Lipid organization in biomimetic model systems –

oligomerization, domain formation, and diffusivity

Ilya Levental

A DISSERTATION

in

BIOENGINEERING

Presented to the Faculties of the University of Pennsylvania in partial fulfillment of the requirements for the Degree of Philosophy

2008

Supervisor of Dissertation: _______________

Graduate Group Chair: _______________

Acknowledgements

Firstly, I would like to thank my committee members, including Dr. Mark Lemmon and the committee chair Dr. Dennis Discher, for facilitating this process and making the writing and presenting of this dissertation much easier and more useful than I ever imagined it would be.

Specifically, I would like to thank Dr. Tobias Baumgart, who has been somewhat of a junior mentor to me in the last two years of my graduate experience, showing me the important (stressful, uncertain) side of academia through the eyes of a junior faculty member.

As the most important and influential non-personal acknowledgement, I would like to sincerely thank my advisor Dr. Paul Janmey. The breadth and depth of his scientific expertise continues to amaze me, even after 5 years of listening to him talk. His unique ability to understand and interpret data from a variety of disparate techniques and projects is, amazingly, surpassed by his ability to take that data and fit it into a holistic picture of that particular project, and the entire lab’s direction and focus. Finally, I can say with confidence that his laissez-faire approach to mentoring has been the single most formative thing about my graduate experience and my current incarnation as a scientist-in-training, despite the fact that I complained about it incessantly.

I would also like to thank all my coworkers from the Janmey lab for being unfailingly helpful, friendly, enthusiastic and supportive, and making work not seem like work at all. In particular, I would like to thank Dr. Robert Bucki for welcoming me into the lab, introducing me to the exciting world of lipid biology, and his consistent confidence in my scientific ability and intellect, in spite of all available evidence to the contrary (at least early on).

Finally, for some personal acknowledgements, I would like to thank all the great friends that I have made during my graduate career at UPenn. Without naming anyone in particular in fear of forgetting someone important, I would just like to thank all of them for making this experience much more fun than it ever had the right to be. Grad school is supposed to be painful and scientists are supposed to be boring, but it wasn’t and it was largely because they weren’t.

Most importantly, I’d like to thank my awesome wife Dr. Kandice Levental. This is not the forum for an appropriate acknowledgement of everything that she has meant to me through this experience, but I would like to thank her for making my work almost entirely stress-free, because as long as she’s around, it doesn’t seem that important.

Oh, and finally, I’d like to thank you, the reader, for actually being interested in my ridiculously long and boring dissertation.

ii

Abstract

Lipid organization in model systems of varying complexity –

oligomerization, domain formation, and diffusivity

By - Ilya Levental

Thesis Advisor - Dr. Paul Janmey

Although significant evidence exists confirming the central role of membrane lipids in

the regulation of numerous cellular functions, much more is known about their

biochemical activity and regulation than about their physical structure and configuration

in native environments, or how this organization affects their functionality. Two distinct

contexts in which lipid organization is proposed to play important roles in functional

regulation of cell behavior are: (1) phase separation of the plasma membrane into

immiscible liquid phases, termed membranes rafts; and (2) localized enrichment of

phosphatidylinositol-bisphosphate (PIP2). In this dissertation, lateral lipid

inhomogeneity was explored in planar lipid model systems of increasing complexity to

evaluate PIP2 lateral organization and cholesterol-dependent phase separation in

biomimetic membranes. The effect of soluble factors on the organization of PIP2 was

explored by determining the subphase-dependence of pressure-area relationships of pure,

naturally-derived PIP2 monolayers. Experimental observation and comparison with

theoretical modeling showed that PIP2 intermolecular organization was not governed

strictly by electrostatic repulsion, but rather by a combination of repulsion and water-

mediated hydrogen bonding. These results were confirmed in mixed bilayer vesicles by

fluorescence resonance energy transfer and neutron scattering, which also yielded the

iii

first estimate of putative PIP2 cluster size. Since cholesterol-induced liquid-liquid phase

coexistence is an important phenomenon in biological lipid organization, its effect on

