mol% PI(4,5)P2. The 2 mol% PI(4,5)P2 was used to mimic more closely physiological plasma membrane concentrations. Again, this change in

PI(4,5)P2 concentration was found to have no impact on the chemical shift of each of the phosphomonoester peaks. As expected, cholesterol has no effect on the chemical shift of the

31 phosphodiester peaks of PC and PI(4,5)P2. Static P NMR spectra were acquired for all samples

(see appendix Figure A8(C)) which confirmed that cholesterol up to 40 mol% did not significantly affect the bilayer organization of our mixed lipid samples.

Figure 5-5A shows representative MAS NMR spectra for PI(4,5)P2 in the presence of increasing amounts of cholesterol ranging from 0 to 40 mol% (bottom to top spectra). As the cholesterol concentration is increased, the 4- and 5- phosphate peaks are observed to shift downfield. The respective average chemical shift peak positions are plotted for increasing

111

amounts of cholesterol in Figure 5-5B. The presence of cholesterol concentrations of 20 mol% did not have a statistically relevant effect on the chemical shift of the 4-phosphate peak. The incorporation of 30-40 mol% cholesterol show a statistically significant, albeit small, shift in the chemical shift of the 4-phosphate peak, while the peak shift observed for the 5-phosphate is more pronounced for all of the cholesterol concentrations tested.

31 Figure 5-5: Effect of cholesterol on the ionization properties of PI(4,5)P2. a). Representative P MAS

NMR spectra for PC / PI(4,5)P2 vesicles with 0, 20, 30, and 40 mol% cholesterol respectively (bottom to top). Mixtures contained 5 mol% PI(4,5)P2 except for the 40 mol% cholesterol sample which contained only 2 mol% PI(4,5)P2, the remainder was DOPC. b) Peak values for the 4- and 5-phosphates of PI(4,5)P2 as a function of cholesterol concentration. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). For the 4-phosphate, the 40 and 30 mol% cholesterol mixtures were significantly different from the PC control. For the 5-phosphate, each mixture containing cholesterol was significantly different from the PC control. Three unique samples were tested for each sample mixture. The pH for all sample mixtures was 7.1 ± 0.1.

As with Ca2+ and Mg2+, the peak shifts observed for the 4- and 5-phosphate peaks in the presence of cholesterol are due to a deshielding of the two phosphomonoester groups leading to an increased negative charge of PI(4,5)P2. However, by itself, the downfield movement of the phosphomonoester peaks (i.e. increase in negative charge) induced by cholesterol does not

112

clearly indicate the nature of the type of interaction. In our previous work, we found that hydrogen-bond donor lipids in the membrane caused increased deprotonation of the PI(4,5)P2 phosphomonoesters (Graber, Jiang et al. 2012) following our earlier model for the deprotonation and protein interaction of phosphatidic acid, namely the electrostatic hydrogen-bond switch mechanism (Kooijman, Carter et al. 2005; Kooijman, Tieleman et al. 2007). The increase in charge induced by cholesterol is estimated using equation 2-1 (see Figures 5-6 to 5-8) with the assumption that cholesterol does not affect the chemical shift of the fully protonated and deprotonated species of each phosphomonoester.

To test the cause of the increase in negative charge by cholesterol we investigated the effect of PE incorporation. In Figure 5-6 we compare the 4- and 5-phosphate chemical shift values in the presence of cholesterol with the respective chemical shift values obtained for lipid systems that contain PE instead of cholesterol ( DOPC/Cholesterol/ PI(4,5)P2 at 58:40:2 molar ratio , and

31 DOPC/DOPE/PI(4,5)P2 at 47.5:47.5:5 molar ratio). Representative MAS P NMR spectra are shown in Figure 5-6A and are quantified in Figure 5-6B. From these data it is clear that the effect of PE, as a potential hydrogen-bond donor, is significantly larger than for cholesterol. The shift observed for the 4-phosphate at pH 7.2 is minor and comparable for both the cholesterol and PE containing membrane. This is consistent with our pH titration data which showed that the change in chemical shift for the 4-phosphate induced by PE is very minor at pH 7.2 (Graber, Jiang et al.

2012). However the downfield shift for the 5-phosphate is considerably larger for PE than for cholesterol. The addition of cholesterol to the ternary PC/PE/PI(4,5)P2 lipid mixture did not change the ionization of the phosphoinositide headgroup significantly. Our results suggest that the interaction between the PE headgroup and PI(4,5)P2 involves the phosphomonoester groups 113

of PI(4,5)P2 (though the 4-phosphate group is less affected), while cholesterol apparently does not interact directly with the phosphomonoester groups. We hypothesize that the change in phosphomonoester ionization in the presence of cholesterol is due to a reduced lateral PI(4,5)P2 density, which will be discussed further below.

31 Figure 5-6: P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with varying amounts 31 of PE and cholesterol. a). P MAS NMR spectra for PC / PI(4,5)P2 vesicles with 0 mol% cholesterol, 40 mol% cholesterol, 47.5 mol% PE, and 29 mol% PE and 40mol% cholesterol. Mixtures contained 2-5 mol% PI(4,5)P2 and the remainder was PC. b) Peak values for the 4- and 5-phosphates of PI(4,5)P2 in each of the vesicle mixtures. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). For the 4-phosphate: * statistically significant difference from the PC control, ** statistically significant difference from the PC control and the 40 mol% cholesterol mixture. For the 5-phosphate: * statistically significant difference from the PC control, ** statistically significant difference from all other mixtures. Three unique samples were tested for each sample mixture. The pH for all sample mixtures was 7.1 ± 0.1.

Comparison of the interaction of cholesterol and Phosphatidylinositol with PI(4,5)P2

Phosphatidylinositol (PI) is present in the inner leaflet of the plasma membrane in concentrations between 6-10% (van Meer, Voelker et al. 2008) and we have previously shown

114

that PI and PI(4,5)P2 co-localize in macroscopic domains in model lipid membranes (Graber,

Jiang et al. 2012). Cholesterol has a similar effect on PI(4,5)P2 in that it promotes macroscopic

PI/PI(4,5)P2 domain formation. Here we compare the effect of cholesterol and PI on the downfield movement of the chemical shift of the 4- and 5-phosphates of PI(4,5)P2. Additionally we examine the combined effect of PI and cholesterol.

Using 2 mol % PI(4,5)P2 we compare the PI(4,5)P2 ionization in the presence of 10 mol% PI with the headgroup ionization in the presence of cholesterol or both PI and cholesterol. The latter lipid mixture is relevant as a significant portion of cholesterol found in the plasma membrane is

31 present in the inner leaflet where both PI and PI(4,5)P2 are confined. Representative P MAS

NMR spectra for each mixture are compared with a PC/PI(4,5)P2 (96.5:3.5 molar ratio) standard as shown in Figure 5-7A. The phosphodiester of PI is not well resolved as it overlaps almost completely with the phosphodiester of PC and PI(4,5)P2, however, a downfield shoulder on the

PC/PI(4,5)P2 phosphodiester peak is observable (see top two spectra in Figure 5-7A). Results are quantified in Figure 5-7B. The presence of PI has no effect on the position of the 4- phosphomonoester peak, while a small but statistically relevant shift of the 5-phosphomonoester peak is observed at pH 7.2. PI has thus a minor effect on the ionization properties of PI(4,5)P2, as observed previously (Graber, Jiang et al. 2012), despite having a major effect on membrane morphology (lateral distribution of PI(4,5)P2). This is due to complex interactions between PI and PI(4,5)P2 induced by the rich functionality of both phosphatidylinositol headgroups. We have explained previously the lack of a shift of the phosphomonoester peaks in the presence of

PI with a competition between the higher negative surface potential due to the negative charge of

PI (leading to enhanced protonation of the PI(4,5)P2 phosphomonoester groups and lower 115

charge) and intermolecular PI/PI(4,5)P2 hydrogen-bond formation (leading to an enhanced phosphomonoester group deprotonation and hence higher charge). In the sum, these two effects almost cancel out.

31 Figure 5-7: P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with varying amounts 31 of PI and cholesterol. a). P MAS NMR spectra for PC / PI(4,5)P2 vesicles with 0 mol% cholesterol, 40 mol% cholesterol, 10 mol% PI, and 10 mol% PI and 40 mol% cholesterol. Mixtures contained 2-5 mol%

PI(4,5)P2 and the remainder was PC. b) Peak values for the 4- and 5-phosphates of PI(4,5)P2 in each of the vesicle mixtures. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). For the 4-phosphate: * statistically significant difference from the PC control. For the 5-phosphate: * statistically significant difference from the PC control. Three unique samples were tested for each sample mixture. The pH for all sample mixtures was 7.1 ± 0.1.

The small downfield shift of the 5-phosphate in the presence of PI contrasts with the downfield shift caused by the presence of cholesterol, cholesterol having a larger effect than PI.

While these mixtures contain 40 mol% cholesterol and only 10 mol% PI, we have previously shown that increasing the concentration of PI does not increase the ionization of PI(4,5)P2. This can be explained based upon the observation that PI(4,5)P2/PI organize into domains at 10 mol%

PI and therefore, the environment of PI(4,5)P2 does not change as the PI concentration is 116

increased beyond 10 mol% (Graber, Jiang et al. 2012). Considering the differences in PI(4,5)P2 ionization behavior in the presence of cholesterol and PI, we can thus conclude that while both cholesterol and PI induce macroscopic domain formation in PI(4,5)P2 containing model membranes (Dasgupta, Bamba et al. 2009; Graber, Jiang et al. 2012), the foundations of the respective interactions are likely to be quite different as might be expected based upon the radically different chemical structures of these membrane lipids (see Figure 5-1). The mixture containing both cholesterol and PI illustrates this further as there is no cumulative effect on ionization of the 5-phosphate of PI(4,5)P2 (see Figure 5-7B). This is further addressed in the discussion.

2+ 2+ Cholesterol and the divalent cations Ca and Mg have a cumulative effect on PI(4,5)P2

Next we investigated the effect of both cholesterol and Ca2+ on the ionization properties of

PI(4,5)P2 as both cause PI(4,5)P2 clustering in model membrane systems. In the natural system,

2+ both cholesterol and Ca are present and may interact with PI(4,5)P2.

We have characterized the combined effect of cholesterol and divalent cations (Ca2+ and

2+ Mg ) using 58:40:2 mol% DOPC:Cholesterol:PI(4,5)P2 MLVs, as we had used for our investigation of cholesterol-PI(4,5)P2 interaction. CaCl2 or MgCl2 solutions were added to the vesicles in order to reach 1 mM bulk cation concentrations. The A23187 ionophore was used again to equilibrate the divalent cations throughout the MLVs. All peaks were sharp and well- resolved; indicating that the cations were well distributed across the MLV bilayers. The static spectra acquired for the samples confirm that the lipids are organized in lipid vesicles (see appendix Figure A8 (F)). 117

Representative 31P MAS NMR spectra for each sample mixture are compared in Figure 5-8A.

In the presence of the divalent cations and cholesterol, both the 4- and the 5-phosphate peaks are shifted further downfield compared to the PC and PC/Cholesterol control experiments. Figure 5-

8B quantifies the downfield movement of the chemical shift values of the 4- and 5-phosphate peaks. The average chemical shift values and estimated charges for the 4- and 5-phosphate of

2+ 2+ PI(4,5)P2 in the presence of cholesterol and cholesterol/(Ca or Mg ) are listed in Table 5-2.

31 Figure 5-8: P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with divalent cations 31 and cholesterol. a). P MAS NMR spectra for PC / Chol / PI(4,5)P2 (58% / 40% / 2%) vesicles with no 2+ 2+ cations, with 1 mM Mg , and with 1 mM Ca . The spectrum for PC / PI(4,5)P2 (96.5% / 3.5%) is included for comparison. b) Peak values for the 4- and 5-phosphates of PI(4,5)P2 in each of the vesicle mixtures. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). * Statistically significant difference from the PC control. ** Statistically significant difference from all other mixtures. Three unique samples were tested for each sample mixture. The pH for all sample mixtures was 7.1 ± 0.1.

Only for 40 mol% cholesterol do we observe a downfield shift for the 4-phosphate peak that is beyond the sensitivity of the method. Addition of 1 mM Mg2+ does not significantly affect this downfield shift of the 4-phosphate peak, while addition of 1 mM Ca2+ does result in a slight shift

118

in comparison to what is observed in the absence of Ca2+ (see 40 mol% cholesterol). This is

2+ consistent with the data obtained for the binary PC/PI(4,5)P2 vesicles in the presence of Mg and Ca2+ as shown in Figures 5-2 and 5-3. The joint effect of cholesterol and Mg2+ or Ca2+ on the

5-phosphate ionization is also consistent with the data for Mg2+ and Ca2+ alone in that the addition of both divalent cations to the lipid mixtures containing cholesterol induces a stronger downfield movement of the chemical shift for the 5-phosphate peak than it is observed for the 4- phosphate peak. These results indicate that the interaction between the phosphomonoesters of

PI(4,5)P2 and the divalent cations are additive to the effect induced by cholesterol. These results are further quantified in Table 5-2 (see Table 5-1 for comparison).

Table 5-2. Comparison of PI(4,5)P2 phosphomonoester chemical shifts in the presence of cholesterol or cholesterol and divalent cations (PC/PI(4,5)P2/Cholesterol (98%-x/2%/x%) Cholesterol Divalent Chemical Shift Charge (mol%) Cations (ppm) 0 -- 3.52 ± 0.01 -1.59 ± 0.01 20 -- 3.55 ± 0.03 -1.61 ± 0.01 -- PI(4,5)P2 30 3.57 ± 0.03 -1.61 ± 0.01 4-phosphate 40 -- 3.59 ± 0.01 -1.62 ± 0.01 2+ 40 1 mM Mg 3.62 ± 0.02 -1.63 ± 0.01 2+ 40 1 mM Ca 3.68 ± 0.03 -1.64 ± 0.01 0 -- 2.38 ± 0.02 -1.42 ± 0.01 20 -- 2.47 ± 0.04 -1.44 ± 0.01 -- PI(4,5)P2 30 2.50 ± 0.05 -1.45 ± 0.01 5-phosphate 40 -- 2.50 ± 0.01 -1.45 ± 0.01 2+ 40 1 mM Mg 2.70 ± 0.01 -1.50 ± 0.01 2+ 40 1 mM Ca 2.74 ± 0.05 -1.51 ± 0.01

119

2+ PE and PI influence Ca binding to PI(4,5)P2

In addition to cholesterol we also observed the impact of the phospholipids PE and PI on

PI(4,5)P2 cation binding. PE has already been demonstrated to form hydrogen-bonds with

PI(4,5)P2 (see chapter 4 and (Graber, Jiang et al. 2012)) and this hydrogen-bond formation may affect the interaction between PI(4,5)P2 and the divalent cations. PI induced PI(4,5)P2 clustering in model membrane systems, and in the natural system is likely to be present when Ca2+ interacts with PI(4,5)P2. Therefore, it is important to investigate how these phospholipids affect the

2+ 2+ interaction between PI(4,5)P2 and Ca . We focused on the interaction between Ca and

2+ PI(4,5)P2 as it is the stronger interaction (as compared to Mg ).

We examined vesicles containing DOPC and PI(4,5)P2 along with either 10 mol% PI or 47.5 mol% PE. CaCl2 was added to the vesicles in order to reach 1 mM bulk cation concentrations, while the A23187 ionophore was used again to equilibrate the divalent cations throughout the

MLVs. The resulting MAS spectra show well-resolved peaks indicating that the Ca2+ is evenly distributed. The presence of lipid vesicles was confirmed by the static spectra (see appendix

Figure A8 (G)).

Representative 31P MAS NMR spectra for each sample mixture are shown in Figure 5-9A.

The data in the presence of the phospholipid and Ca2+ are compared against the data for the

2+ PC:PI(4,5)P2 vesicles and PC:PE:PI(4,5)P2 vesicles in the absence of Ca . In both cases (PE and

PI) the addition of Ca2+ leads to a downfield shift for both phosphates compared to the mixtures in the absence of Ca2+. Figure 5-9B shows the average chemical shift values for each sample mixture. 120

31 Figure 5-9: P MAS NMR spectra and bar graph for PC / PI(4,5)P2 vesicles with divalent cations 31 and the phospholipids PI and PE. a). P MAS NMR spectra for PC / PI(4,5)P2 vesicles with 47.5% PE 2+ or 10% PI and with 1 mM Ca compared against PC / PI(4,5)P2 (96.5% / 3.5%) vesicles with or without 2+ 1 mM Ca and PC / PE / PI(4,5)P2 (47.5% / 47.% / 5%) vesicles. Stars indicate statistical significance as determined by ANOVA pairwise comparison of the data sets (P < 0.005). The pH for all sample mixtures was 7.1 ± 0.1.

The presence of PE leads to a significant shift to higher chemical shift values. Despite this, the presence of Ca2+ in addition to PE is able to increase this shift even further. We can see in

2+ Figure 5-9B that the shift for the PC/PE/PI(4,5)P2 vesicles with 1 mM Ca is roughly equal to the combined shift for 1 mM Ca2+ and PE independently. Thus the effect of PE is additive to the effect of Ca2+. In the presence of PI we showed above (see Figure 5-7B) that PI alone leads to almost no shift relative to the base PC/PI(4,5)P2 mixture. However, when we combine PI and 1 mM Ca2+ we see a shift that is significantly greater than the shift for 1 mM Ca2+ in the absence of

2+ PI. This indicates that PI may actually enhance the interaction between Ca and PI(4,5)P2.

121

Discussion

2+ 2+ Interaction of PI(4,5)P2 with Ca and Mg

2+ 2+ The effect of Ca on PI(4,5)P2 is quite pronounced, even at the lowest Ca concentration

2+ investigated in this study. As Ca is drawn to the bilayer by the negative charge of PI(4,5)P2, it displaces protons from the bilayer interface, thus increasing the interfacial pH and increasing

PI(4,5)P2’s charge. While this effect was significant at the concentrations that we measured, cytosolic Ca2+ concentrations are generally lower than the values we used here, which may limit the physiological relevance of the increased ionizations observed in the presence of Ca2+.

However, local concentrations of Ca2+ may be transiently much higher, e.g. during Ca2+ efflux from the endoplasmic reticulum, and may approach the Ca2+ concentrations used in this study.

2+ The observed interaction of Ca with PI(4,5)P2, indicated by the downfield movement of the chemical shift and increase in negative charge of PI(4,5)P2, is much larger than that induced by

Mg2+. This observation is consistent with previous experimental and simulation studies of

2+ 2+ PI(4,5)P2, which showed a significant difference in the mode of binding of Ca over Mg

(Hauser and Dawson 1967; Slochower, Huwe et al. 2013; Wang, Slochower et al. 2014), even though the binding constants are similar. Specifically, Wang et al. found that divalent cation

2+ 2+ induced PI(4,5)P2 clustering occurred in the presence of Ca but not in the presence of Mg

(Wang, Collins et al. 2012). In our case, even 2 mM Mg2+ had no significant effect on the charge of the 4-phosphate and only a modest effect on the 5-phosphate ionization. This differential effect on the degree of ionization of PI(4,5)P2 could be explained by the much higher hydration energy (and hence larger hydration shell) of Mg2+ as compared to Ca2+ (-1922 vs. -1592 kJ mol-1 respectively) (Wolf and Cittadini 2003). Due to its higher hydration energy, Mg2+ ions hold onto 122

their hydration shell more strongly, which could diminish the electrostatic interaction with

PI(4,5)P2. Simulations performed by Slochower et al. support this hypothesis, as they found in

2+ their simulations that Mg binds more loosely (e.g. further away) to PI(4,5)P2 and is coordinated by water (Slochower, Huwe et al. 2013). While this may partially explain the difference between

Mg2+ and Ca2+, in actuality the situation may be more complex. Ba2+ has a lower hydration energy than Ca2+ or Mg2+ (-1288 kJ mol-1) and Ni2+ has a higher hydration energy than Ca2+ or

Mg2+ (-2092 kJ mol-1) (Hunt 1963). Therefore, if the strength of the binding is purely dependent on hydration energy than we would expect the strength of the binding to be in order of the respective hydration energies, that is Ni2+ < Mg2+ < Ca2+ < Ba2+. Instead we find that Ni2+ binds at comparable levels to Ca2+, and Ba2+ lies somewhere between Ca2+ and Mg2+, and the actual order of interaction is Mg2+ < Ba2+ < Ni2+ ≈ Ca2+. There must be other factors that determine the strength of the interaction, possibly based on the size of the hydrated ion (Mg2+ does have a larger hydrated radius (Nightingale 1959)). In studies of DMPA (1,2-dimyristoyl-sn-glycero-3- phosphate) and Ba2+/Ca2+, Bu et al. found similar results. Despite the lower hydration energy of the Ba2+ ion, Ca2+ had a much higher binding affinity to DMPA (Bu, Flores et al. 2009).

Slochower et al. discuss the main binding site of Ca2+ to be primarily at the 4-phosphate of

PI(4,5)P2. At first glance this appears to be in disagreement with our data, which show upon

2+ interaction of PI(4,5)P2 with Ca a stronger deprotonation of the 5-phosphate group relative to the 4-phosphate group. However, on closer inspection the differences between the two studies can be easily explained. First, it is not surprising that the Ca2+ interacts more strongly with the 4- phosphate in the Slochower et al. study as the charge of the 4-phosphate in their simulations was set to -2 and that of the 5-phosphate to -1. Within the confines of their modeling experiment this 123

is a reasonable choice of parameters for the simulation because our previously published data showed a higher charge at the 4-phosphate (-1.60) than at the 5-phosphate (-1.42), i.e., if one has to assign non-fractional charges to the two phosphate groups, one would assign the 4-phosphate to carry a -2 charge. Second, the simulations do not measure actual ionization properties of each of the phosphates, but instead calculate the lowest energy configuration of the system based on a choice of initial starting conditions. Third, Slochower et al.’s simulation of cation binding was performed with only one PI(4,5)P2 molecule present. In the membrane system with many

PI(4,5)P2 molecules clustering together in the presence of the cations the preferential binding site may be altered due to the formation of intermolecular hydrogen-bonds, as well as the possibility

2+ of cation bridging between molecules. Considering that Ca promotes PI(4,5)P2 clustering, it is likely that Ca2+ bridges adjacent phosphoinositide lipids. It is also possible that the orientation of

PI(4,5)P2 in the membrane environment reduces the preference for the 4-phosphate observed in the simulations. Several studies have suggested that PI(4,5)P2’s headgroup lies relatively flat on the bilayer (Bradshaw, Bushby et al. 1999; Li, Venable et al. 2009; Slochower, Huwe et al.

2013). In this orientation it may be that the 4-phosphate is less accessible than the 5-phosphate.

Indeed, in our experiments we have consistently noted a greater shift for the 5-phosphate in the presence of interaction partners, which indicates it is more sensitive to these interactions (see discussion below). It will be interesting to see what similar simulations on PI(4,5)P2 confined in a membrane will yield in terms of the position and interactions of Ca2+ and Mg2+. In conclusion, both approaches (31P NMR and QM/MM simulation) are complimentary in nature.

2+ Despite the reduced interaction of Mg , it may still play a role in PI(4,5)P2 signaling due to its high cytosolic concentration. Mg2+ concentrations are in the range of a few millimolar inside 124

the cell, and while Ca2+ has a much larger effect on ionization and clustering compared to Mg2+

2+ at the same concentration, the effect of 2 mM Mg on the 5-phosphate of PI(4,5)P2 is significantly greater than the effect of 0.1 mM Ca2+ on the same phosphate. Therefore it is possible that due to the large excess of Mg2+ it may be able to competitively inhibit Ca2+ binding in some situations. However, it is worth noting that Wang et al. did not find that millimolar Mg2+

2+ levels eliminated PI(4,5)P2 clustering in the presence of micromolar concentrations of Ca , although the number of such clusters was reduced (Wang, Collins et al. 2012).

Interactions at the 5-phosphate of PI(4,5)P2

Throughout our experiments we have consistently observed a greater impact on the 5- phosphate of PI(4,5)P2 as compared to the 4-phosphate. This was also observed in our previous work (Graber, Jiang et al. 2012). This indicates that the 5-phosphate may be more available or accessible for interactions with other molecules. There are several possible explanations for this observation. It is possible that the orientation of the PI(4,5)P2 inositol ring makes this phosphate more accessible for hydrogen-bond formation and other interactions. The 5-phosphate also carries a lower charge, which may make it more sensitive to these interactions, thus causing a larger change in chemical shift. Alternatively it may be that the 5-phosphate experiences weaker intramolecular hydrogen bonding. The free hydroxyl at the 6-position of the inositol ring interacts with the 5-phosphate through hydrogen bonding, but may also interact with the phosphodiester, thus providing a weaker hydrogen-bond to the 5-phosphate. In contrast, the hydroxyl at the 3-position can hydrogen-bond almost exclusively with the 4-phosphate, as the hydroxyl in the 2-position is axial (all others are equatorial). However, it should be noted that in

Slochower et al.’s simulations they found that the 2-position hydroxyl group could form a 125

hydrogen-bond with the 5-phosphate. If this occurs in the in vitro environment, then the 5- phosphate may actually experience stronger intramolecular hydrogen bond interactions (2 instead of 1) than the 4-phosphate. Based on our 31P MAS NMR results we suspect that this 2-OH hydrogen-bond with the 5-phosphate is not preserved under physiological conditions in a membrane environment. If this was true then we should have observed a more deprotonated state of the 5-phosphate. Ultimately the types and strengths of intramolecular hydrogen bonds with each of the two phosphomonoesters of PI(4,5)P2 will determine the degree of deprotonation

(keeping headgroup orientation, etc., the same). Our 31P MAS NMR results thus suggest which of these scenarios is more likely under conditions of an extended lipid bilayer in an aqueous environment at pH 7 and physiological salt concentration.

Effect of cholesterol and interaction with PE and PI

The 4- and 5-phosphates of PI(4,5)P2 are increasingly deshielded in the presence of increasing amounts of cholesterol. At physiologically relevant concentrations of cholesterol (30-40%) there is a small shift beyond the detection limits of the method in both the 4-, and 5-phosphate peak. In a related study (Jiang, Redfern et al. 2014), we have found that cholesterol derivatives with a modified hydroxyl failed to support PI(4,5)P2 domain formation. The hydroxyl group of cholesterol is a potential hydrogen-bond donor and may interact with the PI(4,5)P2 headgroup through hydrogen-bond formation. However, the PI(4,5)P2/cholesterol interaction affected to a much smaller extent the ionization state of the phosphomonoester groups than it was found for

PE, which is a common hydrogen-bond donor phospholipid. This suggests that the cholesterol hydroxyl group does not directly interact with the phosphomonoesters of PI(4,5)P2. An interaction of the cholesterol hydroxyl group and the PI(4,5)P2 phosphomonoester groups would 126

have been quite surprising. Even if the PI(4,5)P2 inositol ring is tilted back to the bilayer, such an interaction would require the cholesterol to be situated deep in the polar headgroup region, which is very unlikely considering the non-polar nature of the sterol ring structure. Instead of hydrogen–bonding directly to the PI(4,5)P2 phosphomonoester groups, the PI(4,5)P2 clustering effect induced by cholesterol is better explained by a model where cholesterol acts as a spacer between adjacent PI(4,5)P2 molecules. We have postulated previously that PI(4,5)P2 enriched domains might form through an extensive hydrogen bond network that involves intramolecular hydrogen bonds between the hydroxyl and phosphomonoester and phosphodiester groups as well as intermolecular hydrogen bonds between adjacent PI(4,5)P2 molecules. The formation of intramolecular hydrogen-bonds leads to a reduction of the charge density at the phosphomonoester groups by “smearing” out the charge throughout the inositol phosphate headgroup (in contrast to, for example, phosphatidic acid where the charge is fully localized at the phosphomonoester group). The formation of an intermolecular hydrogen-bond network, which is probably at least in part water mediated, leads to a further reduction of the local charge density by distributing the charge across the bilayer interface. The inclusion of cholesterol in such a molecular structure would increase the area per molecule for PI(4,5)P2, and thus further reduce the negative surface charge density of the lipid. A reduction in the negative surface charge density results in a locally reduced negative membrane potential, increased interfacial pH, and hence a slightly higher charge for PI(4,5)P2, exactly as we observe here. This explanation is also compatible with the macroscopic domain formation of PI(4,5)P2 induced by cholesterol as the electrostatic repulsion between two neighboring PI(4,5)P2 molecules would be reduced

(reducing the free energy of the system). The question arises; What makes cholesterol form a

127

domain together with PI(4,5)P2, since the interaction of the rigid sterol ring with the highly unsaturated arachidonoyl chain is generally unfavorable? At this point, we can only speculate about the answer to this question, however, the fact that the cholesterol hydroxyl group is important for the interaction with PI(4,5)P2 suggests that hydrogen bond formation – water mediated or direct – is relevant. The cholesterol hydroxyl group in a PC bilayer is typically situated approximately where the lipid carbonyl groups are found (Pitman, Suits et al. 2004). It might be that in the presence of PI(4,5)P2 cholesterol moves closer to the headgroup region, which would allow the cholesterol hydroxyl group to interact with the phosphodiester group or even the hydroxyl groups in the 2- or 6-position of the inositol ring. Energetically this might be favorable because the cholesterol molecules could fully participate in intermolecular hydrogen bond network described above.

The results for the ionization of PI(4,5)P2 in the mixtures with cholesterol and the hydrogen- bond donor lipids PE or PI are less intuitive. If PE forms hydrogen bonds with the phosphomonoester groups of PI(4,5)P2 while cholesterol binds elsewhere on the PI(4,5)P2 headgroup, then we might expect that the effects of these individual interactions would be cumulative. Instead we observe no significant increase of the ionization in the presence of cholesterol over what is observed when only PE is present. It is possible that at this concentration of PE, the presence of cholesterol does not alter the hydrogen bond network in a way that it would be reflected in a change of ionization at the phosphomonoester groups. This notion is further supported by our observation that at low pH the cholesterol induced downfield shift of the phosphomonoester groups is more pronounced (see Figure A9).

128

The mixture containing PI and cholesterol is also more complicated than it might have been expected. For the PC/PI/Cholesterol/PI(4,5)P2 system, the 4-phosphate is not shifted compared to the binary PC/PI(4,5)P2 control, while the 5-phosphate is shifted by roughly the same amount as it was observed in the presence of 40 mol% cholesterol without PI present. There might be some level of competition between PI and cholesterol, as the PI reduces the effect of cholesterol on the

4-phosphate. In the case of the 5-phosphate, it may be that the effect due to competition is compensated by hydrogen-bond formation from the PI.

While both cholesterol and the divalent cations were found to interact individually with

PI(4,5)P2 and promote PI(4,5)P2 clustering, the NMR data obtained in the presence of both, cholesterol and either of the divalent cations suggest an additive rather than a synergistic effect on PI(4,5)P2 ionization when both components are present. Thus, the cholesterol and the divalent cations interact independently with PI(4,5)P2. This is not unexpected, as cholesterol is likely to reside fairly low in the bilayer (Pitman, Suits et al. 2004), while the divalent cations approach from the aqueous buffer above the bilayer. Indeed, as discussed earlier, it is likely that cholesterol does not interact directly with the PI(4,5)P2 phosphomonoesters, and thus does not affect the interaction of the divalent cations with the phosphomonoesters.

2+ Effect of PE and PI on the PI(4,5)P2 interaction with Ca

The combination of PE and Ca2+ lead to a cumulative effect on the overall ionization of

2+ PI(4,5)P2. This result is quite similar to the cholesterol/Ca combination. Whereas cholesterol forms in all likelihood a hydrogen bond with the PI(4,5)P2 headgroup at a location somewhat deeper in the membrane, Ca2+ interacts with the headgroup from solution. In contrast, PE is well

129

known to form hydrogen-bonds with the PI(4,5)P2 headgroup, most likely through the 5- phosphate (see chapter 4). This interaction has a much higher chance of interfering with the

2+ Ca /PI(4,5)P2 interaction than the cholesterol hydrogen-bond. This indicates that the PE-

2+ PI(4,5)P2 interaction either does not interfere with the Ca /PI(4,5)P2 interaction, or its interference is offset by the increased charge on PI(4,5)P2 due to the hydrogen bond formation.

Based on our data we cannot say which of these scenarios is more likely to be correct. The combined effect of these interactions does lead to a quite large downfield shift for the PI(4,5)P2 phosphomonoester peak. Based on this shift, we can calculate an increase in the overall charge for PI(4,5)P2 from -4.05 to -4.24. This indicates how the combined effect of membrane components could lead to a fairly large overall change in the charge of PI(4,5)P2.

PI, like PE, can form hydrogen bonds with PI(4,5)P2 (see chapter 4). Unlike PE, PI also carries a negative charge, and the competing effect of the charge and hydrogen bond formation means that PI has very little effect on the ionization of PI(4,5)P2 at pH 7. However, when we add

2+ 2+ Ca in the presence of PI, we see a larger effect than when we add Ca to PC:PI(4,5)P2 vesicles.

2+ Thus, PI actually seems to enhance the binding of Ca to PI(4,5)P2. What could cause this enhanced binding? One possibility is that local Ca2+ concentrations are increased at the membrane due to the increased negative potential of the overall membrane due to the presence of negatively charged PI. However, we found previously that increased PI concentration (leading to negative membrane potential) had no effect on the ionization of PI(4,5)P2 (see chapter 4 and the appendix). We explained this by suggesting that PI and PI(4,5)P2 localize in domains, and therefore PI(4,5)P2 already resides in a PI rich environment and increased global membrane potential (i.e. increased PI concentration) will have no further effect on PI(4,5)P2 ionization. 130

Instead, the effect of PI may be based primarily on the domain formation induced by PI.

2+ Furthermore, PI itself may experience clustering due to the presence of Ca . In a PC/PI(4,5)P2 membrane, PI(4,5)P2 will be more scattered, although there may still be nanoscale PI(4,5)P2 domains forming (Redfern and Gericke 2004; Kooijman, King et al. 2009). With PI(4,5)P2

2+ scattered throughout the membrane, it may be more difficult for Ca to find PI(4,5)P2 molecules to bind. When PI is present, PI(4,5)P2 will cluster into large PI(4,5)P2 rich domains. These domains will be characterized by a large negative potential field. These domains will therefore

2+ provide a platform for Ca to bind to, with multiple PI(4,5)P2 molecules in close proximity so

2+ the Ca could quickly bind to nearby PI(4,5)P2 molecules after it dissociates with a given

2+ PI(4,5)P2 molecule. The enhanced Ca binding in the presence of PI has important implications for the role of PI in PI(4,5)P2 signaling as well as the role of PI(4,5)P2 rich domains in general.

Conclusions

2+ Ca and cholesterol have a potentially significant role in PI(4,5)P2 signaling due to their

2+ ability to promote local accumulation of PI(4,5)P2. In the case of Ca the local accumulation observed experimentally is in terms of nano-scale domains, and in the case of cholesterol it is in the form of macroscopic domain formation (Dasgupta, Bamba et al. 2009; Wang, Collins et al.

2012). Our previous work suggests that PI(4,5)P2 forms nanoclusters in mixed PC/PI(4,5)P2 vesicles even without the presence of Ca2+ or cholesterol (Kooijman, King et al. 2009). This conclusion was based on the observation that PI(4,5)P2 adopts a higher negative charge in lipid bilayers than it was found for the soluble analog of the PI(4,5)P2 headgroup, inositol-4,5- bisphosphate. Due to the accumulation of negative charges in a lipid bilayer with anionic lipids, the interfacial pH is lower than the bulk pH and therefore, a higher degree of protonation and 131

hence a lower charge was expected for the membrane resident lipid. Since the opposite was observed, we concluded that intermolecular hydrogen bonds lead to a larger than expected deprotonation from the PI(4,5)P2 phosphomonoester groups. However, the balance of repulsive forces due to the negative headgroup charge and attractive forces due to hydrogen-bond formation, likely leads to very fragile domains that can easily disintegrate. While this observation is of significance because it suggest that the mutual repulsive forces between

PI(4,5)P2 molecules are smaller than expected, formation of pure PI(4,5)P2 domains is quite unlikely in real biological membranes. However, PI, Ca2+ and cholesterol are able to stabilize

PI(4,5)P2 enriched domains, which has potentially far reaching implications for phosphoinositide mediated signaling events. The results from this study add to our understanding of how these domains may form.

2+ We observed a strong interaction between Ca and PI(4,5)P2, which leads to a quite pronounced change in PI(4,5)P2’s ionization state (see Table 5-1). On the other hand, our experiments in the presence of Mg2+ show that the interaction in terms of changes in ionization of the 4- and 5-phosphate is considerably less than what is observed in the presence of Ca2+, consistent with the work of Wang et al. and Slochower et al. (Wang, Collins et al. 2012;

Slochower, Huwe et al. 2013). In contrast to the recent simulation studies by Slochower et al., we found that both divalent cations have a larger effect on the 5-phosphate ionization than on the

4-phosphate (Slochower, Huwe et al. 2013). It is striking that we observe clear differential

2+ 2+ effects of Ca and Mg on the ionization properties of PI(4,5)P2 which lead to significant differences in the clustering behavior of PI(4,5)P2 due the presence of these ions.

132

Our results concerning the effect of cholesterol on the ionization state of PI(4,5)P2 are consistent with other experimental observations, namely cholesterol induced PI(4,5)P2 macroscopic domain formation. The increase in negative charge induced by cholesterol is small and not consistent with a direct interaction of the cholesterol hydroxyl group with the phosphomonoester groups of PI(4,5)P2. Instead, cholesterol is likely to act as a spacer between

PI(4,5)P2 molecules in the membrane. Apparently the cholesterol hydroxyl group participates in the hydrogen-bond network that is formed between the PI(4,5)P2 molecules, which would lead to a stabilization of these cholesterol containing PI(4,5)P2 enriched domains. The inclusion of cholesterol in the PI(4,5)P2 domains leads to an increase in the PI(4,5)P2 area per molecule and hence decreases the negative surface potential of these clusters. The reduced negative membrane potential in turn increases the interfacial pH, which leads to enhanced deprotonation of the phosphomonoester groups and hence slightly increases their negative charge. The effect of cholesterol is modulated by other common membrane lipids, namely PE and PI. The addition of cholesterol to PC/PE/PI(4,5)P2 mixed vesicles did not alter the ionization state of the PI(4,5)P2 phosphomonoester groups, suggesting that either cholesterol is excluded in this case from the

PE/PI(4,5)P2 hydrogen-bond network or does not alter it to the extent that it would be reflected in the ionization state of the phosphomonoester groups. Based upon our measurements of

PC/PE/PI(4,5)P2/cholesterol vesicles at low pH (see Figure A9), the latter explanation appears to be more likely.

In the case of PI the impact of cholesterol on the PC/PI/PI(4,5)P2 system is more subtle as there is no combined effect of cholesterol and PI on the degree of ionization of PI(4,5)P2. Even though both cholesterol and PI promote macroscopic domain formation of PI(4,5)P2 in model PC 133

membranes, likely by acting as a spacer between PI(4,5)P2 molecules, the additional negative charge of PI and its ability to form hydrogen-bonds with the PI(4,5)P2 phosphomonoester groups compete in terms of their effect on PI(4,5)P2 charge. Hence, the observation that there is no additional effect on the negative charge of PI(4,5)P2, at pH 7.2, in the presence of both cholesterol and PI. The effect of the bivalent cations Mg2+ and Ca2+ and cholesterol on the ionization state of PI(4,5)P2 were found to be additive but not synergistic.

Our results also suggest that the 5-phosphate may be more important for interactions of

PI(4,5)P2 with bivalent cations or other lipid species present in the bilayer. Its location on the inositol ring appears to be uniquely suited to making it highly sensitive to interactions with other molecules.

134

Chapter 6

Competitive cation binding to Phosphatidylinositol-4,5-bisphosphate revealed by X-ray

fluorescence

Introduction

Calcium has been found to interact strongly with PI(4,5)P2. Studies of PI(4,5)P2 in model membranes have found that calcium binds to PI(4,5)P2 and causes PI(4,5)P2 cluster formation

(Wang, Collins et al. 2012; Sarmento, Coutinho et al. 2013). If these clusters exist in vivo, they could have an important role in PI(4,5)P2 signaling. Clustering was observed indirectly via AFM imaging (of the transferred monolayer), as well as through condensation of the PI(4,5)P2 monolayer upon addition of calcium (Wang, Collins et al. 2012). In contrast, limited cluster

2+ formation was observed for Mg , which interacts more weakly with PI(4,5)P2 (Wang, Collins et al. 2012). In addition, when both calcium and magnesium were present, calcium easily out competed magnesium (Wang, Collins et al. 2012).

Calcium flux is commonly used by the cell to trigger signaling events (Bootman, Lipp et al.

2001), and may affect PI(4,5)P2 based signaling. In order to model the crowded cation environment of the cytosol, we examined the competition between K+, Mg2+, and Ca2+ for

PI(4,5)P2 binding sites.