PIP2 organization was assayed in 3-component monolayers including PIP2 and

physiological levels of cholesterol. These experiments demonstrated that the formation

of liquid-ordered domains constrains and segregates PIP2 to the liquid-disordered portion

of the membrane. Finally, the cholesterol-dependence of phase behavior was studied in a

novel cell-derived model system for studying lateral lipid organization in complex

mixtures approximating the plasma membrane. Phase separation and relative abundance

in this system was clearly shown to be a function of cholesterol concentration, with

cholesterol depletion inducing phase separation at physiological temperatures. These

observations were in accordance with simple model systems and demonstrated the utility

of cell-derived vesicles as an intermediate model system to study lipid behavior in

extremely complex mixtures, without the confounding factors of active cellular

processes. Taken together, the results presented demonstrate the potential physiological

importance of lateral lipid organization and the physiological regulation thereof.

iv

Table of Contents

Abstract.........................................................................................................iiiTable of Contents..........................................................................................vFigure List..................................................................................................viiiChapter 1 - Background and Significance: Lateral lipid heterogeneity in biological contexts.........................................................................................1

1.1 - Introduction and Hypothesis................................................................................1

1.2 - Lipid Organization Through PIP2 Domains......................................................3

1.2.1 - Physical chemistry of PIP2............................................................................4

1.2.2 - PIP2 in cell biology.........................................................................................5

1.2.3 - PIP2 cytoplasmic leaflet domains..................................................................8

1.3 - Lipid Organization Through Membrane Rafts................................................10

1.3.1 - Lipid raft composition and reconstitution in model systems....................10

1.3.2 - Giant Plasma Membrane Vesicles..............................................................12

Chapter 2 - Intermolecular Interactions in Pure, Naturally-Derived PIP2 Monolayers.........................................................................................14

2.1 - Experimental Design and Methods....................................................................15

2.1.1 - Lipids and reagents......................................................................................16

2.1.2 - Pressure-area isotherms...............................................................................16

2.1.3 - Time-course experiment..............................................................................17

2.1.4 - Other important experimental considerations..........................................18

2.2 – Experimental results – Combined electrostatics and hydrogen bonding determine intermolecular interactions in PIP2 monolayers....................................19

2.2.1 - Phase behavior of pure, natural PIP2.........................................................19

2.2.2 - Expanding effect of increased ionic strength on monolayers of PIP2......20

2.2.3 - Effects of different counterions...................................................................24

2.2.4 - Expanding effect of non-ionic chaotropes and temperature....................27

2.3 – Theoretical modeling results – Electrostatic contribution to the surface pressure of charged monolayers containing polyphosphoinositides.......................31

2.3.1 – Experimental justification for electrostatic modeling..............................31

2.3.2 - Theoretical model of electrostatic contribution to surface pressure.......33v

2.3.3 – Comparison of model with experimental results......................................41

2.4 - Results discussion, caveats and significance.....................................................46

Chapter 3 – Domain formation, lateral segregation and line tension in mixed lipid systems.....................................................................................52

3.1 - Experimental Design and Methods....................................................................53

3.1.1 - Lipids and reagents......................................................................................54

3.1.2 – Monolayer imaging......................................................................................54

3.1.3 – Neutron scattering and FRET in LUVs.....................................................55

3.1.3 – Edge fluctuation of liquid-liquid domains.................................................58

3.2 – Experimental results – Lipid segregation and domain formation in mixed lipid systems.................................................................................................................59

3.2.1 – FRET detection of PIP2 demixing in LUVs..............................................59

3.2.2 – Neutron scattering observation of PIP2 demixing and domains size......60

3.2.3 – PIP2 segregation in cholesterol-containing monolayers...........................65

3.3 – Experimental results – Line tension in cholesterol-DMPC monolayers........68

3.3.1 - Capillary Wave Theory................................................................................71

3.3.2 – Data analysis.................................................................................................73

3.3.3 – Line tension and dipole density results......................................................78

3.4 - Results discussion, caveats and significance.....................................................85

Chapter 4 – Bridging membrane raft model systems: Cholesterol-dependent phase separation in Giant Plasma Membrane Vesicles........91

4.1 - Justification for GPMV experiments.................................................................92

4.2 - Experimental Design and Methods....................................................................93