135

The use of single lipid monolayers of natural, highly unsaturated PI(4,5)P2 is important as it allows direct evaluation of cation-PI(4,5)P2 interactions which is not possible for mixed lipid monolayers where non-specific ion binding is a problem. Additionally, the formation of pure

PI(4,5)P2 domains in model lipid membranes has been well established and highlights that such domains may exist in living cells (Kooijman, King et al. 2009; Wang, Collins et al. 2012;

Sarmento, Coutinho et al. 2013; Jiang, Redfern et al. 2014; Salvemini, Gau et al. 2014) , especially in regions of active PI(4,5)P2 synthesis and where PI(4,5)P2 clusters exist because of other protein mediated phosphoinositide domains. We used a combination of both x-ray fluorescence and reflectivity to directly characterize cation binding by PI(4,5)P2. Others have examined PI(4,5)P2 via x-ray reflectivity, however, it is unclear which PI(4,5)P2 species was used, and more importantly these studies used DPPC, a phospholipid not found in the native plasma membrane inner leaflet (Ghosh, Castorph et al. 2012).

For the first time we have used x-ray techniques to measure cation binding to a natural, highly unsaturated, PI(4,5)P2 monolayer. We find an eight-fold increase in the interfacial concentration of K+ in the absence of divalent cations. At physiological concentrations of 1 µM Ca2+ we observe a 60 fold increase in interfacial Ca2+. Unexpectedly, there is still a four-fold surface increase of [K+] in the presence of 1 mM Ca2+ and Mg2+ in the bulk solution indicating that K+ is likely important for the binding of PI(4,5)P2 binding partners.

136

Results

Calcium reorients PI(4,5)P2 in model lipid membranes

X-ray reflectivity from highly unsaturated, natural, PI(4,5)P2 at physiologically relevant subphase conditions in the absence and presence of Ca2+ shows that Ca2+ causes significant changes in the structure of the PI(4,5)P2 model membrane (Figure 6-1A). Figure 6-1B shows the corresponding electron (scattering) density across the lipid layer (see methods). The main effect of increased Ca2+ concentration is an increase in the headgroup electron density and thickness of

2+ Figure 6-1. X-ray reflectivity of the PI(4,5)P monolayer in the presence of varying [Ca ]. (a) X- 2 ray reflectivity measured from a PI(4,5)P monolayer spread on a pH 7.2 subphase containing 15 mM 2 Tris, 100 mM KCl, and varying amounts of CaCl (from 0-1,000 µM). The monolayer was held at a 2 constant surface pressure of 30 ± 2 mN/m. (b) Electron density versus distance from the PI(4,5)P 2 headgroup plotted with respect to CaCl concentration. Electron density values are calculated from a 2 2-box fitting of the normalized reflectivity.

137

Table 6-1. PI(4,5)P2 monolayer structural parameters derived from a 2 box fit to the reflectivity curves from Figure 6-1A

Subphase 1 µM 10 µM 100 µM 1 mM

(CaCl2)

lH (Å) 9.2(9) 7.6(4) 7.2(4) 8.2(6) 3 ρH(e/Å ) 0.56(2) 0.65(2) 0.72(2) 0.70(3)

lT (Å) 14.7(4) 16.2(2) 18.1(2) 18.4(3) 3 ρT(e/Å ) 0.321(4) 0.322(2) 0.322(2) 0.307(4)

lH + lT (Å) 24.0(5) 23.8(2) 25.2(2) 26.7(3) σ (Å) 4.3(2) 4.6(1) 4.7(1) 4.5(1) 2 Amol (Å ) 58-62 54-55 48-49 47-49

Structural parameters derived from the reflectivity curves for a PI(4,5)P2 monolayer spread on a pH 7.2 subphase containing 15 mM Tris, 100 mM KCl, as well as 1 µM, 10 µM, 100 µM, or 1 mM CaCl2 3 respectively. lH (Å) describes the length of the headgroup region, ρH (e/Å ) is the electron density of the 3 headgroup region, lT (Å) is the length of the acyl chain region, ρT(e/Å ) is the electron density of the acyl 2 chain region, σ (Å) is the surface roughness, and Amol (Å ) is the area per molecule in the monolayer.

the acyl-chain region (see table 6-1) as evidenced by a shift of the first minimum of the R/RF curves (Kjaer 1994). The increased thickness is likely due to an increase in the relative order of the PI(4,5)P2 acyl chains and a subsequent condensation of the PI(4,5)P2 monolayer in agreement with the condensing effect of divalent cations on PI(4,5)P2 (Levental, Cebers et al. 2008;

Levental, Christian et al. 2009; Wang, Collins et al. 2012). The increased order and decreased molecular area of PI(4,5)P2 can be calculated by assuming that an ordered PI(4,5)P2 molecule

(e.g. gel phase) has an acyl-chain length corresponding to the stearic acid tail at the sn-1 position of natural PI(4,5)P2. Namely, lTmax, can be estimated as ~ 22 Å since each carbon (17 in stearic acid, excluding the carbonyl) contributes ~1.27 Å (Small 1986; Kjaer, Alsnielsen et al. 1989).

The molecular area of the lipid Amol, can be determined based on the maximum (ordered) and the 138

actual acyl-chain length using 2A0 lT,max / lT, where A0 is the minimum area available to a single, untilted alkyl chain (~20 Å, according to measurements with arachidic acid) (Small 1986; Kjaer,

Alsnielsen et al. 1989). Here we assume that the arachidonoyl chain of PI(4,5)P2 contributes an equal area to the overall area of each PI(4,5)P2. Although the organization of arachidonoyl chains is largely unknown, this is a reasonable assumption and leads to a lower bound of the molecular area of PI(4,5)P2 (see table 6-1). The molecular area values were also consistent with the values calculated based on the total area of the monolayer and the amount of PI(4,5)P2 that was spread.

Cells contain up to ~1 mM free Mg2+ and only micromolar concentrations of Ca2+. During

Ca2+ signaling the local concentration of Ca2+ increases dramatically and can potentially reach millimolar concentrations in the vicinity of Ca2+ channels. Figure 6-2A shows x-ray reflectivity from natural, highly unsaturated, PI(4,5)P2 monolayers on buffer subphases containing 1 mM

Mg2+ as a function of [Ca2+]. The effect of 10 µM Ca2+ is negligible but increases dramatically for 0.1 and 1 mM Ca2+, as may occur during Ca2+ signaling. Figure 6-2B shows the electron

(scattering) density profiles for the reflectivity curves of Figure 6-2A (see methods). As in the

2+ 2+ absence of Mg , increases in [Ca ] result in a shift to lower Qz values of the first minimum in

2+ 2+ the R/RF curves. This indicates that even in the presence of 1 mM Mg , Ca is able to further increase the thickness of the PI(4,5)P2 model membrane and increase the headgroup density (see

Figure 6-2B and Table 6-2).

139

140

2+ Figure 6-2. X-ray reflectivity of the PI(4,5)P2 monolayer in the presence of 1 mM Mg and 2+ varying [Ca ]. (a) X-ray reflectivity measured from a PI(4,5)P2 monolayer spread on a pH 7.2

subphase containing 15 mM Tris, 100 mM KCl, 1 mM MgCl2, and varying amounts of CaCl2 (from 0- 1,000 µM). The monolayer was held at a constant surface pressure of 30 ± 2 mN/m (b) Electron

density versus distance from the PI(4,5)P2 headgroup plotted with respect to CaCl2 concentration. Electron density values are calculated from a 2-box fitting of the normalized reflectivity.

Table 6-2. PI(4,5)P monolayer structural parameters derived from 2 a 2 box fit to the reflectivity curves from Figure 6-2A. 2+ Figure 6-2. X-ray reflectivity of the PI(4,5)P2 monolayer in the presence of 1 mM Mg and 2+ 2+ 2+ 2+ 2+ 2+ Mgvarying 1 mM [Ca ].0 (a)µM X Ca-ray reflect10ivityµM measuredCa 100 fromµM a CaPI(4,5)P 1mM2 monolayer Ca spread on a pH 7.2 subphase containing 15mM Tris, 100mM KCl, 1 mM MgCl , and varying amounts of CaCl (from 0- l 10(1) 9(1) 9(2) 2 10.0(8) 2 1000H (Å) µM). The monolayer was held at a constant surface pressure of 30 ± 2 mN/m (b) Electron 3 densityρH(e/Å ) versus distance0.57(4) from the PI(4,5)P0.60(4) 2 headgroup0.63(5) plotted with0.63(3) respect to CaCl2 concentration. Electron density values are calculated from a 2-box fitting of the normalized reflectivity. lT (Å) 15.4(6) 16.0(6) 16.9(8) 17.0(4) 3 ρT(e/Å ) 0.303(7) 0.308(8) 0.30(1) 0.300(5)

lH + lT (Å) 25.4(6) 25.3(6) 25.8(9) 27.0(4) σ (Å) 4.2(2) 4.3(2) 4.3(3) 4.3(2) 2 Amol (Å ) 55-64 53-57 50-55 51-58

Structural parameters derived from the reflectivity curves for a PI(4,5)P2 monolayer spread on a pH 7.2 subphase containing 15 mM Tris, 100 mM KCl, 1 mM MgCl2, as well as 0 µM, 10 µM, 100 µM, or 1 mM CaCl2 respectively.

2+ 2+ To clearly show the effect of both Mg and Ca on the structure of the PI(4,5)P2 model membrane the reflectivity for PI(4,5)P2 on a subphase without divalent cations, a subphase with only Ca2+, a subphase with only Mg2+, and a subphase with Mg2+ and Ca2+ together is compared in Figure 6-2C. Figure 6-2D shows the corresponding electron density and Table 6-3 the fitting parameters.

141

Table 6-3. PI(4,5)P2 monolayer structural parameters derived from a 2 box fit to the reflectivity curves from Figure 6-2C

Subphase Buffer MgCl2 CaCl2 MgCl2/CaCl2

lH (Å) 10.3 (7) 10 (1) 8.2(6) 10.0(8) 3 ρH(e/Å ) 0.51(1) 0.57(4) 0.70(3) 0.63(3)

lT (Å) 13.4(3) 15.4(6) 18.4(3) 17.0(4) 3 ρT(e/Å ) 0.323(3) 0.303(7) 0.307(4) 0.300(5)

lH + lT (Å) 23.6(4) 25.4(6) 26.7(3) 27.0(4) σ (Å) 4.2(1) 4.2(2) 4.5(1) 4.3(2) 2 Amol (Å ) 64 – 69 55 – 64 47 – 52 51 – 58 M2+/PI(4,5)P2 n/a 2.0 ± 1.8 1.6 ± 0.9 n/a

Structural parameters derived from the reflectivity curves for a PI(4,5)P2 monolayer spread on a pH 7.2 subphase containing 15 mM Tris, 100 mM KCl, as well as no divalent cations, 1 mM

MgCl2, 1 mM CaCl2, or 1 mM MgCl2 and CaCl2 respectively.

The reflectivity data allows for an estimation of the number of divalent cations (M2+) bound to the headgroup of PI(4,5)P2, based on a space-filling model (Bu, Flores et al. 2009; Vaknin

2012; Wang, Anderson et al. 2012). The model assumes the headgroup is the sum of volumes of

+ 2+ the K ions, M ions, water molecules, and the PI(4,5)P2 head group. The conservation of electron numbers and constrains on the total volume contained in the two boxes can be expressed as follows:

142

Amol lH =VH,dry + 30nH2O +nK+VK+ + nM2+VM2+, (6-1)

e e AmolρHlH = N H,dry + 10nH2O +18nK+ + n M2+N M2+, (6-2)

+ 2+ Assuming the number of K per PI(4,5)P2 is unchanged prior to and after M binding, we obtain

2+ the nM2+, the number of M per PI(4,5)P2 molecule (see Table 6-3) (Wang, Anderson et al.

2012). These binding numbers serve as lower limits of actual binding number of Ca2+ and Mg2+ per PI(4,5)P2.

X-Ray Fluorescence

A more direct way to determine cation binding to PI(4,5)P2 model membranes is x-ray fluorescence. Figure 6-3A shows x-ray fluorescence data (Qz < 0.0217) for a pure (no PI(4,5)P2 present) 100 mM KCl, buffer interface (black squares) compared to the same interface in the presence of PI(4,5)P2 (red squares). The high negative charge of PI(4,5)P2 increases the surface

+ concentration of K eight fold (bulk 0.1 M, vs. PI(4,5)P2 headgroup interface 0.8 M). As we increase the [Ca2+] in the bulk, the surface [Ca2+] increases dramatically (green, blue, and orange data points). The increase in [Ca2+] coincides with a decrease in [K+], as Ca2+ displaces K+ from the headgroup interface which indicates a much higher binding constant for Ca2+ compared to K+ as expected.

143

A B

Figure 6-3. X-ray fluorescence reveals cation binding to the PI(4,5)P monolayer. (a) X-ray 2 + 2+ fluorescence from surface K and Ca ions with either no monolayer (black) or a PI(4,5)P monolayer 2 (red) spread on a pH 7.2 subphase containing 15 mM Tris, 100 mM KCl, and varying amounts of CaCl 2 (from 0-1000 µM). The monolayer was held at a constant surface pressure of 30 ± 2 mN/m. The -1 fluorescence was integrated for Q values from 0.010-0.021 Å , which results in a penetration depth of z ~50-100 Å. Fluorescence in the presence of 1 µM (green), 0.1 mM (blue), and 1 mM (orange) calcium 2+ are shown. (b) Fluorescence measurements in the presence of additional 1 mM Mg (dark green). Fluorescence in the presence of 10 µM (light green), 0.1 mM (blue), and 1 mM (orange) calcium are shown. The cytoplasm also contains roughly 0.8-1 mM free magnesium ions (Lodish 2013). In order to investigate the effect of magnesium on PI(4,5)P2 ion binding we investigated a buffer

2+ subphase containing 1 mM MgCl2 and varied [Ca ]. Figure 6-3B compares the surface fluorescence (Qz < 0.0217) for a PI(4,5)P2 monolayer with 1 mM MgCl2 and varying amounts of calcium with that of the plain subphase with no monolayer. With 1 mM Mg2+ we observe a significant decrease in the interfacial [K+] (dark green triangles). As increasing amounts of Ca2+ are added (light green diamonds and light blue triangles) the interfacial [Ca2+] increases. The

144

Ca2+ fluorescence in the presence of Mg2+ is compared against the case without Mg2+ (dark blue

2+ 2+ triangles). The Mg ions cannot be observed directly as the Kα and Kβ emission lines of Mg are too low in energy to be excited by the 8 keV beam. However, Mg2+ binding can be observed indirectly based on the displaced potassium ions (compare 0 mM Mg2+ (red data) with 1 mM

2+ 2+ Mg (dark green data) in Figure 6-3B). In the presence of 1 mM Mg , the potassium Kα peak is reduced to roughly equivalent levels as we observe for the subphase containing 1 mM Ca2+ without Mg2+ (compare dark green with blue), indicating that Mg2+ binds to a similar degree as

Ca2+. This is consistent with estimated binding constants for Ca2+ and Mg2+ which are of the same order of magnitude (Hendrickson and Ballou 1964; Dawson 1965; Hauser and Dawson

1967; Buckley and Hawthorne 1972; Toner, Vaio et al. 1988). Addition of 0.1 mM Ca2+ to the subphase, leads to the observation of bound calcium. The calcium fluorescence increases at 1 mM Ca2+ as more calcium binds (cyan data vs green data). Interestingly, the [K+] is unchanged by the addition of Ca2+, indicating that Ca2+ replaces Mg2+ bound at the surface instead of K+.

Interfacial [Ca2+] is reduced compared to the solution without Mg2+. This shows that Mg2+

2+ reduces Ca binding and thus stays at the PI(4,5)P2 headgroup interface.

The interfacial ion density is determined from equation 6-3:

ns = Is(αi)/Ib(αi) D(αi) nb, (6-3)

where nb is the bulk concentration of the ion, D(αi) is the penetration depth for the incident beam

(~5-10 nm), Ib(αi) is the fluorescence intensity of the solution without the monolayer, and Is(αi) is

145

the fluorescence intensity of the solution without the monolayer after the subtraction of the bulk contribution Ib(αi). By multiplying the ion density by the area per lipid molecule (determined from the applied area per molecule) we obtain the number of ions per lipid. Figure 6-4A and B shows the observed ion density and ions / lipid with respect to calcium concentration, with and without 1 mM Mg2+, respectively. When no calcium is present the K+ density is high, with

+ 2+ + roughly 1.7 ± 0.2 K ions associate with each PI(4,5)P2 molecule. As the [Ca ] increases, the K density drops off as K+ is replaced by Ca2+, while the Ca2+ ion density increases dramatically. At

2+ + 1 mM Ca the potassium levels have dropped to around 0.5 ± 0.3 K ions per PI(4,5)P2

A B

+ 2+ Figure 6-4. Calculated ion density for K and Ca at the PI(4,5)P monolayer. (a) Ion density 2 2+ + 2+ + plotted vs [Ca ] for K and Ca ions at the surface of the PI(4,5)P monolayer. Ion density of K and 2 2+ Ca is compared for samples in the presence (blue and magenta respectively) and absence (black and 2+ 2+ + 2+ red) of 1 mM Mg . (b) Ions per lipid plotted vs. [Ca ] for K and Ca . Lines are meant to guide the eye.

146

2+ molecule, while there are 1.6 ± 0.3 Ca ions per PI(4,5)P2. Magnesium ion density could not be directly determined. However, the presence of 1mM Mg2+ leads to a reduction in K+ and more significantly Ca2+ ion density due to competition. In the absence of Ca2+, 1 mM Mg2+ reduces the

K+ density to similar levels as observed for 1 mM Ca2+ without Mg2+. As Ca2+ is added to the

Mg2+/K+ subphase, the K+ ion density stays nearly constant. Surprisingly, even with 1 mM Mg2+

2+ + 2+ and Ca , there is still K present at the PI(4,5)P2 interface. Additionally, [Ca ] is also reduced by the presence of Mg2+. The observation that [K+] stays increased, despite the presence of

2+ 2+ + divalent Mg and Ca suggests that K plays an important role in PI(4,5)P2 signaling as its binding to amino acids could lead to changes in protein structure and thus likely function.

Discussion

The results of our x-ray experiments indicate the complexity of interactions between common cellular cations and PI(4,5)P2. The many ion species within the cell must compete for the charged PI(4,5)P2 headgroups. Cation competition is likely important for controlling PI(4,5)P2 signaling.

As may be expected based on simple electrostatics, we see significant surface enhancement of potassium in the presence of PI(4,5)P2 as the potassium ions are drawn to the negatively charged

PI(4,5)P2. At pH 7.2, PI(4,5)P2 has a single remaining exchangeable proton and carries a charge of -4 (Kooijman, King et al. 2009). Our experiment shows that 1-2 K+ ions accumulate around each PI(4,5)P2 headgroup and the net charge is -2 or -3. These potassium levels are consistent with what previous studies have suggested, although our number is a bit higher (Toner, Vaio et al. 1988). The overall potassium ion concentration increases from 0.1 M in the bulk to a striking 147

+ 0.8 M at the interface. Our work shows that the K -PI(4,5)P2 interaction isn’t strong enough to bring sufficient potassium to the interface to completely neutralize the PI(4,5)P2 charge.

However, calcium has a much stronger interaction with PI(4,5)P2 due to its higher charge density. Even with calcium concentrations that are 102-105 fold lower than the potassium concentrations, the calcium still binds to PI(4,5)P2. As the calcium concentration increases, calcium replaces the potassium ions at the interface. At 1 mM CaCl2, the calcium ions actually outnumber potassium at the interface, despite the much lower calcium concentration (~1.6 Ca2+ to ~0.5 K+). Thus, the binding constant for calcium is at least 100 times greater than the potassium binding constant. With 1 mM CaCl2 and 100 mM KCl, PI(4,5)P2’s charge is nearly completely neutralized. The original charge of -4 (according to NMR measurements (Kooijman,

King et al. 2009)) is reduced to -0.5 with 1.6 Ca2+ and 0.5 K+ ions bound. However, this charge will also be increased slightly as the calcium binding induces further deprotonation due to increased positive surface potential (an increase of roughly -0.25 (see chapter 5, (Graber, Gericke et al. 2014))).