4.2.1 – Cell culture and treatment..........................................................................93

4.2.2 – GPMV isolation and visualization..............................................................94

4.2.3 – Cholesterol mol fraction quantification.....................................................95

4.2.3 – Quantification of SMase treatment............................................................95

4.2.4 – Fluorescence correlation spectroscopy......................................................96

4.2.5 – Detergent resistant membrane quantification..........................................97

4.3 - Experimental results - Cholesterol-dependent phase separation in GPMVs 99

4.3.1 – Lo phase comprises majority of GPMV surface area...............................99

4.3.2 – Cholesterol depletion/loading affects phase separation and Lo phase fraction....................................................................................................................102

vi

4.3.3 – Sphingomyelin depletion has no effect on Lo phase................................105

4.3.4 – Cholesterol level determine phase separation temperature...................107

4.3.5 – Correlation between GPMV phase separation and presence of DRM. 108

4.3.6 – Lipid tracer diffusivity is 3x slower in Lo phase than Ld phase.............110

4.4 - Results discussion, caveats and significance...................................................115

Chapter 5 – Conclusions...........................................................................1195.1 – A combination of electrostatics and hydrogen bonding determine PIP2 organization................................................................................................................119

5.2 – Theoretical modeling of the electrostatic contribution to surface pressure of charged monolayers...................................................................................................120

5.3 – PIP2 domain formation and segregation in mixed lipid systems.................121

5.4 – Line tension and dipole density differences in cholesterol-containing monolayers.................................................................................................................123

5.5 – Cholesterol-dependent phase separation in GPMVs.....................................123

Chapter 6 – Future Directions.................................................................1256.1 – Calcium-induced mesoscopic domains in PIP2-containing monolayers.....125

6.2 – Continuation of neutron scattering experiments...........................................128

6.3 – Monolayer behavior of PIP3............................................................................129

6.4 – Effect of PIP2 on line tension in cholesterol-containing monolayers...........129

6.5 – Influence of lipid composition perturbation on demixing in GPMVs.........130

6.6 – Protein sorting in membrane rafts..................................................................131

References...................................................................................................133

vii

Figure List

Figure 1-1. Structure of PIP2...............................................................................................4Figure 1-2. Colocalization of CTB and Lo phase in GPMVs............................................13Figure 2-1. Expanding effect of NaCl on PIP2 monolayers..............................................21Figure 2-2. Specificity of salt-expanding effect to PIP2...................................................23Figure 2-3. Effects of various counterions........................................................................26Figure 2-4. Evidence for water-mediated intermolecular hydrogen bonding....................29Figure 2-5. PIP2 isomer specificity of subphase NaCl expansion effect...........................30Figure 2-6. Experimental justification for electrostatic model..........................................32Figure 2-7. Electrostatic model results..............................................................................40Figure 2-8. Area per charge of PIP2..................................................................................42Figure 2-9. Comparison of model with experimental isotherms.......................................45Figure 3-1. FRET in LUVs................................................................................................63Figure 3-2. Neutron scattering from LUVs.......................................................................64Figure 3-3. Liquid-liquid domain formation in mixed lipid monolayers..........................67Figure 3-4. Co-localization of PIP2 and Ld phase.............................................................67Figure 3-5. Analysis of mode power fluctuation spectra at 30 mol% Dchol....................76Figure 3-6. Line tension and dipole density differences at critical composition...............81Figure 3-7. Composition dependence of line tension and dipole density difference.........84Figure 4-1. Cellular cholesterol level determines Lo/Ld ratio..........................................100Figure 4-2. Direct cholesterol modulation in GPMVs.....................................................104Figure 4-3. Sphingomyelin depletion has no effect on GPMV phase behavior..............106Figure 4-4. Cholesterol determines temperature-dependent phase separation................109Figure 4-5. Phase separation induces two distinct populations of diffusivities...............112Figure 4-6. Triton-extracted cholesterol distribution in sucrose gradients......................114Figure 6-1. Calcium-induced domain formation in PIP2 containing monolayers...........127

viii

Chapter 1 - Background and Significance: Lateral lipid heterogeneity in biological contexts

1.1 - Introduction and Hypothesis

The plasma membrane of protozoan and animal cells functions as the interface between

the enclosed environment of the protoplasm and the outside world. As part of this

function, membrane constituents are responsible for signal transduction across the

membrane, control of permeability to a variety of biologically-active molecules,

mechanical integrity of the cell through interaction with the cytoskeleton, and attachment

of cells to extracellular matrices. The canonical explanation for the structure and

function of the plasma membrane is the “fluid mosaic” model proposed by Singer and

Nicholson1 in 1978. In this model, the many functions of the cell membrane are

performed by integral and peripheral proteins dissolved in a two-dimensional, freely

diffusing, homogeneous fluid comprised of the lipid constituents of the membrane,

specifically phospho- and sphingolipids and sterols.

For several decades, the fluid mosaic model stood as the definitive description of the

plasma membrane, until the discovery2, 3 and extensive characterization (reviewed in4) of

Detergent Insoluble Glycolipid (DIG) enriched fractions of cellular membranes

(otherwise known as Detergent Resistant Membranes (DRMs)). The presence of these

fractions strongly suggested that the plasma membrane is not laterally uniform (as the

fluid mosaic model suggests) but instead contains regions of molecular and biophysical

heterogeneity. These findings (along with numerous confirmatory studies using

immunolocalization5, 6, fluorescence resonance energy transfer7, and single molecule

tracking8) have led to the “lipid raft” model of cellular plasma membranes9 which

suggests that cholesterol and glycosphingolipid-rich “rafts” exist as an immiscible liquid

phase in the bulk plasma membrane, and that these serve to segregate and localize certain

lipid and protein components4.

In addition to lipid heterogeneity in the context of membrane rafts, recent work has

implicated the localization and organization of specific lipids as an important factor

regulating their biological activity. Of particular interest in cell biology is phosphatidyl

inositol (4,5) bisphosphate (PIP2), a lipid that is unique in both its physicochemical

properties and biological importance. The cellular functions of PIP2 include regulation

of extracellular signal transduction, calcium signaling, cytoskeletal dynamics, ion

channels, membrane trafficking, vesicle fusion, and the localization of cytoskeletal and

other proteinaceous components to the plasma membrane, among others. Although the

significance of PIP2 is well-documented, there remains an unanswered question of how

such a small, inabundant, and membrane-bound molecule can be responsible for so many

diverse and important cellular roles. It has been suggested that the molecular

organization (i.e. single free lipid, several lipids bound together, a large complex of many

lipids) of PIP2 could be a determinant of the specific function of a particular pool of

PIP2. Previous work has suggested the existence of lateral differences in the

concentration of PIP2 arising from local production or sequestration, although no

conclusive mechanism for PIP2 domain formation yet exists.

In this work, the results of experiments and theoretical modeling will be presented in the

aim of addressing the hypothesis that lateral lipid organization in planar model systems

can be affected by variation of physiologically-relevant factors and that this variation 2

can impact the biologically-relevant properties of the component lipids. This hypothesis

will be addressed by investigation in three distinct contexts, starting with the most basic

and building to the most complex: (1 – Chapter 2) experimentation and modeling of a

pure PIP2 monolayer to investigate the relative influence of intermolecular hydrogen

bonding attraction and electrostatic repulsion in determining PIP2 interaction and

organization; (2 – Chapter 3) evaluation of domain formation in mixed lipid systems and

the influence of lipid demixing on line tension and PIP2 localization and segregation; and

(3 – Chapter 4) investigation of the effect of cholesterol level on phase coexistence and

behavior in complex cell-derived lipid and protein Giant Plasma Membrane Vesicles.

1.2 - Lipid Organization Through PIP2 Domains

Phosphatidyl inositol 4,5-bisphosphate (PI(4,5)P2 or PIP2) is a member of the

phosphoinositide (PPI) family of lipids. This family is defined by the inositol (6-carbon

cyclic sugar) headgroup attached to a phospholipid tail through a glycerol backbone (Fig.

1.1).

3

Figure 1-1. Structure of PIP2.(top) space filling model of PI (4,5) P2; (bottom) chemical structure of the tri-ammonium salt of PIP2.

Drawings adapted from Avanti website with permission.