Magnesium further complicates the situation. When just magnesium and potassium are present (1 mM Mg2+, 100 mM KCl), the divalent cation again heavily out-competes the potassium. While the magnesium ions cannot be directly counted, the potassium ions are reduced

2+ + to nearly the same level as in the presence of 1 mM Ca (0.75 K ions per PI(4,5)P2). We can estimate the number of magnesium ions based on our electron density profile as well as the change in ion density of the potassium and calcium ions. Based on the electron density profile, there are two magnesium ions per PI(4,5)P2. By assuming one magnesium ion for each displaced potassium and calcium ion we can also estimate the magnesium ion density from the decrease in 148

the potassium and calcium ion density. From this we calculate roughly 1 magnesium ion per

PI(4,5)P2 for 1 mM MgCl2 and 100 mM KCl. As the calcium concentration increases we observe very little change in the potassium concentration and we can therefore assume that calcium is mostly replacing magnesium ions (see figure 6-4A in blue). When the buffer subphase has 1 mM

CaCl2, 1 mM MgCl2, and 100 mM KCl, the potassium density is actually slightly higher in the presence of magnesium, while the calcium density is slightly lower. However, in the absence of magnesium the monolayer is somewhat more condensed, and when this is accounted for there appears to be almost no change to the number of calcium and potassium ions. Even though one has to keep in mind that our analysis of the interfacial [Mg2+] is indirect, this appears to show that at a 1:1 ratio the calcium almost completely replaces the magnesium. This is quite surprising given that there is still potassium present and the magnesium clearly binds much more strongly than the potassium. The last potassium ion may be bound quite tightly, despite its lower charge density, and may have a more selective size specific interaction with PI(4,5)P2.

In the cell cytosol, potassium ion concentration is between 100 and 150 mM. According to our

+ results this would indicate that 1.7 K ions bind to each PI(4,5)P2 molecule in the cell plasma membrane inner leaflet. While one potassium ion binds quite loosely and probably remains

+ hydrated, another 0.5 K ions/lipid remains closely associated with the PI(4,5)P2 headgroup despite the presence of Mg2+ and Ca2+ at millimolar levels. This suggests that the second K+ ion may be shared between two PI(4,5)P2 lipids and binds fairly strongly. The cytosol also commonly contains roughly 1 mM of free Mg2+ (Lodish 2013). At this concentration roughly one magnesium ion binds to each PI(4,5)P2 lipid, replacing a single potassium ion. MD simulations suggest that this magnesium binds to PI(4,5)P2 through a water mediated, and 149

therefore weaker, interaction and remains hydrated (Slochower, Huwe et al. 2013). The effective charge of PI(4,5)P2 will therefore decrease from -2.5 to -1.5. In a resting cell, calcium concentrations are quite low and only reach the micromolar range. According to our data, at these concentrations some calcium binds to PI(4,5)P2 at a ratio of roughly 0.1 ions/PI(4,5)P2.

Even with these reduced levels of calcium we see an effect on the lipid packing with the presence of calcium leading to some condensation of the PI(4,5)P2 monolayer. In the presence of magnesium this effect is reduced. Previous studies have also suggested that 1 µM Ca2+ is enough to cause formation of small PI(4,5)P2 clusters (Wang, Collins et al. 2012; Sarmento, Coutinho et al. 2013). However, these studies were done in the absence of K+ ions, which may reduce the

2+ impact of 1 µM Ca and prevent the formation of calcium induced PI(4,5)P2 clusters. We find

2+ 2+ that in the presence of 100 mM KCl there are only 0.1 Ca ions/PI(4,5)P2 at micromolar [Ca ], which may not be enough to induce clustering. However, during calcium signaling events, large amounts of calcium are released via calcium channels, and calcium concentrations reach 10-100

µM (Bootman, Lipp et al. 2001). It is possible that near the calcium channels local concentrations may even reach millimolar concentrations. This will trigger large amounts of

2+ calcium binding to PI(4,5)P2 to a peak of around 1.6 Ca ions per lipid and could cause calcium induced clustering. The first calcium ion is likely to bind to the phosphomonoesters of PI(4,5)P2.

While the 4-phosphate could be the preferred target, due to its slightly larger charge (Kooijman,

King et al. 2009; Slochower, Huwe et al. 2013), our NMR studies show a larger effect of Ca2+ on the 5-phosphate. This may possible be due to the orientation of the PI(4,5)P2 headgroup. The second calcium ion might be shared between two PI(4,5)P2 molecules, acting as a bridge to promote PI(4,5)P2 clustering. Thus, varying calcium concentrations may have a significant effect

150

on PI(4,5)P2 mediated signaling events. The presence of these calcium ions virtually neutralizes the effective charge of PI(4,5)P2 and can also lead to PI(4,5)P2 clustering (Wang, Collins et al.

2012; Sarmento, Coutinho et al. 2013). Many proteins that bind to PI(4,5)P2 rely at least partially

2+ on electrostatic interactions to dock to PI(4,5)P2, and the presence of these Ca ions may impact binding between some proteins and PI(4,5)P2.

Conclusions

For the first time we successfully used surface sensitive (synchrotron) x-ray techniques to measure the structure of naturally, highly unsaturated, PI(4,5)P2 model membranes under physiologically relevant conditions, specifically, the interaction of this model membrane with the

+ 2+ 2+ + cations K , Ca , and Mg . The monovalent cation, K , is enriched eight-fold at the PI(4,5)P2 lipid headgroup interface. The addition of Ca2+ at physiologically relevant concentrations results in significant changes to the organization (thickness) of PI(4,5)P2 in the model membrane. As

Ca2+ binds it replaces K+ at the headgroup interface so that at 1 mM bulk Ca2+ there are 1.6 Ca2+

2+ 2+ bound to each PI(4,5)P2 (to a 1 M surface concentration of Ca ). Mg also binds to PI(4,5)P2 headgroups but affects the organization of the PI(4,5)P2 model membrane to a lesser degree, consistent with our own and others observations (Levental, Cebers et al. 2008; Wang, Collins et al. 2012; Graber, Gericke et al. 2014). When both Mg2+ and Ca2+ are present in the solution, Ca2+ successfully competes with Mg2+ for binding sites, and appears to replace most of the bound

2+ + 2+ Mg ions. Surprisingly, 0.5 K ions/PI(4,5)P2 remain despite the presence of 1mM Ca and 1 mM Mg2+, indicating a tightly associated potassium ion (K+ specific binding site). Increased cellular calcium flux during calcium signaling is likely to have a significant impact on PI(4,5)P2

151

signaling affecting which and how many ions bind to the PI(4,5)P2 enriched membrane and thus affecting the affinity of PI(4,5)P2 binding proteins.

152

Chapter 7

NMR investigation of the highly specific PTEN/PI(4,5)P2 interaction

Introduction

Throughout this work I have investigated cation-phosphoinositide interactions primarily based on electrostatics and the ionization properties of PI(4,5)P2. In this chapter I investigate the

PTEN/PI(4,5)P2 interaction, an interaction that is partially based on electrostatics, but is highly specific for PI(4,5)P2 (Redfern, Redfern et al. 2008). As discussed in the introduction, the

PTEN/PI(4,5)P2 interaction is critical for activating PTEN, which in turn regulates the

PI3K/AKT signaling pathway (Campbell, Liu et al. 2003). The N-terminal sequence of PTEN, where the PI(4,5)P2 binding site is located, is occupied with multiple cationic residues. Within the region of residues 10-16, which is known to be critical for the interaction (Redfern, Redfern et al. 2008), there are four cationic residues (three arginines and one lysine). Overall there is a net charge of +2 in the first 21 amino acids. The highly negatively charged phosphoinositides can interact electrostatically with this positively charged binding region of PTEN. However,

PI(4,5)P2 binds much more strongly than PI(3,4)P2, even though PI(3,4)P2 has a similar charge to

PI(4,5)P2 (-3.96 vs. -4.04) (Redfern, Redfern et al. 2008). Thus the PTEN N-terminus is specifically structured to accept PI(4,5)P2 and not PI(3,4)P2, and hydrogen-bond formation likely plays a large role in the interaction.

153

NMR studies have been used extensively to investigate the structure of many proteins

(Wuthrich 1990; Wuthrich 2003; Markwick, Malliavin et al. 2008; Opella 2013). Protein ligand interactions can also be examined using NMR, either by NOESY investigation of through-space coupling, or by examining chemical shift or dynamics changes to the protein upon ligand binding. NOESY through-space coupling can be observed between nuclei that are near each other but are not covalently attached. Thus, observed NOESY coupling between a ligand and protein can indicate the ligand binding site. Chemical shift changes occur due to a change in the environment around nuclei, which can be due to the presence of ligand at a binding site or allosteric changes in the protein structure resulting in a change in environment. Here I will attempt to use NMR to characterize the interaction between PI(4,5)P2 and peptides based on the

N-terminal sequence of PTEN. In order to avoid issues with peptide solubility and aggregation, I used residues 10-16 and 12-18 to study this interaction. These short peptides include the crucial binding region around the K13 residue but are short enough to be soluble in a simple buffer (as

31 observed in DLS measurements). P NMR and T1 experiments were done using MAS of 9:1

PC:PI(4,5)P2 MLVs. This allowed us to use the natural, polyunsaturated PI(4,5)P2 lipid and to study it in a realistic membrane model. For solution NMR experiments, a soluble form of

PI(4,5)P2 was required to allow us to perform these experiments. We used PI(4,5)P2-amine (see figure 7-6 for the chemical structure) or dioctanoyl-PI(4,5)P2.

31 P NMR reveals an interaction between PTEN peptides and PI(4,5)P2

31 Using P NMR we can examine the effect of PTEN on the phosphomonoesters of PI(4,5)P2 directly. Hydrogen bond formation or other interactions between the peptide and the

154

31 Figure 7-1. MAS P NMR chemical shift of PI(4,5)P2 with PTEN10-16. Average chemical shift

values for the 4- and 5-phosphate of PI(4,5)P2 in PC / PI(4,5)P2 (90% / 10%) MLVs with no

PTEN10-16 (black), or PTEN10-16 at a 1:1 molar ratio with PI(4,5)P2 (red). Error bars are the standard deviation from three separate measurements. phosphomonoesters will be observed by a change in the chemical shift of the phosphate groups.

MLVs were formed with 90 mol% PC and 10 mol% PI(4,5)P2. The 10 mol% of PI(4,5)P2 was used to decrease measurement times. Although this concentration of PI(4,5)P2 is not strictly physiological, we have found in our previous measurements that the PI(4,5)P2 content in our vesicles does not significantly affect the position of the PI(4,5)P2 phosphate group peaks

(Kooijman, King et al. 2009). It should be noted that the local concentration of PI(4,5)P2 may be much higher as well, due to the possible formation of PI(4,5)P2 rich clusters (Redfern and

Gericke 2004; Redfern and Gericke 2005; Kooijman, King et al. 2009). The chemical shift was

31 observed using MAS P NMR in the presence and absence of the peptide PTEN10-16. PTEN10-16 was added at a 1:1 total molar ratio with PI(4,5)P2 to ensure an observable interaction. Average chemical shift values for the phosphate peaks are shown in Figure 7-1. In the presence of the peptide there is a downfield shift of both phosphate groups, indicating increased deprotonation of 155

the phosphate. The shift is somewhat larger for the 5-phosphate. This increased deprotonation suggests the formation of hydrogen-bonds between PTEN and PI(4,5)P2. This is similar to the proposed “electrostatic hydrogen-bond switch model” for PA (Kooijman, Tieleman et al. 2007;

Loew, Kooijman et al. 2013). The presence of this downfield shift confirms that the PTEN10-16 peptide interacts with PI(4,5)P2 and can be used as a model for the PTEN N-terminus.

The altered chemical shifts for the PI(4,5)P2 phosphomonester groups clearly reveals the interaction with the PTEN peptide. In our solution NMR experiments, we will use a soluble analog for PI(4,5)P2, rather than the PC:PI(4,5)P2 membrane model. This secondary model for the interaction can be tested using the same 31P measurements. Upon mixing the soluble

PI(4,5)P2 and the PTEN peptide a similar effect should be observed. PI(4,5)P2-amine was mixed with PTEN10-16 at five different ratios—0:1, 1:10, 1:5, 1:2, and 1:1 molar ratios of peptide to lipid. Figure 7-2 shows the shift of the 4- and 5-phosphate peaks for each ratio. The 4-phosphate is the furthest downfield at 3.9 ppm (Figure 7-2A), while the 5-phosphate is at 3.28 ppm (Figure

7-2B). Without the interference from the PC phosphodiester the PI(4,5)P2 phosphodiester can be observed at 0.21 ppm (see appendix Figure A10). At low peptide to lipid ratios, there is no significant change in any of the peaks. With a 1:2 or 1:1 ratio the peaks shift downfield, as observed for the PI(4,5)P2 MLVs in the presence of PTEN10-16. The phosphodiester shows no change with the addition of PTEN10-16. This is consistent with what we would expect to observe, as for the natural PI(4,5)P2 we would not expect that PTEN would interact with the phosphodiester that is buried in the membrane. However, the lack of shift for the phosphodiester could also be due to the fact that it is fully deprotonated at the observed pH values and thus it may be less affected by any interaction. In addition, the 5-phosphate shows a larger downfield 156

31 Figure 7-2. P NMR spectra of PI(4,5)P2-amine in the presence of varying PTEN10-16. PI(4,5)P2- amine and PTEN10-16 were mixed together in five different molar ratios—1:0, 10:1, 5:1, 2:1, and 1:1. 31P NMR spectra were acquired for each mixture and overlaid. Intensity values were normalized by the highest peak (P4). A) Close up of the 4-phosphate peak. B) Close up of the 5-phosphate. shift than the 4-phosphate, just as was observed for the PC:PI(4,5)P2 MLV model. Overall these results show that PI(4,5)P2-amine and PTEN10-16 do interact and can be a good model for the

PI(4,5)P2/PTEN interaction.

Given that the interaction between PI(4,5)P2 and PTEN can be observed based on the chemical shift of the phosphomonoesters, this allows us in a straightforward way to examine the

2+ level of PTEN binding. Previously we have looked at Ca binding to PI(4,5)P2 and discussed its implications for PI(4,5)P2 signaling. Since we observed strong binding between PI(4,5)P2 and

2+ Ca it is important to investigate how this interaction may affect other interactions of PI(4,5)P2.

Therefore, the interaction between PI(4,5)P2 and PTEN was investigated in the presence of 1 mM Ca2+. While 1 mM Ca2+ is somewhat higher than physiological concentrations, this

157

concentration was chosen as we previously observed a large interaction between PI(4,5)P2 and

Ca2+ at this concentration and we should therefore be able to easily see any possible effect from

2+ 2+ the presence of Ca . MLVs were formed containing 9:1 PC:PI(4,5)P2. Ca was added to these vesicles to reach a bulk concentration of 1 mM. The A23187 ionophore was used to equilibrate

2+ Ca across the bilayers. PTEN10-16 was added to a 1:1 molar ratio with the PI(4,5)P2. Multiple freeze/thaw cycles were performed to help incorporate the peptide throughout the MLVs. The chemical shift values for the phosphomonoesters were recorded and averaged over multiple

2+ samples. Figure 7-3 shows the results for vesicles with PTEN10-16 and Ca . These results are

2+ compared against PC:PI(4,5)P2 alone, with Ca alone, or with PTEN10-16 alone. The combined

2+ presence of PTEN10-16 and Ca leads to a downfield shift of both of the phosphomonoesters, as

31 2+ Figure 7-3. MAS P NMR chemical shift of PI(4,5)P2 with PTEN10-16 and Ca . Average chemical

shift values for the 4- and 5-phosphate of PI(4,5)P2 in PC / PI(4,5)P2 (90% / 10%) MLVs with 1 mM 2+ Ca and PTEN10-16 at a 1:1 molar ratio with PI(4,5)P2 (green) compared against PC / PI(4,5)P2 alone, 2+ or with 1:1 PTEN10-16 (red), or with 1 mM Ca (blue). Error bars are the standard deviation from three separate measurements.

158

we expect. However, the shift is not significantly greater than the shift observed for Ca2+ or

2+ PTEN10-16 independently. This shows that the presence of Ca and PTEN10-16 does not lead to a cumulative effect on PI(4,5)P2 deprotonation, which suggests that they bind independently.

Dynamics of the PI(4,5)P2 headgroup

Spin-lattice relaxation times give an indication of the dynamics of the studied nuclei and can be measured using NMR experiments. In a membrane environment, an increase in the lipid relaxation time can indicate a decreased axial rotation of the lipid (Cullis, De Kruyff et al. 1976;

Dufourc, Mayer et al. 1992; Lu, Damodaran et al. 2005). The decreased rotation allows fewer opportunities for the lipid to interact with the surrounding lipids and relax itself by passing along its energy. The dynamics of the PI(4,5)P2 headgroup were examined by T1 relaxation measurements. By measuring the T1 relaxation time in the presence and absence of PTEN the effect of binding on the lipid dynamics can be determined. The T1 relaxation value was measured using a T1 inversion recovery experiment consisting of a single 180° pulse followed by a variable delay time and then a final 90° pulse. MLVs were formed with a 9:1 molar ratio of PC and

PI(4,5)P2. T1 experiments were done with MAS in order to remove the CSA (chemical shift anisotropy) from the spectrum. Figure 7-4 shows the resulting MAS 31P NMR spectra for the 16 delay times tested. The 4-phosphate and 5-phosphate are clearly resolved and are observed downfield. The PC phosphodiester is observed upfield at -0.6 ppm. The PI(4,5)P2 phosphodiester is hidden by the overlap with the PC phosphodiester as we have observed in previous experiments. When the delay time is much longer than the T1 value (e.g. 10s) the system is able to completely relax after the initial 180° pulse. For very short times (e.g. 0.001s), the system

159

31 31 Figure 7-4. MAS P NMR inversion recovery experiment with PC:PI(4,5)P2 MLVs. MAS P

NMR spectra are plotted with respect to delay time for PC / PI(4,5)P2 (90% / 10%) MLVs. does not have time to relax and the full peak is shifted to the –x axis and we observe an inverted peak. The actual T1 time for a given peak is roughly equal to the time where no peak is observed

(i.e. the time at which the peak has relaxed halfway back from the 180° pulse) divided by ln(2).

T1 values were calculated quantitatively by plotting the measured peak area with respect to the delay time and fitting it (see methods). The relaxation curve and calculated T1 values for the

PI(4,5)P2 phosphomonoesters and PC phosphodiester in the PC / PI(4,5)P2 (90% / 10%) MLVs are shown in Figure 7-5A. The phosphomonoesters have considerably longer relaxation times as compared to the PC phosphodiester. The longer relaxation times are likely due to the increased solvent exposure of the two phosphomonoesters. PTEN10-16 was then added at a 1:1

PI(4,5)P2/peptide molar ratio.

160

Figure 7-5. T1 spin-lattice relaxation vales for PI(4,5)P2 and PC in PC / PI(4,5)P2 MLVs with

PTEN10-16. The normalized peak area is plotted with respect to delay time and fit to provide the

resulting T1 value for the 4- and 5-phosphates of PI(4,5)P2 and the phosphodiester of PC in PC /

PI(4,5)P2 (90% / 10%) MLVs with (a) 0 µmol PTEN10-16 or (b) 1 µmol PTEN10-16.

Table 7-1. Relaxation times for PI(4,5)P2 in the presence and absence of PTEN10-16 T P4 P5 PC 1

PC:PI(4,5)P 9:1 1.38 ± 0.06 1.28 ± 0.04 0.92 ± 0.01 2 + 1µmol PTEN 1.45 ± 0.01 1.33 ± 0.01 0.931 ± 0.008 10-16

Δ 0.07 0.05 0.008

T1 values are an average over three measurements. The uncertainty is determined from the standard deviation. Δ represents the difference in T1 between the presence and absence of the peptide.

The new T1 values are shown in Figure 7-5B. When PTEN10-16 is added (at a 1:1 molar ratio) an increase in the relaxation values is measured. For PC the change is insignificant showing that

PTEN does not interact with PC. For the phosphomonoesters the increased relaxation time indicates a decrease in the rotation of the lipid. The observed change in T1 is comparable to the shift observed in other studies of antimicrobial peptides interacting with membrane lipids (Lu, 161

Damodaran et al. 2005). This shows that the peptide binds tightly to the lipid and affects the dynamics of the lipid.

PTEN peptide NMR assignment

Solution NMR experiments were performed using PTEN10-16 and PTEN12-18 (see Figure 7-6 for peptide structures). The respective peptide was dissolved in a PBS buffer with a concentration of ~5 mM. The peptides were assigned using COSY, TOCSY, and NOESY experiments on the peptide. To assist in assigning the peptide peaks, the peak shifts were compared against a standard amino acid chemical shift table (see table 7-2). The PTEN10-16 assignment is shown in Figure 7-7. The peptide peaks were assigned based on the spin system and the expected chemical shift values. The TOCSY spectrum (see appendix Figure A11) was used to assist and confirm the assignments based on the COSY. The three arginine residues could not be resolved. The serine NH was not observed, possibly due to a high exchange rate.