Although the five primary hydroxyl groups of phosphotidyl inositol (PI) are potential

targets for enzymatic phosphorylation in cells, only C3, C4, and C5 are physiologically

phosphorylated to form the other PPI family members, likely due to steric restrictions.

These lipids are unique among biologically relevant (and particularly cytoplasmic leaflet)

lipids due to the large size and charge of the phosphorylated inositol headgroup.

1.2.1 - Physical chemistry of PIP2

At physiological pH, PIP2 carries a charge of between -3 and -410, 11 and has a cross-

sectional area of ~75 Å2/molecule, compared with 58 Å2/molecule for phosphatidyl serine

(PS), another important anionic inner leaflet lipid12. PIP2 has a total of five ionizable

groups (2 per phosphomonoester, 1 on phosphodiester), and the protonation of each of

4

these is strongly dependent on both the pH and the ionic strength and composition of its

environment. For example, although only one of the phosphodiester groups appears to be

cation-bound at physiological pH and salt concentration11 (charge of -4), changing the salt

from sodium to potassium changes the charge to -3, while addition of calcium and other

multivalent cations (such as spermine) can further neutralize the charge of this lipid10.

The high valence of PIP2 leads to relatively high charge densities when these molecules

are packed into compressed planar configurations, and this high charge density has

important consequences for the electrostatic lateral surface pressure governing the

molecular packing of these lipids. For zwitterionic and monoanionic lipids, surface

pressures at physiologically relevant densities are dominated by the length and

unsaturation of acyl chains and the size of the headgroup12, but for more highly charged

lipids, a significant surface pressure can arise from electrostatic repulsions between

phospholipids13. Additionally, electrostatic effects become important when considering

the interactions of charged lipids with soluble ionic components, such as salts and

polyionic macromolecules. While screening of surface charge by soluble counterions is

the typically considered mode of interaction, lipid headgroup deprotonation by soluble

ions13 has been shown to be an important determinant of lipid packing14, phase

transitions15, domain morphology16, and enzymatic lability17. In Chapter 2, modeling of

the electrostatics in highly charge planar system will show that electrostatic repulsion is a

major contributor to the lateral surface pressure of polyphosphoinositide monolayers.

1.2.2 - PIP2 in cell biology

Although the fluid mosaic model originally viewed lipids as an inert substrate for the

solution and activity of membrane-bound proteins, the constituent lipids of the plasma 5

membrane have been show to be key components in various cellular processes including

inflammation18, apoptosis19, migration20, and proliferation21, 22. Among membrane-bound

lipids, PIP2 enjoys a unique importance in the regulation of cell function. Despite its

structural simplicity and relative scarcity in cells (<1% of all plasma membrane lipids23, 24,

it has been shown to be a critical mediator for a variety of vital cellular processes. The

most widely recognized function of PIP2 is as a second messenger in several critical

cellular pathways. PIP2 is the substrate for phosphorylation by PI 3-kinase to produce

the signaling lipid PIP3 for mitogenic signaling through Protein Kinase B (PKB or Akt)25.

Additionally, stimulation of some G-protein coupled receptors (GPCRs) activates

hydrolytic cleavage of PIP2 by phospholipase C (PLC) to produce diacyl glycerol (DAG)

and inositol trisphosphate (IP3), which are effectors of the protein kinase C and calcium

signaling pathways, respectively (reviewed in26).

The cellular roles of PIP2 are not limited to being a passive substrate for enzymatic

modulation. One of the most significant functions of PIP2 appears to be in the regulation

of proteins responsible for the maintenance and dynamics of the actin cytoskeleton20, 27, as

well as the attachment of these cytoskeletal structures to the actin cytoskeleton28. The

discovery of gelsolin, an actin binding and severing protein that was shown to not only

bind PIP2, but to be functionally regulated by interaction with PIP229, began a line of

inquiry that has shown PIP2 to be a critical mediator of cytoskeletal assembly. It has

been shown to bind or activate a variety of actin-binding proteins including N-WASP,

Cdc42, cofilin/ADF, CapZ, profilin, and gelsolin, leading researchers to believe that this

lipid may be a critical checkpoint for actin-based protrusion and motility30-32. In addition

to its vital roles as a second messenger and cytoskeletal regulator, this lipid has been

6

shown to be involved in a large variety of other critical cell functions, including

membrane trafficking33 and attachment34, regulation of ion channels35, endocytosis and

exocytosis33, and synaptic vesicle fusion36.

1.2.2a - PIP2 Production

The enzymatic production and consumption of PIP2 involves several enzymes unique to

phosphoinositol metabolism. Typically, the production of PIP2 begins with the

phosphorylation of the relatively abundant (~10%) phosphotidyl inositol by PI4 Kinase in

Golgi, ER, and plasma membranes. This step is succeeded by phosphorylation of the 5’

carbon of inositol by a type I PIP 5-Kinase. Depending on the specific enzymes

involved, these steps can be inverted, and it is possible that the production mechanism

affects the cellular function of the final PIP2 product37. Although most cellular PIP2 is

metabolized by the phopholipase activity of PLC described above, it can be converted

back into the monophosphate product PI4P by PIP2 5’-phosphatase, which in turn can be

enzymatically hydrolyzed to PI38, 39. Similar mechanisms exist for the production of the

four PPI family members not described here.

1.2.2b - PIP2 binding domains

The ability of PIP2 to mediate its many cellular functions derives from its ability to bind,

activate, and localize effector proteins to the inner leaflet of the plasma membrane where

cellular PIP2 is found. This binding is mediated by several protein domains with specific

phosphoinositide-binding function. The most common of these phosphoinositide binding

domains is the Pleckstrin Homology (PH) domain, originally discovered as part of the N-

terminal of pleckstrin, the main PKC substrate in platelets. Since their discovery, over

7

250 unique PH domains have been suggested and many have been shown to be critical to

the membrane localization and function of a large variety of proteins. Although the

specificities and affinities of the PH domains are highly variable, some (e.g. PH from

PLC-δ) are highly specific for PI(4,5)P240. Several other specific PIP2 binding domains

exist (ENTH, FERM, Tubby, PX) that serve to not only localize proteins containing these

domains, but also change their function upon binding PIP2 (reviewed in23). For example,

the FERM domain of radixin, which links actin to the plasma membrane, releases

autoinhibited portions of radixin upon interaction with PIP2 allowing protein binding to

the previously cryptic domain41.

1.2.3 - PIP2 cytoplasmic leaflet domains

The discovery of multi-molecular lipid and protein aggregates in the outer leaflet of the

plasma bilayer suggests a critical role for lipid organization in regulating the function of

both lipids and proteins. Although most recent research into lipid organization has

focused on outer leaflet complexes, inner leaflet lipid structures have been proposed. The

large number and variety of binding partners and cellular functions of PIP2 beg the

question of how it is possible for a small (~1kD) membrane-bound molecule to be

involved in so many specific and diverse cellular functions. A recently proposed answer

is that control of PIP2 signaling comes not only from enzymatic regulation of the

abundance of this lipid, but also from regulation of its spatial organization, with respect

to both cellular effectors as well as adjacent lipid molecules. Some of the first evidence

supporting this hypothesis was the finding that a significant fraction of PIP2 in the cell

membrane was inaccessible for PLC hydrolysis42, 43. Later, detergent-resistant membrane

fractions, the putative membrane-localized signaling complexes termed “lipid rafts”, 8

were shown to be enriched in PIP244, 45. Other imaging methods, including GFP-tagged

PIP2-binding domains34, 46 and fluorescent anti-PIP2 antibodies44, 47 have likewise

confirmed the possibility of structurally distinct PIP2 fractions. Although the existence

of these domains and their functional significance remains disputed48, 49, spatial

segregation of PIP2 continues to be an intriguing possibility for regulation of this critical

lipid messenger.