162

Figure 7-6. Chemical structures of the tested peptides and soluble PI(4,5)P -amine. A) Structure 2 of PTEN , sequence SRNKRRY. B) Structure of PTEN , sequence NKRRYQE. C) PI(4,5)P - 10-16 12-18 2 amine, a soluble analog of PI(4,5)P . 2

163

Figure 7-7. COSY assignment of PTEN . COSY spectrum was acquired from a 5 mM sample of 10-16 PTEN dissolved in 15 mM phosphate buffer with 100 mM NaCl at pH ~5. Red lines indicate 10-16 cross peaks that were used to assign the peptide peaks.

164

Figure 7-8. NOESY assignment of PTEN . NOESY spectrum was acquired from a 5 mM sample 10-16 of PTEN dissolved in 15 mM phosphate buffer with 100 mM NaCl at pH ~5. Red lines indicate 10-16 cross peaks that were used to assign the peptide peaks.

165

Table 7-2. Standard amino acid chemical shift values NH αH βH

Serine (S) 8.38 4.50 3.88, 3.88

γCH2 1.70, 1.70 1.89, Arginine (R) 8.27 4.38 δCH 3.32, 3.32 1.79 2 NH 7.17, 6.62

Asparagine 8.75 4.75 2.83, γNH2 7.59, 6.91 (N) 2.75

γCH2 1.45, 145

1.85, δCH2 1.70, 1.70 Lysine (K) 8.41 4.36 1.76 εCH2 3.02, 3.02 + εNH3 7.52 3.13, 2,6H 7.15 Tyrosine (Y) 8.18 4.60 2.92 3,5H 6.86 Random coil 1H chemical shifts for the amino acids found in PTEN10-16. Data are determined from the tetrapeptide GGXA. Chemical shift values are from (Wüthrich 1986).

Table 7-3. Assigned PTEN10-16 amino acid chemical shift values NH αH βH Serine (S) 4.50 4.09, 3.89

γCH2 1.59 Arginine (R) 8.27 4.17 δCH2 3.08 NH

Asparagine 8.45 4.58 γNH2 (N)

γCH2 δCH 1.59 Lysine (K) 8.69 4.29 1.64 2 εCH2 2.89 + εNH3

Arginine (R2) γCH2 1.46

Arginine (R3) γCH2 1.24 2,6H Tyrosine (Y) 8.00 4.41 3.03, 2.81 3,5H

Amino acid chemical shift peak values for PTEN10-16 obtained from assignment based on Figure 7-7 and 7-8.

166

PI(4,5)P2-amine NMR assignment

The PI(4,5)P2-amine peaks were also assigned using COSY experiments. In order to assign the inositol ring, the chemical shifts were compared with assignments others have made for the

PI(4,5)P2 inositol ring (Lindon, Baker et al. 1986; Reid and Gajjar 1987). The PI(4,5)P2-amine assignment is shown in Figure 7-9.

Figure 7-9. COSY assignment of PI(4,5)P2-amine. COSY spectrum was acquired from a 5 mM

sample of PI(4,5)P2-amine dissolved in 15 mM phosphate buffer with 100 mM NaCl at pH ~5.

167

NMR investigation of PTEN/PI(4,5)P2 interaction

PTEN10-16 and PI(4,5)P2-amine were combined at a 1:1 molar ratio. A low pH (~5) PBS buffer was used. The lower pH values enabled the exchangeable protons to be observed. A phosphorous shift was still observed at this lower pH value, showing that the higher [H+] does not eliminate the interaction (not shown). The peak assignments of the individual PI(4,5)P2-amine and PTEN peptides were compared against the mixture to assign all peaks. COSY spectra were obtained to show the changes in chemical shifts or in peak intensities. The COSY spectra of PTEN10-16 in the presence or absence of PI(4,5)P2-amine is shown in figure 7-10.

Figure 7-10. Comparison of COSY spectra of PTEN10-16 in the presence and absence of PI(4,5)P -amine. COSY spectrum was acquired from a 5 mM sample of PTEN dissolved in 15 2 10-16 mM phosphate buffer with 100 mM NaCl at pH ~5, with (black) and without (red) an equimolar

amount of PI(4,5)P2-amine.

168

The red peaks are the spectrum in the absence of PI(4,5)P2-amine, the black peaks include

PI(4,5)P2-amine (notice the extra black inositol ring peaks around 4 ppm). There are several large changes. First, the KNH peak is completely gone in the spectrum with PI(4,5)P2-amine.

The disappearance of the KNH peak could be due to peak broadening, either as a result of reduced mobility of the lysine residue due to the interaction or due to rapid exchange rate or hydrogen bond formation with this proton. The shifts for the tyrosine peaks indicates a change in environment for the tyrosine residue, and suggest that this residue plays a role in the interaction.

The NOESY spectrum of the peptide-lipid mix was obtained in order to try to observe through space correlations between the peptide and the lipid. The NOESY spectrum of the mixture is shown in Figure 7-11. Unfortunately no direct cross peaks between the lipid and the peptide are observed. This is probably due to the fact that the main interaction likely occurs between the peptide and the phosphomonoesters or inositol ring hydroxyl groups. These protons are not observable due to their high exchange rate with water (or broadening due to hydrogen bond formation with water molecules). The guanidinium NH peak of the arginine groups does disappear though, indicating that the arginine guanidinum groups interact with the lipid. Many of the interpeptide NOESY crosspeaks disappear as well, which may indicate that the peptide stretches out around the PI(4,5)P2-headgroup. However, without observing actual crosspeaks between the peptide and the lipid we cannot determine exactly where the peptide is positioned.

169

Figure 7-11. NOESY spectra of a PTEN :PI(4,5)P -amine mixture. The NOESY spectrum was 10-16 2 acquired from a 5 mM sample of 1:1 mix of PTEN :PI(4,5)P -amine dissolved in 15 mM 10-16 2 phosphate buffer with 100 mM NaCl at pH ~5.

170

Discussion

The PI(4,5)P2 5-phosphate is important for the PTEN/PI(4,5)P2 interaction

In our 31P measurements, we find that the 5-phosphate consistently has a stronger downfield shift in the presence of the PTEN peptides. This larger shift indicates the importance of the 5- phosphate for the interaction. Interestingly, the increased shift for the 5-phosphate was observed for the soluble PI(4,5)P2 analog as well as the natural PI(4,5)P2 molecule within the membrane model. Thus the greater effect on the 5-phosphate is not simply due to the orientation of the

PI(4,5)P2 headgroup at the lipid interface. Although it may be possible that PI(4,5)P2-amine may form micelles, in which case the orientation of the headgroup with respect to the micelle may affect the availability of the two phosphates. Coincidentally, the 5-phosphate is the primary difference between PI(4,5)P2 and PI(3,4)P2. In PI(3,4)P2 this phosphate is instead at the 3- position. Thus, the unique properties of the 5-position (and the 5-phosphate), may be one of the reasons for the large difference in affinity between these two phosphatidylinositol bisphosphates.

2+ Ca and PTEN compete for PI(4,5)P2

The combination of PTEN and Ca2+ did not lead to a cumulative increase in the 31P shift for

2+ PI(4,5)P2. Previously we observed that the effect of Ca binding and PE hydrogen-bond

2+ formation with PI(4,5)P2 was cumulative (see Figure 5-9), which suggests that Ca binding and hydrogen-bond formation can occur simultaneously and increase PI(4,5)P2 deprotonation. In this

2+ case we do not observe a cumulative effect, which suggests that Ca and PTEN10-16 bind to

2+ PI(4,5)P2 independently. This shows that Ca does not enhance PTEN binding, although we can’t conclusively say whether Ca2+ reduces the PTEN binding. We only see the effect of the 171

2+ 2+ PTEN and the Ca on PI(4,5)P2 and so we cannot distinguish between Ca and PTEN binding.

However, we should note that this is an indirect method of measuring PTEN binding, and we

2+ cannot definitively conclude that Ca and PTEN do not bind to PI(4,5)P2 simultaneously. Given

2+ the strong interaction between Ca and PI(4,5)P2 as well as between PTEN and PI(4,5)P2, it is worth investigating this further in the future. Further studies using more direct binding measurements will be necessary to confirm these results.

PTEN binds tightly to PI(4,5)P2 and alters its dynamics

The T1 measurements show that PTEN binds tightly to PI(4,5)P2 and is able to alter its motion within the membrane. The increased relaxation time suggests that peptide binding reduces the axial rotation of the lipid. Alternatively, the change in relaxation time may be due to decreased access to the solvent due to the bound peptide. Unlike the 31P chemical shift measurements, the

T1 measurement does not show a clear preference for either phosphate. This may indicate that the

T1 increase is more of a general effect, affecting the whole lipid headgroup. Thus both phosphates are affected comparably.

The tyrosine, lysine and arginine residues of PTEN’s n-terminal end interact with PI(4,5)P2

The 2D NMR measurements can show which residues of the PTEN peptides interact with

PI(4,5)P2. The lack of cross peaks between the peptide and the lipid is disappointing. This makes it much more difficult to draw conclusions about how PTEN wraps itself around the PI(4,5)P2 headgroup. Further solid-state experiments, such as measuring 31P-13C distances with REDOR, will be necessary to learn more about the structure of the PI(4,5)P2/PTEN interaction. However, based on the chemical shifts and the broadening of several of the peptide peaks, we can make 172

some conclusions about the residues that interact with PI(4,5)P2. The K13, Y16 and at least one, possibly all three of the arginine residues (R12, R14 and R15), are all affected by the presence of

PI(4,5)P2. The lysine backbone amide peak may actually interact with the lipid, as it disappears in the presence of PI(4,5)P2. The tyrosine is still observed, but several of its peaks shift. For the arginine, the only observed guanidinium peak (which may be a contribution from multiple arginines) is eliminated in the presence of the lipid. Thus further experimentation should focus on these critical residues.

Conclusions

PTEN interacts strongly with PI(4,5)P2, as evidenced by our NMR data. A peptide derived from the N-terminus of PTEN interacts with and causes a large downfield shift of the PI(4,5)P2 phosphate peaks. This shift is larger for the 5-phosphate, indicating the importance of this phosphate for the interaction. The peptide PTEN10-16 along with the soluble PI(4,5)P2 analog was

2+ found to be a good model for the PI(4,5)P2/PTEN interaction. Ca was found to compete with

PTEN, reducing the binding of one or both. The dynamics of the PI(4,5)P2 headgroup was found to be strongly affected by PTEN binding, revealing the tight binding of PTEN to the headgroup.

Solution NMR measurements showed that the K13, Y16, and arginine residues (R12, R14, and

R15) are each affected by the addition of PI(4,5)P2, and thus are likely to be important for the interaction. Additional NMR experiments are needed to further characterize this interaction and to determine the unique geometry that gives it such high specificity.

173

Chapter 8

Project summation

In this thesis we have extensively studied the ionization properties of PI(4,5)P2 and its interactions with common components of the plasma membrane inner leaflet and the cytosol.

Using a new fitting procedure we have developed a model for the ionization of the phosphatidylinositol polyphosphates within the membrane system. While this ionization had been described before in a qualitative manner, this new fitting model has allowed us to quantitatively describe the ionization and measure pKa values for each ionization step. These pKa values will be useful for future modeling projects as well as giving us a deeper understanding of the ionization of these lipids. In addition to revisiting the ionization of phosphatidylinositol polyphosphates in simple binary membrane systems, we also investigated the ionization of the important phosphatidylinositol bisphosphate, PI(4,5)P2, in complex ternary lipid systems. We investigated PI(4,5)P2 ionization with each of the plasma membrane inner leaflet lipids, PE, PI, and PS. We found evidence for a significant interaction between PE and PI(4,5)P2. PE forms a hydrogen-bond with the PI(4,5)P2 headgroup, leading to a shift in the ionization of PI(4,5)P2 to lower pKa values. The hydrogen-bond appears to form preferentially between PE and the 5- phosphate of PI(4,5)P2, as seen by the stronger effect of PE on the 5-phosphate. The interaction between PS and PI(4,5)P2 was found to be somewhat enigmatic. We observed surprisingly little effect from PS, although we expected to find a shift to higher pKa values based on the indirect

174

effect of the increased negative membrane potential in the presence of PS. This indicates that the effect of increasing the membrane potential can be somewhat complicated. In this case it seems likely that PS negates the effect of its negative charge by hydrogen-bond formation with

PI(4,5)P2. No macroscopic domain formation was observed for PC:PS:PI(4,5)P2 vesicles. We found an important interaction between PI and PI(4,5)P2. Like PS, PI increases the negative potential of the membrane. However, again we observed little effect of PI on the ionization of

PI(4,5)P2, and we suggest that this is caused by the competing effect of hydrogen-bond formation. The importance of this interaction was highlighted by GUV fluorescence microscopy studies, which showed that PI promoted the formation of large bulge shaped PI(4,5)P2 rich domains.

The ionization of PI(4,5)P2 was also examined in the presence of cholesterol and the divalent cations Ca2+ and Mg2+. Cholesterol was found to have a small but significant effect on the ionization of PI(4,5)P2. This indicates that cholesterol interacts with PI(4,5)P2 but not directly through the phosphomonoesters. This interaction may be part of the mechanism for cholesterol

2+ induced PI(4,5)P2 domain formation. Ca was found to have a large effect on PI(4,5)P2 ionization, while Mg2+ had a smaller effect. The larger effect of Ca2+ may be due to its lower hydration energy or the smaller radius of its hydrated state (Wang, Collins et al. 2012;

Slochower, Huwe et al. 2013). However, we also found that Ba2+ has a smaller effect than Ca2+ despite the fact that it has a lower hydration energy, showing that the strength of the interaction is not purely dependent on hydration energy. In complex mixtures containing varying combinations of cholesterol, Ca2+, PE, and PI, we found that many of these components have a cumulative effect on PI(4,5)P2 ionization. These combined effects could result in a fairly 175

significant increase in PI(4,5)P2 charge as compared to the binary PC:PI(4,5)P2 mixture. PI actually seemed to enhance the effect of Ca2+, as the resulting chemical shift was higher than the combined effect of PI and Ca2+ alone.

2+ 2+ The interaction between Ca , Mg and PI(4,5)P2 was investigated in greater detail using X- ray techniques. These experiments were carried out in the presence of 100 mM KCl, to mimic the salt levels in the cell and provide some competition for the Ca2+ and Mg2+ ions. K+ ions were

+ found to be enriched at the surface of the PI(4,5)P2 monolayer, reaching 0.8 M overall K concentration at the surface. The addition of Ca2+ to the bulk leads to a peak of 1,000 fold

2+ 2+ surface enrichment of Ca ions to a total of 1.6 Ca ions / PI(4,5)P2 or 1 M (at 1 mM bulk

2+ concentration). The binding of Ca leads to condensation of the PI(4,5)P2 molecules as revealed by x-ray reflectivity. Mg2+ was found to compete with Ca2+ and reduce Ca2+ binding, but Ca2+ was able to out-compete Mg2+ and seems to eliminate Mg2+ at a 1:1 molar ratio. A small fraction

+ of K ions was found to remain associated with PI(4,5)P2 despite the presence of the divalent

+ cations. This fraction of K ions could have an important effect on PI(4,5)P2, specifically in protein interactions.

The specific interaction between PTEN and PI(4,5)P2 was examined using NMR techniques.

31 P NMR clearly revealed this interaction via the effect on the ionization of PI(4,5)P2. The PTEN peptide caused a downfield shift, indicating increased deprotonation of the PI(4,5)P2 phosphomonoesters. This deprotonation could be attributed to the formation of hydrogen-bonds between the peptide and PI(4,5)P2. The peptide caused a greater effect on the 5-phosphate, indicating the importance of this phosphomonoester for the interaction. The presence of Ca2+ did

176

not lead to a significant increase in the effect of PTEN, supporting the notion that PTEN binds

2+ PI(4,5)P2 independent of Ca . Relaxation measurements showed that the PTEN peptide binds tightly to PI(4,5)P2, constraining the movement of the lipid and increasing relaxation times. The peptide NMR peaks were assigned using 2D NMR experiments. Using 2D NMR of the peptide and the lipid together, we found that the K13 residue was significantly affected by the interaction. The Y16 residue also seemed to participate in the interaction, while the arginine and serine residues were involved as well.

Future work

We found strong evidence that lipids within the membrane can interact with PI(4,5)P2 and affect its localization and ionization. Whether these interactions occur in vivo and how they

2+ affect PI(4,5)P2 signaling is unknown. PI, cholesterol, and Ca have a strong potential to affect

PI(4,5)P2/protein interactions as they have been shown to affect PI(4,5)P2 localization in vitro.

+ PE also may affect PI(4,5)P2 due to the formation of hydrogen bonds. K is a component that is easily overlooked, but according to our experiments it may be tightly associated with PI(4,5)P2 and may affect its signaling. It will be very important to characterize how these interactions can effect PI(4,5)P2 mediated signaling pathways.

PTEN can act as a good model system for the effect of these components on PI(4,5)P2 signaling, as it binds specifically to PI(4,5)P2 and it is part of a very important signaling pathway. Therefore, I think it will be important to investigate PTEN binding to PI(4,5)P2 in the presence of varying lipid components. This could be accomplished by measuring the

PTEN/PI(4,5)P2 binding constant using SPR or ITC experiments with PI(4,5)P2 in various

177

membrane environments (containing PI, cholesterol, or PE) or in the presence of other cations

2+ 2+ + (Ca , Mg , or K ). SPR would allow a direct measurement of PTEN binding to PI(4,5)P2 membrane models, as well as kinetic measurements. However, SPR measurements have proven to be quite complicated with lipid vesicles containing phosphoinositides. It should be noted that

ITC experiments may be complicated by the fact that PTEN can also cause PI(4,5)P2 domain formation upon binding. Thus, the observed enthalpy will be equal to the combined enthalpy of the domain formation and the PTEN binding enthalpy. This problem may be avoided by using vesicle mixtures that already have strong domain formation, as in this case the effect of PTEN on the localization of PI(4,5)P2 should be minimal. The strong domain formation could be promoted by adding PI or cholesterol, or it could also be promoted simply by lowering the temperature to

10°C (Redfern and Gericke 2004). Thus we may be able to separate the domain formation enthalpy from the binding enthalpy.

In addition, it will be important to establish whether PI(4,5)P2 cluster formation can occur in vivo. One of the difficulties of showing PI(4,5)P2 cluster formation is that often fluorescently labeled PI(4,5)P2 molecules or non-physiological PI(4,5)P2 concentrations must be used to visualize these domains (Redfern and Gericke 2004; Fernandes, Loura et al. 2006; Jiang 2010;

Wang, Collins et al. 2012; Sarmento, Coutinho et al. 2013; Jiang, Redfern et al. 2014). CARS spectroscopy may be able to help resolve this problem, as CARS can distinguish lipids based on their Raman scattering modes. Thus, PI(4,5)P2 could be labeled with deuterium to provide contrast. This should have little effect on the physical properties of PI(4,5)P2. It may be difficult to actually label PI(4,5)P2 and observe it within living cells. However, more complex membrane models, such as asymmetric GUVs with complex membrane composition (Chiantia, Schwille et 178

al. 2011; Hu, Li et al. 2011; Matosevic and Paegel 2013; Coyne, Patel et al. 2014; Lin and

London 2014), could be used as more accurate models of the actual cellular membrane.

Our investigation of cation binding to a PI(4,5)P2 was highly successful, and we were able to determine quantitative numbers for cation binding to PI(4,5)P2. However, it should be noted that we used a pure PI(4,5)P2 monolayer, while in the cellular plasma membrane there are many other components present. We justified this choice by pointing out that PI(4,5)P2 rich domains may occur in vivo. We have found that other components (e.g. PI and cholesterol) may be enriched in

PI(4,5)P2 domains as well. Thus, it may be important to extend our x-ray studies by investigating systems composed of PI, cholesterol, and PI(4,5)P2, in order to determine how the changing composition of PI(4,5)P2 clusters may influence the number of cations that are enriched at the

PI(4,5)P2 headgroup region. We have already obtained some preliminary data for

PI(4,5)P2/Cholesterol monolayers. These experiments will be significantly more complicated, as it will be difficult to differentiate between ions that are bound to PI or cholesterol, and ions that are actually associated with PI(4,5)P2 molecules.