Despite the mounting evidence for the existence of spatially distinct pools of PIP2, the

mechanism for the formation of such domains has yet to be explained. Recent research

argues strongly for interaction between unstructured polybasic domains of proteins such

as the myristoylated alanine-rich C-kinase substrate (MARCKS) and multiple PIP2

molecules, allowing concentration of this lipid through non-specific, electrostatic

attraction23, 44, 50-53. This hypothesis views the interactions between neighboring PIP2

molecules as dominated by electrostatic repulsion between the charge-dense poly-anionic

headgroups. On the other hand, recent experiments with liposomes containing PIP2

argue for the existence of PIP2 domains due to attractive interactions through hydrogen

bonding54, 55 by mechanisms similar to those previously proposed for phosphatidic acid56

or sphingomyelin57, 58. Finally, experiments showing colocalization of PIP2 with lipid raft

markers44, 45 suggest the possibility of domain formation by interaction with spatially

segregated proteins or lipids.

In Chapters 2 and 3 of this work, data will be presented to show that PIP2 intermolecular

interactions are not governed solely by electrostatic repulsion, but instead by a

combination of electrostatics and hydrogen bonding which can regulate PIP2 packing

through a variety of physiologically-relevant effectors. Additionally, it will be shown 9

that this attraction through hydrogen-bonding can induce PIP2 clustering in mixed lipid

bilayers. Finally, it will be demonstrated how large-scale cholesterol-induced domain

formation can segregate and concentrate PIP2 in plasma membrane-model monolayers.

1.3 - Lipid Organization Through Membrane Rafts

Originally identified for their role in sorting membrane components from the Golgi

apparatus to the apical side of the plasma membrane in epithelial cells2, 3, lipid rafts (now

more generally referred to as “membrane rafts”59) have been implicated in a large number

of diverse cellular functions including endo/exocytosis60, viral entry61 and budding62,

immune system signaling through Fc63, T-cell64, and B-cell receptors65, growth factor

signaling66, differentiation and pattern formation during development67, and the formation

and maintenance of focal adhesion complexes between cells and extracellular supports68.

In addition to their importance in the cellular mechanisms listed above, lipid rafts have

been proposed to be “hijacked” by numerous pathogens as part of their cell entry,

immune evasion, and proliferation mechanisms (reviewed in69). The various raft-related

functions are hypothesized to depend on the ability of rafts to regulate protein function

through a variety of distinct mechanisms including clustering of raft-associated proteins

and raft coalescence inducing protein interaction70, as well as possible regulation of

receptor affinity by raft association71.

1.3.1 - Lipid raft composition and reconstitution in model systems

Biochemical characterization of lipid raft components (as defined by their resistance to

detergent solubilization at 4oC) found enrichment of several lipid and protein plasma

membrane components relative to the detergent labile fraction. The most prominent

10

lipidic components enriched in lipid rafts are sterols such as cholesterol, glycosylated

lipids including GM1, and phospholipids with long-chain saturated fatty acid tails (most

prominently sphingomyelin)2, 72. Protein markers typically associated with membrane

rafts are GPI-anchored proteins such as Thy1, cholesterol binding proteins including

caveolin, and lipid-anchored proteins, among many other specific components of the

cellular processes mentioned above73.

Concurrent experiments with purified lipids in model membrane systems like monolayers

and Giant Unilammelar Vesicles (GUVs) have begun to clarify the physicochemical

bases for lateral lipid heterogeneity. The phase behavior of biologically relevant

phospholipids has been thoroughly investigated and can be simplified to a temperature

dependent transition between a highly-ordered, crystalline, solid-like phase (so) at low

temperature, and a disordered liquid phase (L or Ld) at higher temperatures.

Interestingly, inclusion of cholesterol into pure phospholipid systems induces the

formation of a second liquid phase (Lo), distinct from the Ld phase in the high degree of

conformation ordering and packing of the lipid acyl chains in the ordered phase74. While

the Lo phase also has a lower translational and rotational diffusivity75 than the L phase,

translational mobility is much greater in the Lo phase than in the crystalline phase. Thus,

Lo represents a cholesterol-dependent liquid phase with physical properties intermediate

to those of the Ld and so phases. The two liquid phases can coexist and form microscopic,

immiscible domains at a variety of physiologically-relevant conditions76-78, with a

significant enrichment of cholesterol, sphingomyelin and other saturated phospholipids.

Additionally, the canonical raft components GM1 and Thy1 reconstituted into supported

monolayers were shown to independently partition into the Lo phase79. The studies cited

11

here, along with many others, have contributed to the existing paradigm that liquid-liquid

coexistence in model systems is related to raft formation in cellular plasma membranes,

and that the cholesterol-enriched Lo phase is analogous to the raft phase.