The PTEN/PI(4,5)P2 interaction could be further characterized with more complex NMR experiments. Solution NMR experiments are useful for structure determination, however, the

PTEN/PI(4,5)P2 interaction may be significantly different at the membrane surface. Therefore it will be important to study it with solid-state NMR experiments. In particular, REDOR NMR experiments can be used to obtain distance constraints between the PI(4,5)P2 phosphates and labeled carbon or nitrogen nuclei within a PTEN peptide. Fluorine labels could also be added to

179

the peptide to measure distance constraints. In addition, 1H-31P correlation experiments could also be done to investigate the effect of the PTEN binding to PI(4,5)P2.

180

References

Almeida, P. F., A. Best, et al. (2011). "Monte Carlo simulation of protein-induced lipid demixing in a membrane with interactions derived from experiment." Biophys J 101(8): 1930-1937.

Als-Nielsen, J. and D. McMorrow (2011). Elements of modern X-ray physics. Hoboken, Wiley.

Baumgart, T., A. T. Hammond, et al. (2007). "Large-scale fluid/fluid phase separation of proteins and lipids in giant plasma membrane vesicles." Proc Natl Acad Sci U S A 104(9): 3165-3170.

Blin, G., E. Margeat, et al. (2008). "Quantitative analysis of the binding of ezrin to large unilamellar vesicles containing phosphatidylinositol 4,5 bisphosphate." Biophys J 94(3): 1021-1033.

Bootman, M. D., P. Lipp, et al. (2001). "The organisation and functions of local Ca(2+) signals." J Cell Sci 114(Pt 12): 2213-2222.

Bradshaw, J. P., R. J. Bushby, et al. (1999). "Orientation of the headgroup of phosphatidylinositol in a model biomembrane as determined by neutron diffraction." Biochemistry 38(26): 8393-8401.

Bu, W., K. Flores, et al. (2009). "Preferential affinity of calcium ions to charged phosphatidic acid surface from a mixed calcium/barium solution: X-ray reflectivity and fluorescence studies." Langmuir 25(2): 1068-1073.

Bu, W., P. J. Ryan, et al. (2006). "Ion distributions at charged aqueous surfaces by near- resonance X-ray spectroscopy." J Synchrotron Radiat 13(Pt 6): 459-463.

Buckley, J. T. and J. N. Hawthorne (1972). "Erythrocyte membrane polyphosphoinositide metabolism and the regulation of calcium binding." J Biol Chem 247(22): 7218-7223.

Campbell, R. B., F. Liu, et al. (2003). "Allosteric activation of PTEN phosphatase by phosphatidylinositol 4,5-bisphosphate." J Biol Chem 278(36): 33617-33620.

181

Catimel, B., C. Schieber, et al. (2008). "The PI(3,5)P2 and PI(4,5)P2 interactomes." J Proteome Res 7(12): 5295-5313.

Chiantia, S., P. Schwille, et al. (2011). "Asymmetric GUVs prepared by MbetaCD-mediated lipid exchange: an FCS study." Biophys J 100(1): L1-3.

Cho, W. (2006). "Building signaling complexes at the membrane." Sci STKE 2006(321): pe7.

Chun, Y. S., S. Shin, et al. (2010). "Cholesterol modulates ion channels via down-regulation of phosphatidylinositol 4,5-bisphosphate." J Neurochem 112(5): 1286-1294.

Cinar, B., N. K. Mukhopadhyay, et al. (2007). "Phosphoinositide 3-kinase-independent non- genomic signals transit from the androgen receptor to Akt1 in membrane raft microdomains." J Biol Chem 282(40): 29584-29593.

Coyne, C. W., K. Patel, et al. (2014). "Lipid bilayer vesicle generation using microfluidic jetting." J Vis Exp(84): e51510.

Cullis, P. R., B. De Kruyff, et al. (1976). "Factors affecting the motion of the polar headgroup in phospholipid bilayers. A 31P NMR study of unsonicated phosphatidylcholine liposomes." Biochim Biophys Acta 426(3): 433-446.

Cullis, P. R. and B. Dekruijff (1979). "Lipid Polymorphism and the Functional Roles of Lipids in Biological-Membranes." Biochim Biophys Acta 559(4): 399-420.

Daillant, J., L. Bosio, et al. (1991). "Interaction of Cations with a Fatty-Acid Monolayer - a Grazing-Incidence X-Ray-Fluorescence and Reflectivity Study." Langmuir 7(4): 611- 614.

Das, S., J. E. Dixon, et al. (2003). "Membrane-binding and activation mechanism of PTEN." Proc Natl Acad Sci U S A 100(13): 7491-7496.

Dasgupta, U., T. Bamba, et al. (2009). "Ceramide kinase regulates phospholipase C and phosphatidylinositol 4, 5, bisphosphate in phototransduction." Proc Natl Acad Sci U S A 106(47): 20063-20068.

182

Dawson, R. M. (1965). "'Phosphatido-peptide'-like complexes formed by the interaction of calcium triphosphoinositide with protein." Biochem J 97(1): 134-138.

Dickson, E. J., J. B. Jensen, et al. (2014). "Golgi and plasma membrane pools of PI(4)P contribute to plasma membrane PI(4,5)P2 and maintenance of KCNQ2/3 ion channel current." Proc Natl Acad Sci U S A.

Duerr, E. M., B. Rollbrocker, et al. (1998). "PTEN mutations in gliomas and glioneuronal tumors." Oncogene 16(17): 2259-2264.

Dufourc, E. J., C. Mayer, et al. (1992). "Dynamics of phosphate head groups in biomembranes. Comprehensive analysis using phosphorus-31 nuclear magnetic resonance lineshape and relaxation time measurements." Biophys J 61(1): 42-57.

Elhyany, S., E. Assa-Kunik, et al. (2004). "The integrity of cholesterol-enriched microdomains is essential for the constitutive high activity of protein kinase B in tumour cells." Biochem Soc Trans 32(Pt 5): 837-839.

Evans, J. H., S. H. Gerber, et al. (2004). "The calcium binding loops of the cytosolic phospholipase A2 C2 domain specify targeting to Golgi and ER in live cells." Mol Biol Cell 15(1): 371-383.

Fairn, G. D., M. Hermansson, et al. (2011). "Phosphatidylserine is polarized and required for proper Cdc42 localization and for development of cell polarity." Nat Cell Biol 13(12): 1424-1430.

Fairn, G. D., N. L. Schieber, et al. (2011). "High-resolution mapping reveals topologically distinct cellular pools of phosphatidylserine." J Cell Biol 194(2): 257-275.

Falke, J. J. and B. P. Ziemba (2014). "Interplay between phosphoinositide lipids and calcium signals at the leading edge of chemotaxing ameboid cells." Chem Phys Lipids.

Felemez, M., P. Bernard, et al. (2000). "Inframolecular protonation process of myo-inositol 1,4,5-tris(phosphate) and related compounds: Dynamics of the intramolecular interactions and evidence of C-H center dot center dot center dot O hydrogen bending." J Am Chem Soc 122(13): 3156-3165.

183

Fernandes, F., L. M. Loura, et al. (2006). "Absence of clustering of phosphatidylinositol-(4,5)- bisphosphate in fluid phosphatidylcholine." J Lipid Res 47(7): 1521-1525.

Folch, J., M. Lees, et al. (1957). "A simple method for the isolation and purification of total lipides from animal tissues." J Biol Chem 226(1): 497-509.

Franca-Koh, J., Y. Kamimura, et al. (2007). "Leading-edge research: PtdIns(3,4,5)P3 and directed migration." Nat Cell Biol 9(1): 15-17.

Gambhir, A., G. Hangyas-Mihalyne, et al. (2004). "Electrostatic sequestration of PIP2 on phospholipid membranes by basic/aromatic regions of proteins." Biophys J 86(4): 2188- 2207.

Gamper, N. and M. S. Shapiro (2007). "Target-specific PIP(2) signalling: how might it work?" J Physiol 582(Pt 3): 967-975.

Gao, X., P. R. Lowry, et al. (2011). "PI3K/Akt signaling requires spatial compartmentalization in plasma membrane microdomains." Proc Natl Acad Sci U S A 108(35): 14509-14514.

Gericke, A., M. Munson, et al. (2006). "Regulation of the PTEN phosphatase." Gene 374: 1-9.

Ghosh, S. K., S. Castorph, et al. (2012). "Measuring Ca2+-induced structural changes in lipid monolayers: implications for synaptic vesicle exocytosis." Biophys J 102(6): 1394-1402.

Graber, Z. T., A. Gericke, et al. (2014). "Phosphatidylinositol-4,5-bisphosphate ionization in the presence of cholesterol, calcium or magnesium ions." Chem Phys Lipids 182: 62-72.

Graber, Z. T., Z. Jiang, et al. (2012). "Phosphatidylinositol-4,5-bisphosphate ionization and domain formation in the presence of lipids with hydrogen bond donor capabilities." Chem Phys Lipids 165(6): 696-704.

Graber, Z. T. and E. E. Kooijman (2013). "Ionization behavior of polyphosphoinositides determined via the preparation of pH titration curves using solid-state 31P NMR." Methods Mol Biol 1009: 129-142.

184

Hauser, H. (1989). "Mechanism of spontaneous vesiculation." Proc Natl Acad Sci U S A 86(14): 5351-5355.

Hauser, H. and R. M. Dawson (1967). "The binding of calcium at lipid-water interfaces." Eur J Biochem 1(1): 61-69.

Hendrickson, H. S. and C. E. Ballou (1964). "Ion Exchange Chromatography of Intact Brain Phosphoinositides on Diethylaminoethyl Cellulose by Gradient Salt Elution in a Mixed Solvent System." J Biol Chem 239: 1369-1373.

Hope, H. R. and L. J. Pike (1996). "Phosphoinositides and phosphoinositide-utilizing enzymes in detergent-insoluble lipid domains." Mol Biol Cell 7(6): 843-851.

Hu, P. C., S. Li, et al. (2011). "Microfluidic fabrication of asymmetric giant lipid vesicles." ACS Appl Mater Interfaces 3(5): 1434-1440.

Hunt, J. P. (1963). Metal Ions in Aqueous Solution, W.A. Benjamin.

Iijima, M., Y. E. Huang, et al. (2004). "Novel mechanism of PTEN regulation by its phosphatidylinositol 4,5-bisphosphate binding motif is critical for chemotaxis." J Biol Chem 279(16): 16606-16613.

James, D. J., C. Khodthong, et al. (2008). "Phosphatidylinositol 4,5-bisphosphate regulates SNARE-dependent membrane fusion." J Cell Biol 182(2): 355-366.

Jiang, Z. (2010). Characterization of phosphoinositide and sphingolipid domain formation in model membranes Doctor of Philosophy, Kent State University.

Jiang, Z., R. E. Redfern, et al. (2014). "Cholesterol stabilizes fluid phosphoinositide domains." Chem Phys Lipids.

Johnson, C. M., G. R. Chichili, et al. (2008). "Compartmentalization of phosphatidylinositol 4,5- bisphosphate signaling evidenced using targeted phosphatases." J Biol Chem 283(44): 29920-29928.

185

Kjaer, K. (1994). "Some Simple Ideas on X-Ray Reflection and Grazing-Incidence Diffraction from Thin Surfactant Films." Physica B 198(1-3): 100-109.

Kjaer, K., J. Alsnielsen, et al. (1989). "Synchrotron X-Ray-Diffraction and Reflection Studies of Arachidic Acid Monolayers at the Air-Water-Interface." Journal of Physical Chemistry 93(8): 3200-3206.

Kolsch, V., P. G. Charest, et al. (2008). "The regulation of cell motility and chemotaxis by phospholipid signaling." J Cell Sci 121(Pt 5): 551-559.

Kooijman, E. E. and K. N. Burger (2009). "Biophysics and function of phosphatidic acid: a molecular perspective." Biochim Biophys Acta 1791(9): 881-888.

Kooijman, E. E., K. M. Carter, et al. (2005). "What makes the bioactive lipids phosphatidic acid and lysophosphatidic acid so special?" Biochemistry 44(51): 17007-17015.

Kooijman, E. E., K. E. King, et al. (2009). "Ionization properties of phosphatidylinositol polyphosphates in mixed model membranes." Biochemistry 48(40): 9360-9371.

Kooijman, E. E., S. R. Kuzenko, et al. (2011). "Phosphatidylinositol 4,5-bisphosphate stimulates alveolar epithelial fluid clearance in male and female adult rats." Am J Physiol Lung Cell Mol Physiol 301(5): L804-811.

Kooijman, E. E., D. P. Tieleman, et al. (2007). "An electrostatic/hydrogen bond switch as the basis for the specific interaction of phosphatidic acid with proteins." J Biol Chem 282(15): 11356-11364.

Kooijman, E. E., D. Vaknin, et al. (2009). "Structure of ceramide-1-phosphate at the air-water solution interface in the absence and presence of Ca2+." Biophys J 96(6): 2204-2215.

Koreh, K. and M. E. Monaco (1986). "The relationship of hormone-sensitive and hormone- insensitive phosphatidylinositol to phosphatidylinositol 4,5-bisphosphate in the WRK-1 cell." J Biol Chem 261(1): 88-91.

186

Koushik, A. B., R. R. Powell, et al. (2013). "Localization of phosphatidylinositol 4,5- bisphosphate to lipid rafts and uroids in the human protozoan parasite Entamoeba histolytica." Infect Immun 81(6): 2145-2155.

Kwiatkowska, K. (2010). "One lipid, multiple functions: how various pools of PI(4,5)P(2) are created in the plasma membrane." Cell Mol Life Sci 67(23): 3927-3946.

Lasserre, R., X. J. Guo, et al. (2008). "Raft nanodomains contribute to Akt/PKB plasma membrane recruitment and activation." Nat Chem Biol 4(9): 538-547.

Lemmon, M. A. (2007). "Pleckstrin homology (PH) domains and phosphoinositides." Biochem Soc Symp(74): 81-93.

Levental, I., A. Cebers, et al. (2008). "Combined electrostatics and hydrogen bonding determine intermolecular interactions between polyphosphoinositides." J Am Chem Soc 130(28): 9025-9030.

Levental, I., D. A. Christian, et al. (2009). "Calcium-dependent lateral organization in phosphatidylinositol 4,5-bisphosphate (PIP2)- and cholesterol-containing monolayers." Biochemistry 48(34): 8241-8248.

Levental, I., D. Lingwood, et al. (2010). "Palmitoylation regulates raft affinity for the majority of integral raft proteins." Proc Natl Acad Sci U S A 107(51): 22050-22054.

Li, Z., R. M. Venable, et al. (2009). "Molecular dynamics simulations of PIP2 and PIP3 in lipid bilayers: determination of ring orientation, and the effects of surface roughness on a Poisson-Boltzmann description." Biophys J 97(1): 155-163.

Lin, Q. and E. London (2014). "Preparation of artificial plasma membrane mimicking vesicles with lipid asymmetry." PLoS One 9(1): e87903.

Lindon, J. C., D. J. Baker, et al. (1986). "1H, 13C and 31P n.m.r. spectra and molecular conformation of myo-inositol 1,4,5-triphosphate." Biochem J 233(1): 275-277.

Liu, A. P. and D. A. Fletcher (2006). "Actin polymerization serves as a membrane domain switch in model lipid bilayers." Biophys J 91(11): 4064-4070.

187

Lodish, H. F. B., Arnold; Kaiser, Chris; Krieger, Monty; Bretscher, Anthony; Ploegh, Hidde L.; Amon, Angelika; Scott, Matthew P. (2013). Molecular cell biology.

Loew, S., E. E. Kooijman, et al. (2013). "Increased pH-sensitivity of protein binding to lipid membranes through the electrostatic-hydrogen bond switch." Chem Phys Lipids 169: 9- 18.

London, E. and D. A. Brown (2000). "Insolubility of lipids in triton X-100: physical origin and relationship to sphingolipid/cholesterol membrane domains (rafts)." Biochim Biophys Acta 1508(1-2): 182-195.

Lu, J. X., K. Damodaran, et al. (2005). "Solid-state nuclear magnetic resonance relaxation studies of the interaction mechanism of antimicrobial peptides with phospholipid bilayer membranes." Biochemistry 44(30): 10208-10217.

Lyakhova, T. A. and J. D. Knight (2014). "The C2 domains of granuphilin are high-affinity sensors for plasma membrane lipids." Chem Phys Lipids 182: 29-37.

Markwick, P. R., T. Malliavin, et al. (2008). "Structural biology by NMR: structure, dynamics, and interactions." PLoS Comput Biol 4(9): e1000168.

Marsh, D. (2013). Handbook of Lipid Bilayers. Boca Raton, CRC Press.

Matosevic, S. and B. M. Paegel (2013). "Layer-by-layer cell membrane assembly." Nat Chem 5(11): 958-963.

McLaughlin, S. and D. Murray (2005). "Plasma membrane phosphoinositide organization by protein electrostatics." Nature 438(7068): 605-611.

McLaughlin, S., J. Wang, et al. (2002). "PIP(2) and proteins: interactions, organization, and information flow." Annu Rev Biophys Biomol Struct 31: 151-175.

Murray, D. H. and L. K. Tamm (2011). "Molecular mechanism of cholesterol- and polyphosphoinositide-mediated syntaxin clustering." Biochemistry 50(42): 9014-9022.

188

Nightingale, E. R. (1959). "Phenomenological Theory of Ion Solvation. Effective Radii of Hydrated Ions." The Journal of Physical Chemistry 63(9): 1381-1387.

Ohki, S., M. Muller, et al. (2010). "Surface potential of phosphoinositide membranes: comparison between theory and experiment." Colloids Surf B Biointerfaces 79(1): 210- 218.

Opella, S. J. (2013). "Structure determination of membrane proteins by nuclear magnetic resonance spectroscopy." Annu Rev Anal Chem (Palo Alto Calif) 6: 305-328.

Owen, D. M., D. J. Williamson, et al. (2012). "Sub-resolution lipid domains exist in the plasma membrane and regulate protein diffusion and distribution." Nat Commun 3: 1256.

Pentz, L. and E. R. Thornton (1967). "Isotope Effects on Basicity of 2-Nitrophenoxide 2,4- Dinitrophenoxide Hydroxide and Imidazole in Protium Oxide-Deuterium Oxide Mixtures." J Am Chem Soc 89(26): 6931-&.

Pershan, P. S. and M. L. Schlossman (2012). Liquid Surfaces and Interfaces: Synchrotron X-ray methods. New York, Cambridge University Press.

Pfeiffer, D. R. and H. A. Lardy (1976). "Ionophore A23187 - Effect of H+ Concentration on Complex-Formation with Divalent and Monovalent Cations and Demonstration of K+- Transport in Mitochondria Mediated by A23187." Biochemistry 15(5): 935-943.

Pike, L. J. and L. Casey (1996). "Localization and turnover of phosphatidylinositol 4,5- bisphosphate in caveolin-enriched membrane domains." J Biol Chem 271(43): 26453- 26456.

Pike, L. J. and J. M. Miller (1998). "Cholesterol depletion delocalizes phosphatidylinositol bisphosphate and inhibits hormone-stimulated phosphatidylinositol turnover." J Biol Chem 273(35): 22298-22304.

Pitman, M. C., F. Suits, et al. (2004). "Molecular-level organization of saturated and polyunsaturated fatty acids in a phosphatidylcholine bilayer containing cholesterol." Biochemistry 43(49): 15318-15328.

189

Rauch, M. E., C. G. Ferguson, et al. (2002). "Myristoylated alanine-rich C kinase substrate (MARCKS) sequesters spin-labeled phosphatidylinositol 4,5-bisphosphate in lipid bilayers." J Biol Chem 277(16): 14068-14076.

Redfern, D. A. and A. Gericke (2004). "Domain formation in phosphatidylinositol monophosphate/phosphatidylcholine mixed vesicles." Biophys J 86(5): 2980-2992.

Redfern, D. A. and A. Gericke (2005). "pH-dependent domain formation in phosphatidylinositol polyphosphate/phosphatidylcholine mixed vesicles." J Lipid Res 46(3): 504-515.

Redfern, R. E., D. Redfern, et al. (2008). "PTEN phosphatase selectively binds phosphoinositides and undergoes structural changes." Biochemistry 47(7): 2162-2171.

Reid, D. G. and K. Gajjar (1987). "A proton and carbon 13 nuclear magnetic resonance study of neomycin B and its interactions with phosphatidylinositol 4,5-bisphosphate." J Biol Chem 262(17): 7967-7972.

Ross, A. H. and A. Gericke (2009). "Phosphorylation keeps PTEN phosphatase closed for business." Proc Natl Acad Sci U S A 106(5): 1297-1298.

Rouser, G., S. Fkeischer, et al. (1970). "Two dimensional then layer chromatographic separation of polar lipids and determination of phospholipids by phosphorus analysis of spots." Lipids 5(5): 494-496.

Salvemini, I. L., D. M. Gau, et al. (2014). "Low PIP(2) molar fractions induce nanometer size clustering in giant unilamellar vesicles." Chem Phys Lipids 177: 51-63.

Sarmento, M. J., A. Coutinho, et al. (2013). "Ca induces PI(4,5)P clusters on lipid bilayers at physiological PI(4,5)P and Ca concentrations." Biochim Biophys Acta.

Schink, K. O., C. Raiborg, et al. (2013). "Phosphatidylinositol 3-phosphate, a lipid that regulates membrane dynamics, protein sorting and cell signalling." Bioessays.

Schlewer, G., P. Guedat, et al. (1998). Inositol Phosphates: Inframolecular physico-chemical studies: correlation with binding properties. Phosphoinositides: Chemistry. K. S. Bruzik. Washington, DC., American Chemical Society: 255-270.

190

Schmitt, L., P. Bortmann, et al. (1993). "Myoinositol 1,4,5-Triphosphate and Related- Compounds Protonation Sequence - Potentiometric and P-31 Nmr-Studies." Journal of the Chemical Society-Perkin Transactions 2(11): 2257-2263.

Shaw, J. E., R. F. Epand, et al. (2006). "Tracking peptide-membrane interactions: insights from in situ coupled confocal-atomic force microscopy imaging of NAP-22 peptide insertion and assembly." J Struct Biol 155(3): 458-469.

Shenoy, S., P. Shekhar, et al. (2012). "Membrane association of the PTEN tumor suppressor: molecular details of the protein-membrane complex from SPR binding studies and neutron reflection." PLoS One 7(4): e32591.

Simons, K. and E. Ikonen (1997). "Functional rafts in cell membranes." Nature 387(6633): 569- 572.

Simons, K. and G. van Meer (1988). "Lipid sorting in epithelial cells." Biochemistry 27(17): 6197-6202.

Singer, S. J. and G. L. Nicolson (1972). "The fluid mosaic model of the structure of cell membranes." Science 175(4023): 720-731.

Singh, G., L. Odriozola, et al. (2011). "Characterization of a novel PTEN mutation in MDA-MB- 453 breast carcinoma cell line." BMC Cancer 11: 490.

Slochower, D. R., P. J. Huwe, et al. (2013). "Quantum and All-Atom Molecular Dynamics Simulations of Protonation and Divalent Ion Binding to Phosphatidylinositol 4,5- Bisphosphate (PIP2)." J Phys Chem B 117(28): 8322-8329.

Small, D. M. (1986). The Physical Chemistry of Lipids: From Alkanes to Phospholipids. Handbook of Lipid Research. New York and London, Plenum Press. 4.

Stahelin, R. V., J. L. Scott, et al. (2014). "Cellular and molecular interactions of phosphoinositides and peripheral proteins." Chem Phys Lipids.

Stokoe, D., L. R. Stephens, et al. (1997). "Dual role of phosphatidylinositol-3,4,5-trisphosphate in the activation of protein kinase B." Science 277(5325): 567-570.

191

Suh, B. C. and B. Hille (2008). "PIP2 is a necessary cofactor for ion channel function: how and why?" Annu Rev Biophys 37: 175-195.

Sun, H., R. Lesche, et al. (1999). "PTEN modulates cell cycle progression and cell survival by regulating phosphatidylinositol 3,4,5,-trisphosphate and Akt/protein kinase B signaling pathway." Proc Natl Acad Sci U S A 96(11): 6199-6204.

Swairjo, M. A., B. A. Seaton, et al. (1994). "Effect of vesicle composition and curvature on the dissociation of phosphatidic acid in small unilamellar vesicles--a 31P-NMR study." Biochim Biophys Acta 1191(2): 354-361.

Takasuga, S., Y. Horie, et al. (2013). "Critical roles of type III phosphatidylinositol phosphate kinase in murine embryonic visceral endoderm and adult intestine." Proc Natl Acad Sci U S A 110(5): 1726-1731.

Toner, M., G. Vaio, et al. (1988). "Adsorption of cations to phosphatidylinositol 4,5- bisphosphate." Biochemistry 27(19): 7435-7443.

Tong, J., L. Nguyen, et al. (2008). "Role of GAP-43 in sequestering phosphatidylinositol 4,5- bisphosphate to Raft bilayers." Biophys J 94(1): 125-133.

Truong-Quang, B. A. and P. F. Lenne (2014). "Membrane microdomains: from seeing to understanding." Front Plant Sci 5: 18.

Tsui, F. C., D. M. Ojcius, et al. (1986). "The intrinsic pKa values for phosphatidylserine and phosphatidylethanolamine in phosphatidylcholine host bilayers." Biophys J 49(2): 459- 468.

Vaknin, D. (2012). X-ray Diffraction and Spectroscopic Techniques for Liquid Surfaces and Interfaces. Characterization of Materials. E. N. Kaufmann. New York, John Wiley & Sons. 2: 1393-1423. van Dijck, P. W., B. de Kruijff, et al. (1978). "Phase transitions in phospholipid model membranes of different curvature." Biochim Biophys Acta 506(2): 183-191.

192

van Meer, G., D. R. Voelker, et al. (2008). "Membrane lipids: where they are and how they behave." Nat Rev Mol Cell Biol 9(2): 112-124. van Paridon, P. A., B. de Kruijff, et al. (1986). "Polyphosphoinositides undergo charge neutralization in the physiological pH range: a 31P-NMR study." Biochim Biophys Acta 877(1): 216-219. van Rheenen, J., E. M. Achame, et al. (2005). "PIP2 signaling in lipid domains: a critical re- evaluation." EMBO J 24(9): 1664-1673.

Varnai, P. and T. Balla (2006). "Live cell imaging of phosphoinositide dynamics with fluorescent protein domains." Biochim Biophys Acta 1761(8): 957-967.

Varnai, P., B. Thyagarajan, et al. (2006). "Rapidly inducible changes in phosphatidylinositol 4,5- bisphosphate levels influence multiple regulatory functions of the lipid in intact living cells." J Cell Biol 175(3): 377-382.

Wang, J., A. Gambhir, et al. (2002). "Lateral sequestration of phosphatidylinositol 4,5- bisphosphate by the basic effector domain of myristoylated alanine-rich C kinase substrate is due to nonspecific electrostatic interactions." J Biol Chem 277(37): 34401- 34412.

Wang, J. Y., S. McLaughlin, et al. (2003). "Lateral sequestration of PIP2 by the effector domain of MARCKS." Biophys J 84(2): 461a-461a.

Wang, W., N. A. Anderson, et al. (2012). "Regulation of the electric charge in phosphatidic acid domains." J Phys Chem B 116(24): 7213-7220.

Wang, W. J., N. S. Murthy, et al. (2013). "Structure of Biodegradable Films at Aqueous Surfaces: X-ray Diffraction and Spectroscopy Studies of Polylactides and Tyrosine- Derived Polycarbonates." Langmuir 29(36): 11420-11430.

Wang, Y. H., A. Collins, et al. (2012). "Divalent cation-induced cluster formation by polyphosphoinositides in model membranes." J Am Chem Soc 134(7): 3387-3395.

193

Wang, Y. H., D. R. Slochower, et al. (2014). "Counterion-mediated cluster formation by polyphosphoinositides." Chem Phys Lipids.

Watts, A. (1998). "Solid-state NMR approaches for studying the interaction of peptides and proteins with membranes." Biochim Biophys Acta 1376(3): 297-318.

Wolf, F. I. and A. Cittadini (2003). "Chemistry and biochemistry of magnesium." Mol Aspects Med 24(1-3): 3-9.

Wuthrich, K. (1990). "Protein structure determination in solution by NMR spectroscopy." J Biol Chem 265(36): 22059-22062.

Wuthrich, K. (2003). "NMR studies of structure and function of biological macromolecules (Nobel Lecture)." J Biomol NMR 27(1): 13-39.

Wüthrich, K. (1986). NMR of proteins and nucleic acids, Wiley.

Yun, W. B. and J. M. Bloch (1990). "X-Ray near Total External Fluorescence Method - Experiment and Analysis." Journal of Applied Physics 68(4): 1421-1428.

Zhendre, V., A. Grelard, et al. (2011). "Key role of polyphosphoinositides in dynamics of fusogenic nuclear membrane vesicles." PLoS One 6(9): e23859.

194

Appendix List of Materials

Chemical Name Abbreviation Chemical Company Formula Purity Formula Weight 1,2,-dioleoyl-sn-glycero- DOPC C44H84NO8P Avanti 786.113 >99% 3-phosphocholine polar Lipids 1-palmitoyl-2-oleoyl-sn- POPC C42H82NO8P Avanti 760.076 >99% glycero-3- polar phosphocholine Lipids 1,2,-dioleoyl-sn-glycero- DOPE C41H78NO8P Avanti 744.034 >99% 3-phosphoethanolamine polar Lipids 1,2-dioleoyl-sn-glycero-3- DOPS C42H77NO10PNa Avanti 810.025 >99% phospho-L-serine (sodium polar salt) Lipids L-α-phosphatidylinositol- bPI(4,5)P2 C47H94N3O19P3 Avanti 1096.385 >99% 4,5-bisphosphate (Brain, polar Porcine) (ammonium salt) Lipids 1,2-dioleoyl-sn-glycero-3- RhB DOPE C68H109N4O14PS2 Avanti 1301.715 >99% phosphoethanolamine-N- polar (lissamine rhodamine B Lipids sulfonyl) (ammonium salt) L-α-phosphatidylinositol liver PI C47H82Na O13P Avanti 902.133 >99% (Liver, Bovine) (sodium polar salt) Lipids HPLC Chloroform chloroform CHCl3 Fisher 119.38 99.8% Scientific HPLC Methanol MeOH CH4O Fisher 32.04 99.9% Scientific HPLC Water water H2O Fisher 18.02 >99.9 Scientific % Citric Acid citric acid C6H8O7 Sigma 192.12 99.5% Aldrich 2-(N- MES C6H13NO4S * Sigma 195.24 99.5% Morpholino)ethanesulfo xH2O Aldrich nic acid hydrate 4-(2- HEPES C8H18N2O4S Amresco 238.30 99.7% Hydroxyethyl)piperasine -1-ethanesulfonic acid Glycine glycine C2H5NO2 Amresco 75.07 99% Sodium Chloride NaCl NaCl Sigma 58.44 99.5%

Aldrich Ethylenediaminetetra EDTA C10H16N2O8 Sigma 292.24 99.995 acetic acid Aldrich % Ethanol EtOH C2H6O BDH 46.07 >94% Phosphoric acid H3PO4 H3PO4 1,4-Piperazinediethane PIPES C8H18N2O6S2 Sigma 302.37 99% sulfonic acid Aldrich Potassium phosphate, KH2PO4 KH2PO4 BDH 136.09 99% monobasic Sodium phosphate Na2HPO4 Na2HPO4 BDH 141.96 >98% dibasic Cholesterol (ovine wool) cholesterol C27H46O Avanti 386.654 >98% polar Lipids Phosphatidylinositol plant PI C45H78O13PNa Matreya 858 + Na 98% (sodium salt) Calcium Ionophore A23187 C29H37N3O6 Sigma 523.62 >98% A23187 Aldrich Calcium Chloride CaCl2 CaCl2 Sigma 110.98 99.99 Aldrich % Magnesium Chloride MgCl2 MgCl2 Sigma 95.21 99.99 Aldrich % Nickel Chloride NiCl2 NiCl2 Barium Chloride BaCl2 BaCl2 Manganese Chloride MnCl2 MnCl2 Hydrochloric acid (34%) HCl HCl Sigma 36.46 <10ppt Aldrich Sodium Hydroxide NaOH NaOH Alfa Aesar 39.997 98% Deuterium oxide D2O D2O Sigma 20.03 99.9% Aldrich Trizma Base Tris C4H11NO3 Sigma 121.14 99.9% Aldrich Potassium chloride KCl KCl Sigma 74.55 99% Aldrich SRNKRRY PTEN10-16 C42H73N19O11 LifeTein 1020.17 >95% PI(4,5)P2-amine PI(4,5)P2- C12H23NP3O13N Michael 628.98 amine a5 Best (University of Tennessee) NKRRYQE PTEN12-18 C43H71N17O13 LifeTein 1034.14 >95%

196

31P MAS NMR pulse program and experimental parameters

Pulse Program:

;zg ;avance-version (06/11/09) ;1D sequence ; ;$CLASS=HighRes ;$DIM=1D ;$TYPE= ;$SUBTYPE= ;$COMMENT=

#include

"acqt0=-p1*2/3.1416"

1 ze 2 30m d1 p1 ph1 go=2 ph31 30m mc #0 to 2 F0(zd) exit

ph1=0 2 2 0 1 3 3 1 ph31=0 2 2 0 1 3 3 1

;pl1: f1 channel - power level for pulse (default) ;p1 : f1 channel - high power pulse ;d1 : relaxation delay; 1-5 * T1 ;NS : 1 * n, total number of scans: NS * TD0

;$Id: zg,v 1.9 2006/11/10 10:56:44 ber Exp $

197

Experimental parameters: NAME S1P1 EXPNO 12 PROCNO 1 Date_ 20091116 Time 12.07 INSTRUM spect PROBHD 5 mm TXI 1H-13 PULPROG zgig TD 8192 SOLVENT NA NS 12996 DS 4 SWH 81521.742 Hz FIDRES 9.951385 Hz AQ 0.0502943 sec RG 2050 DW 6.133 usec DE 10.00 usec TE 294.0 K D1 1.00000000 sec D11 0.03000000 sec TD0 75 ======CHANNEL f1 ======NUC1 31P P1 5.25 usec PL1 8.00 dB PL1W 62.50282669 W SFO1 161.9765351 MHz ======CHANNEL f2 ======CPDPRG2 spinal64 NUC2 1H PCPD2 22.25 usec PL2 120.00 dB PL12 14.00 dB PL2W 0.00000000 W PL12W 2.95000005 W SFO2 400.1360019 MHz SI 8192 SF 161.9766731 MHz WDW EM SSB 0 LB 50.00 Hz GB 0 PC 4.00

198

Static 31P NMR pulse program

Pulse Program:

;zgig ;avance-version (07/04/03) ;1D sequence with inverse gated decoupling ; ;$CLASS=HighRes ;$DIM=1D ;$TYPE= ;$SUBTYPE= ;$COMMENT=

#include

"d11=30m"

"acqt0=-p1*2/3.1416"

1 ze d11 pl12:f2 2 30m do:f2 d1 p1 ph1 go=2 ph31 cpd2:f2 30m do:f2 mc #0 to 2 F0(zd) exit

ph1=0 2 2 0 1 3 3 1 ph31=0 2 2 0 1 3 3 1

;pl1 : f1 channel - power level for pulse (default) ;pl12: f2 channel - power level for CPD/BB decoupling ;p1 : f1 channel - high power pulse ;d1 : relaxation delay; 1-5 * T1 ;d11: delay for disk I/O [30 msec] ;NS: 1 * n, total number of scans: NS * TD0 ;cpd2: decoupling according to sequence defined by cpdprg2 ;pcpd2: f2 channel - 90 degree pulse for decoupling sequence

199

;$Id: zgig,v 1.9 2007/04/11 13:34:31 ber Exp $

PIP Ionization Fitting Files

PC-PI(4,5)P2.m m-file for fitting PI(4,5)P2

%% This m-file will fit the PC:PI(4,5)P2 data close all clear all clc

%% data pH = [4.4, 4.7, 5.23, 5.82, 6.15, 6.66, 6.72, 7.11, 7.63, 7.98, 8.43, 8.7, 9.32];

P4 = [1.5142, 1.5742, 1.826, 2.2805,2.7302,3.2565,3.2774,3.5106,4.0292,4.3576,4.6702,4.7281,4.8957];

P5 = [0.6801, 0.7233, 0.9203, 1.2656, 1.6285, 2.1424, 2.1002, 2.46, 3.2774, 3.8286, 4.3712, 4.4567, 4.7603];

% assume +/- 0.1 error P4_err = 0.06*ones(size(P4)); P5_err = 0.06*ones(size(P5));

% PI(4,5)P2(pH,P4,P5,P4_err,P5_err);

%% fitting P4 % A = P4H,P5H; B = P4-P5H; C = P4HP5-; D = P4-P5-; Param_PI(4,5)P2 = [min(P4), min(P5), max(P4), max(P5), 6, 6, 7]; LB = [min(P4)-0.1, min(P5)-1, max(P4)-0.1, max(P5)-0.1, 3, 3, 4]; UB = [min(P4)+0.1, min(P5)+0.1, max(P4)+1, max(P5)+0.1, 8, 8, 10]; PI(4,5)P2y = [P4;P5]; xfit = [4:0.02:10]; % [p_opt, chisq] = lsqcurvefit('PI(4,5)P2_pKa_fit', Param_PI(4,5)P2, pH, [P4;P5], LB, UB); [p_opt, chisq, p_err, dchisq] = err_fit(@PI(4,5)P2_pKa_fit, Param_PI(4,5)P2, pH, [P4;P5], [P4_err;P5_err], LB, UB); [PI(4,5)P2_fit, pKas] = PI(4,5)P2_pKa_fit(p_opt, xfit); pKas(3) = 10^-pKas(3); K_3 = 10^(pKas(1)-pKas(2));

200

% pKas_err = [0.07; 0.07; 0; 0.06; 0;]; pKas_err = [p_err(5); p_err(6); 0; p_err(7); 0]; pKas_err(3) = sqrt(pKas_err(2)^2 + pKas_err(1)^2); pKas_err(5) = sqrt(pKas_err(3)^2 + pKas_err(4)^2);

%% calculate ratios BoA = 10.^(xfit - pKas(1)); CoA = 10.^(xfit - pKas(2)); DoA = 10.^(2.*xfit - pKas(1) - pKas(4)); ToA = 1 + BoA + CoA + DoA; f_A = 1./ToA; f_B = BoA./ToA; f_C = CoA./ToA; f_D = DoA./ToA; f_P5p = f_A + f_B; f_P4p = f_A + f_C; ch_P4 = f_P4p - 2; ch_P5 = f_P5p - 2; ch_T = ch_P4 + ch_P5 - 1; ch_T2 = -4.*f_D + -3.*f_B + -3.*f_C + -2.*f_A - 1;

%% plot PI(4,5)P2 fit close all figure hold on plot(xfit,PI(4,5)P2_fit(1,:),'b-','linewidth', 3); plot(xfit,PI(4,5)P2_fit(2,:),'r-','linewidth', 3);

H1 = plot(pH, P4, 'bo'); set(H1, 'LineWidth',3,'MarkerSize',12,'MarkerFaceColor','w'); H2 = plot(pH, P5, 'rv'); set(H2, 'LineWidth',3,'MarkerSize',12,'MarkerFaceColor','w'); hold off set(gca,'box','on','LineWidth',3); set(gca,'FontSize',30,'FontName','times'); set(gca, 'xlim', [3,11], 'ylim', [0.25,5.5]); axis square

%% plot PI(4,5)P2 charge figure hold on plot(xfit,ch_P4,'b-','linewidth', 3); plot(xfit,ch_P5,'r-','linewidth', 3); hold off set(gca,'box','on','LineWidth',3); set(gca,'FontSize',30,'FontName','times'); set(gca, 'xlim', [4,10], 'ylim', [-2,-1]); axis square

201

%% plot PI(4,5)P2 fractions figure hold on plot(xfit,f_A,'k--','linewidth', 3); plot(xfit,f_B,'b--','linewidth', 3); plot(xfit,f_C,'r--','linewidth', 3); plot(xfit,f_D,'k-','linewidth', 3); hold off set(gca,'box','on','LineWidth',3); set(gca,'FontSize',30,'FontName','times'); set(gca, 'xlim', [4,10], 'ylim', [0,1]); axis square

PI(4,5)P2_pKa_fit.m m-file with PI(4,5)P2 fitting function

%% PI(4,5)P2 pKa fitting function function [out, pKs] = PI(4,5)P2_pKa_fit(Param, x) dA = Param(1); dB = Param(3); dC = Param(1); dD = Param(3); pH = x; pKa1 = Param(5); pKa2 = Param(6); pKa4 = Param(7); pKa3 = pKa2 - pKa1; pKa5 = pKa4 - pKa3; pKs = [pKa1; pKa2; pKa3; pKa4; pKa5]; y1 = (dA + dB.*10.^(pH - pKa1) + dC.*10.^(pH - pKa2) + dD.*10.^(2.*pH - pKa1 - pKa4))./(1 + 10.^(pH - pKa1) + 10.^(pH - pKa2) + 10.^(2.*pH - pKa1 - pKa4)); dA = Param(2); dB = Param(2); dC = Param(4); dD = Param(4); y2 = (dA + dB.*10.^(pH - pKa1) + dC.*10.^(pH - pKa2) + dD.*10.^(2.*pH - pKa1 - pKa4))./(1 + 10.^(pH - pKa1) + 10.^(pH - pKa2) + 10.^(2.*pH - pKa1 - pKa4)); out = [y1;y2];

PIP3.m m-file for fitting PI(3,4,5)P3

%% This m-file will fit the PC:PIP3 data

202

close all clear all clc

%% data pH = [4.33, 4.73, 5.72, 6.15, 6.74, 7.06, 7.18, 7.6, 7.9, 8.4, 9.34, 10.24, 11.3];

P3 = [0.1615, 0.2027, 0.4945, 0.7433, 1.0363, 1.424, 1.6463, 2.0417, 2.4153, 2.8114, 3.2386, 3.5495, 3.841];

P4 = [1.0586, 1.1088, 1.5648, 1.8872, 2.1178, 2.0576, 2.0915, 2.0417, 1.9001, 1.8396, 1.903, 2.86, 3.3];

P5 = [0.4813, 0.5389, 0.8639, 1.1421, 1.5091, 2.0576, 2.23, 2.8511, 3.2976, 3.8478, 4.3627, 4.6546, 4.9388];

% assume +/- 0.1 error P3_err = 0.06*ones(size(P3)); P4_err = 0.06*ones(size(P4)); P5_err = 0.06*ones(size(P5));

%% fitting PIP3 % A = P4H,P5H; B = P4-P5H; C = P4HP5-; D = P4-P5-; Param_PIP3 = [min(P3), min(P4), min(P5), max(P3), max(P4), max(P5), 6, 6, 6, 7, 7, 7, 8];

LB = [min(P3)-0.1, min(P4)-0.1, min(P5)-0.1, max(P3)-0.1, max(P4)-0.1, max(P5)-0.1 3, 3, 3, 3, 3, 3, 3]; UB = [min(P3)+0.1, min(P4)+0.1, min(P5)+0.1 max(P3)+1, max(P4)+0.1, max(P5)+0.1, 10, 10, 10, 10, 10, 10, 10]; PI(4,5)P2y = [P3;P4;P5];

%% fit simultaneously xfit = [4:0.02:11.5]; % [p_opt, chisq] = lsqcurvefit('PIP3_pKa_fit', Param_PIP3, pH, [P3;P4;P5], LB, UB); [p_opt, chisq, p_err, dchisq] = err_fit(@PIP3_pKa_fit, Param_PIP3, pH, [P3;P4;P5], [P3_err;P4_err;P5_err], LB, UB); [PIP3_fit, pKas] = PIP3_pKa_fit(p_opt, xfit); pKas(4) = 10^-pKas(4); pKas(5) = 10^-pKas(5); pKas(6) = 10^-pKas(6); pKas(13) = 10^-pKas(13); pKas(14) = 10^-pKas(14); pKas(15) = 10^-pKas(15);

%% 203

pKas_err = [p_err(7); p_err(8); p_err(9); 0; 0; 0; p_err(10); p_err(11); 0; p_err(12); 0; 0; 0; 0; 0; p_err(13); 0; 0]; pKas_err(4) = sqrt(pKas_err(2)^2 + pKas_err(1)^2); pKas_err(5) = sqrt(pKas_err(3)^2 + pKas_err(1)^2); pKas_err(6) = sqrt(pKas_err(2)^2 + pKas_err(3)^2); pKas_err(9) = sqrt(pKas_err(7)^2 + pKas_err(2)^2 + pKas_err(1)^2); pKas_err(11) = sqrt(pKas_err(8)^2 + pKas_err(3)^2 + pKas_err(1)^2); pKas_err(12) = sqrt(pKas_err(10)^2 + pKas_err(2)^2 + pKas_err(3)^2); pKas_err(13) = sqrt(pKas_err(8)^2 + pKas_err(7)^2); pKas_err(14) = sqrt(pKas_err(10)^2 + pKas_err(9)^2); pKas_err(15) = sqrt(pKas_err(12)^2 + pKas_err(11)^2); pKas_err(17) = sqrt(pKas_err(7)^2 + pKas_err(8)^2 + pKas_err(16)^2); pKas_err(18) = sqrt(pKas_err(16)^2 + pKas_err(14)^2);

%% calculate ratios r3 = 10.^(xfit - pKas(1)); r4 = 10.^(xfit - pKas(2)); r5 = 10.^(xfit - pKas(3)); r34 = 10.^(2.*xfit - pKas(1) - pKas(7)); r45 = 10.^(2.*xfit - pKas(2) - pKas(10)); r35 = 10.^(2.*xfit - pKas(1) - pKas(8)); r345 = 10.^(3.*xfit - pKas(1) - pKas(7) - pKas(16)); rT = 1 + r3 + r4 + r5 + r34 + r45 + r35 + r345; f_0 = 1./rT; f_3 = r3./rT; f_4 = r4./rT; f_5 = r5./rT; f_34 = r34./rT; f_45 = r45./rT; f_35 = r35./rT; f_345 = r345./rT; f_P3p = f_0 + f_4 + f_5 + f_45; f_P4p = f_0 + f_3 + f_5 + f_35; f_P5p = f_0 + f_3 + f_4 + f_34; ch_P3 = f_P3p - 2; ch_P4 = f_P4p - 2; ch_P5 = f_P5p - 2; ch_T = ch_P3 + ch_P4 + ch_P5 - 1;

%% plot PIP3 fit close all figure hold on plot(xfit,PIP3_fit(1,:),'g-','linewidth', 3); plot(xfit,PIP3_fit(2,:),'b-','linewidth', 3); plot(xfit,PIP3_fit(3,:),'r-','linewidth', 3);

H1 = plot(pH, P3, 'gs'); set(H1, 'LineWidth',3,'MarkerSize',12,'MarkerFaceColor','w'); H2 = plot(pH, P4, 'bo'); set(H2, 'LineWidth',3,'MarkerSize',12,'MarkerFaceColor','w'); H3 = plot(pH, P5, 'rv'); set(H3, 'LineWidth',3,'MarkerSize',12,'MarkerFaceColor','w'); 204

hold off set(gca,'box','on','LineWidth',3); set(gca,'FontSize',30,'FontName','times'); set(gca, 'xlim', [3,12], 'ylim', [0,5.5]); axis square

%% plot PI(4,5)P2 charge figure hold on plot(xfit,ch_P3,'g-','linewidth',3); plot(xfit,ch_P4,'b-','linewidth', 3); plot(xfit,ch_P5,'r-','linewidth', 3); hold off set(gca,'box','on','LineWidth',3); set(gca,'FontSize',30,'FontName','times'); set(gca, 'xlim', [4,11.5], 'ylim', [-2,-1]); set(gca, 'xtick', [5,7,9,11]); axis square

%% plot PI(4,5)P2 fractions figure hold on plot(xfit,f_0,'k--','linewidth', 3); plot(xfit,f_3,'g--','linewidth', 3); plot(xfit,f_4,'b--','linewidth', 3); plot(xfit,f_5,'r--','linewidth', 3); plot(xfit,f_34,'-', 'color', [0 0.5 0.5], 'linewidth', 3); plot(xfit,f_45,'-','color', [0.5 0 0.5], 'linewidth', 3); plot(xfit,f_35,'-','color', [0.5 0.5 0], 'linewidth', 3); plot(xfit,f_345,'k-','linewidth', 3); hold off set(gca,'box','on','LineWidth',3); set(gca,'FontSize',30,'FontName','times'); set(gca, 'xlim', [4,11.5], 'ylim', [0,1]); set(gca, 'xtick', [3,5,7,9,11]); axis square

PIP3_pKa_fit.m PI(3,4,5)P3 fitting function

%% PIP3 pKa fitting function function [out, pKs] = PIP3_pKa_fit(Param, x) d0 = Param(1); d3 = Param(4); d4 = Param(1); d5 = Param(1); d34 = Param(4); d45 = Param(1); d35 = Param(4); 205

d345 = Param(4); pH = x; pKs(1) = Param(7); pKs(2) = Param(8); pKs(3) = Param(9); pKs(4) = pKs(2) - pKs(1); pKs(5) = pKs(3) - pKs(1); pKs(6) = pKs(3) - pKs(2); pKs(7) = Param(10); pKs(8) = Param(11); pKs(9) = pKs(7) + pKs(1) - pKs(2); pKs(10) = Param(12); pKs(11) = pKs(8) + pKs(1) - pKs(3); pKs(12) = pKs(10) + pKs(2) - pKs(3); pKs(13) = pKs(8) - pKs(7); pKs(14) = pKs(10) - pKs(9); pKs(15) = pKs(12) - pKs(11); pKs(16) = Param(13); pKs(17) = pKs(16) + pKs(7) - pKs(8); pKs(18) = pKs(16) + pKs(7) + pKs(1) - pKs(10) - pKs(2); num = d0 + d3.*10.^(pH - pKs(1)) + d4.*10.^(pH - pKs(2)) + d5.*10.^(pH - pKs(3)) + d34.*10.^(2.*pH - pKs(1) - pKs(7)) + d45.*10.^(2.*pH - pKs(2) - pKs(10)) + d35.*10.^(2.*pH - pKs(1) - pKs(8)) + d345.*10.^(3.*pH - pKs(1) - pKs(7) - pKs(16)); denom = 1 + 10.^(pH - pKs(1)) + 10.^(pH - pKs(2)) + 10.^(pH - pKs(3)) + 10.^(2.*pH - pKs(1) - pKs(7)) + 10.^(2.*pH - pKs(2) - pKs(10)) + 10.^(2.*pH - pKs(1) - pKs(8)) + 10.^(3.*pH - pKs(1) - pKs(7) - pKs(16)); y1 = num./denom; d0 = Param(2); d3 = Param(2); d4 = Param(5); d5 = Param(2); d34 = Param(5); d45 = Param(5); d35 = Param(2); d345 = Param(5); num = d0 + d3.*10.^(pH - pKs(1)) + d4.*10.^(pH - pKs(2)) + d5.*10.^(pH - pKs(3)) + d34.*10.^(2.*pH - pKs(1) - pKs(7)) + d45.*10.^(2.*pH - pKs(2) - pKs(10)) + d35.*10.^(2.*pH - pKs(1) - pKs(8)) + d345.*10.^(3.*pH - pKs(1) - pKs(7) - pKs(16)); denom = 1 + 10.^(pH - pKs(1)) + 10.^(pH - pKs(2)) + 10.^(pH - pKs(3)) + 10.^(2.*pH - pKs(1) - pKs(7)) + 10.^(2.*pH - pKs(2) - pKs(10)) + 10.^(2.*pH - pKs(1) - pKs(8)) + 10.^(3.*pH - pKs(1) - pKs(7) - pKs(16)); y2 = num./denom;

d0 = Param(3); d3 = Param(3); d4 = Param(3); 206

d5 = Param(6); d34 = Param(3); d45 = Param(6); d35 = Param(6); d345 = Param(6); num = d0 + d3.*10.^(pH - pKs(1)) + d4.*10.^(pH - pKs(2)) + d5.*10.^(pH - pKs(3)) + d34.*10.^(2.*pH - pKs(1) - pKs(7)) + d45.*10.^(2.*pH - pKs(2) - pKs(10)) + d35.*10.^(2.*pH - pKs(1) - pKs(8)) + d345.*10.^(3.*pH - pKs(1) - pKs(7) - pKs(16)); denom = 1 + 10.^(pH - pKs(1)) + 10.^(pH - pKs(2)) + 10.^(pH - pKs(3)) + 10.^(2.*pH - pKs(1) - pKs(7)) + 10.^(2.*pH - pKs(2) - pKs(10)) + 10.^(2.*pH - pKs(1) - pKs(8)) + 10.^(3.*pH - pKs(1) - pKs(7) - pKs(16)); y3 = num./denom; out = [y1;y2;y3];

207

T1 Pulse Program

;t1ir ;avance-version (12/01/11) ;T1 measurement using inversion recovery ; ;$CLASS=HighRes ;$DIM=2D ;$TYPE= ;$SUBTYPE= ;$COMMENT=

#include

"p2=p1*2" "d11=30m"

"acqt0=-p1*2/3.1416"

1 ze 2 d1 p2 ph1 vd p1 ph2 go=2 ph31 d11 wr #0 if #0 ivd lo to 1 times td1 exit

ph1=0 2 ph2=0 0 2 2 1 1 3 3 ph31=0 0 2 2 1 1 3 3

;pl1 : f1 channel - power level for pulse (default) ;p1 : f1 channel - 90 degree high power pulse ;p2 : f1 channel - 180 degree high power pulse ;d1 : relaxation delay; 1-5 * T1 ;d11: delay for disk I/O [30 msec] ;vd : variable delay, taken from vd-list 208

;ns: 8 * n ;ds: 4 ;td1: number of experiments = number of delays in vd-list ;FnMODE: undefined

;define VDLIST

;this pulse program produces a ser-file (PARMOD = 2D)

;$Id: t1ir,v 1.12.8.1 2012/01/31 17:56:37 ber Exp $

Delay list 10, 0.001, 7.5, 10m, 6, 0.05, 5, 0.1, 3.5, 0.2, 2.5, 0.5, 2, .75, 1.25, 1

209

210

Figure A1

Figure A1. MAS 31P NMR spectra for a 1:1 mixture of PC and PE, at pH ~7 The PE phosphodiester has a slightly higher intensity than the PC phosphodiester due to overlap with the asymmetric tail of the PC peak and its reduced broadening (lower FWHM).

211

Figure A2

A B

212

31 Figure A2. Solid state static P NMR spectra for ternary mixtures of PC, PI(4,5)P2, and PE, PS, or PI at three pH values covering the pH range investigated in the pH titration curves. A, PC / PE /

PI(4,5)P2 (47.5% / 47.5% / 5%), B, PC/PS/PI(4,5)P2 (78% / 20% / 2%), and C, PC/PI/PI(4,5)P2 (88% / 10% / 2%).

Figure A3

Figure A3. Ionization behavior of PS as determined by solid state 31P NMR.

The chemical shift of the phosphodiester of PS is recorded as a function of pH for membranes of three different lipid compositions as indicated. The model membrane systems were composed of PC/PS (80% /

20%), PC/PS/PI(4,5)P2 (78% / 20% / 2%), and PC/PS/PA (75% / 20% / 5%) molar ratios. The solid line is

213

a fit to an equation derived from the Henderson-Hasselbalch equation as described previously (Kooijman et al., 2005). Chemical shift is recorded relative to an external 85% H3PO4 standard.

Figure A4

Figure A4. MAS 31P NMR spectra for MLVs containing varying ratios of PC and PI, at pH ~7. The PI phosphodiester peak is at -0.05 ppm in 100% PI vesicles (black), but shifts to lower ppm values as the ratio of PC increases. At a 1:1 ratio (blue) the PI phosphodiester can be observed as a large shoulder overlapping with the PC phosphodiester. As the PC content increases the PI phosphodiester merges further with the PC peak. A possible explanation for this behavior is that as the PC concentration is increasing, the intermolecular interaction between PI hydroxyl groups and phosphodiester groups diminishes.

214

PI(4,5)P2 P4 P5 PC

215

31 Figure A5: P MAS NMR spectra and pH titration curves for PC / PI / PI(4,5)P2 (78% / 31 20% / 2%). A) P MAS NMR spectra as a function of pH for 2 mol% PI(4,5)P2 in 78% PC / 20% PI vesicles. b). Peak positions of the 4- and 5-phosphate of PI(4,5)P2 as a function of pH. c). Ionization model and pKa values for deprotonation of PI(4,5)P2 with PI. d) Charge of the 4- and

5-phosphate of PI(4,5)P2 as a function of pH. e) Plot showing the relative prevalence of each ionization state as a function of pH. f0 is the protonated fraction, f4 is the fraction with P4 deprotonated, f5 is the fraction with P5 deprotonated, and f45 is the fully deprotonated fraction.

216

Figure A6

31 Figure A6. Comparison of P MAS NMR PI(4,5)P2 titration curves for two concentrations of PI. Shown are curves for the 4, and 5-phosphate of PI(4,5)P2 in mixtures of PC and PC/PI at the indicated concentrations of PI (compilation of the data from figure 4-6 and A5).

217

Figure A7

2+ 31 Figure A7: pH dependent effect of Ca on ionization of PI(4,5)P2. a). Representative P

MAS NMR spectra as a function of pH for 3.5 mol% PI(4,5)P2 in 96.5% PC vesicles in the 2+ presence of 1 mM Ca . b) Peak values for the 4- and 5-phosphates of PI(4,5)P2 as a function of pH. Black symbols indicate peak values for 3.5 mol% PI(4,5)P2 in 96.5% PC vesicles in the presence of 1 mM Ca2+, while the blue and the red are the peak values of the 4- and 5-phosphates 2+ of 5 mol% PI(4,5)P2 in 95% PC vesicles respectively in the absence of Ca (Data for 95%/5%

PC/PI(4,5)P2 vesicles taken with permission from Kooijman et al. (Kooijman, King et al. 2009)).

218

Figure A8

219

220

221

222

223

224

31 Figure A8: Solid-state Static P NMR spectra for PC / PI(4,5)P2 vesicles with varying 2+ composition. a). PC / PI(4,5)P2 96.5%/3.5% vesicles in the presence of varying amounts of Ca . 2+ b). PC / PI(4,5)P2 96.5%/3.5% vesicles in the presence of varying amounts of Mg . c). PC /

Cholesterol / PI(4,5)P2 vesicles with varying cholesterol content. d). PC / PE / Cholesterol /

PI(4,5)P2 vesicles with varying PE and cholesterol content. e). PC / PI / Cholesterol / PI(4,5)P2 vesicles with varying PI and cholesterol content. f). PC / Cholesterol / PI(4,5)P2 58% / 40% / 2% 2+ 2+ vesicles in the presence of no divalent cations, 1 mM Mg , or 1 mM Ca . g) PC / PI(4,5)P2 96.5%/3.5% vesicles in the presence of 1 mM Ba2+ or Ni2+.

225

Figure A9

31 Figure A9: P MAS NMR chemical shift bar graph for PC / PI(4,5)P2 vesicles with varying amounts of PE and cholesterol at pH 5. a). Peak values for the 4- and 5-phosphates of 31 PI(4,5)P2 from P MAS NMR spectra PC / PI(4,5)P2 vesicles with 0 mol% cholesterol, 40 mol% cholesterol, 47.5 mol% PE, and 29 mol% PE and 40mol% cholesterol. Mixtures contained 2-5 mol% PI(4,5)P2 and the remainder was PC. For all mixture the pH was ~5.

226

Figure A10

Figure A10. 31P NMR spectra of PI(4,5)P -amine in the presence of varying PTEN . 2 10-16 PI(4,5)P2-amine and PTEN10-16 were mixed together in five different molar ratios—1:0, 10:1, 5:1, 2:1, and 1:1. 31P NMR spectra were acquired for each mixture and overlayed. Intensity values were normalized by the highest peak (P4). A close up of the phosphodiester peak is shown.

31 Figure A11. P NMR spectra of PI(4,5)P2-amine in the presence of varying PTEN10-16. PI(4,5)P2-amine and PTEN10-16 were mixed together in five different molar ratios—1:0, 10:1, 5:1, 2:1, and 1:1. 31P NMR spectra were acquired for each mixture and overlayed. Intensity values were normalized by the highest peak (P4). A close up of the phosphodiester peak is shown.

227

Figure A11

Figure A11. PTEN10-16 TOCSY in the presence of PI(4,5)P2-amine. TOCSY spectra of a 5 mM PTEN10-16 solution at pH 5 and 100 mM NaCl with (black) and without (red) a equimolar amount of PI(4,5)P2-amine.

228

Figure A12

Figure A12. PTEN12-18 TOCSY in the presence of PI(4,5)P2-amine. TOCSY spectra of a 5 mM PTEN12-18 solution at pH 5 and 100 mM NaCl with (black) and without (red) a equimolar amount of PI(4,5)P2-amine.

229

Figure A13

Figure A13. DSC of PC:Cholesterol:PI:PI(4,5)P2 vesicles. DSC scans acquired of multilamellar vesicles composed of PC / Cholesterol / PI / PI(4,5)P2 (45% / 40% / 10% / 5%) in pH 7.2 buffer wih 100 mM HEPES, 100 mM NaCl, and (A) 0 mM or (B) 1 mM CaCl2.

230