1.3.2 - Giant Plasma Membrane Vesicles

Despite the success of lipid model systems in reproducing many of the proposed

properties of cholesterol-dependent lipid phase coexistence thought to be responsible for

lipid raft formation, a persistent criticism remains that these findings are biologically

irrelevant because no model system appropriately recapitulates the lipid complexity,

much less the diverse and abundant protein load, of the cell plasma membrane. However,

recent experiments with Giant Plasma Membrane Vesicles (GPMVs - isolated, large (up

to 10 m), spherical plasma membrane projections that presumably maintain the lipid80

and protein81 diversity of the cellular membrane) showed that complex mixtures such as

cellular membranes can phase separate into two liquid phases and sort membrane

components in a way consistent with the raft hypothesis (82 and Fig 1-2).

12

Figure 1-2. Colocalization of CTB and Lo phase in GPMVs.Napthopyrene stains the Lo phase while rhodamine-PE is the Ld phase marker. Counterstaining clearly

shows partitioning of the B subunit of cholera toxin (CTB), which stains the membrane raft component

GM1, into the Lo phase and away from the Ld phase.

These GPMVs provide an extremely useful model for studying phase separation in the

highly complex system of the plasma membrane and in Chapter 4, data will be presented

to show the effects of cholesterol concentration on phase separation and relative

abundance of the Lo and Ld phases in these cell-derived vesicles. Additionally, the

temperature dependence of two-phase coexistence, and the cholesterol dependence of this

temperature profile will be examined. Finally, the diffusivities of the two phases will be

measured using Fluorescence Correlation Spectroscopy to quantify the functional

consequence of phase coexistence in these model cell membranes. This data will be

compared to the results of previous experiments with mixed lipid vesicles and live cells

to draw conclusions about the generality of the observed behaviors to all cholesterol-

containing phospholipid membranes.13

Chapter 2 - Intermolecular Interactions in Pure, Naturally-

Derived PIP2 Monolayers

Membrane lipids are active contributors to cell function as key mediators in signaling

pathways controlling cell functions including inflammation, apoptosis, migration, and

proliferation. Recent work on multi-molecular lipid structures suggests that in addition to

the functionality of individual lipid moieties, there exists a critical role for lipid

organization in regulating the function of both lipids and proteins. Of particular interest

in this context are the polyphosphoinositides (PPI’s), especially phosphatidylinositol

(4,5) bisphosphate (PIP2). The cellular functions of PIP2 are numerous and well-

characterized, whereas the organization of PIP2 in the inner leaflet of the plasma

membrane, the nature of the factors controlling targeting of PIP2 to specific proteins, and

the functional consequence of this organization remain poorly understood.

To analyze the organization of PIP2 in a simplified planar system, we have used

Langmuir monolayers to study the effects of subphase conditions on monolayers of

purified naturally-derived PIP2 and other anionic or zwitterionic phospholipids. We

report a significant, unexpected molecular area expanding effect of subphase monovalent

salts on PIP2 at biologically relevant surface densities. This effect is shown to be specific

to PIP2 and independent of subphase pH. Theoretical modeling of the electrostatic

repulsion in this system reveals that this expansion is partly, but not entirely, due to ionic

strength-dependent protonation of the PIP2 headgroups. This effect is specific to PIP2

because electrostatic contributions to surface pressures in monolayers of phospholipids

become significant only when the charge spacing is below the Bjerrum length, and for 14

molecules with steric cross-sections typical of phospholipids in the cell membrane (~50

Å2), only polyphosphoinositides (of which PIP(4,5)P2 is the most abundant) achieve this

threshold. Theory and experiment show that surface pressure increases linearly with

PIP2 net charge and reveal crossing of high and low ionic strength pressure-area

isotherms, due to opposing effects of ionic strength in compressed and expanded

monolayers. The validity of the theoretical model is confirmed by comparison of area-

pressure isotherms of PIP2 with other acidic phospholipids over a range of subphase

conditions, revealing the extent to which electrostatic effects contribute to membrane

surface pressure, while quantitative differences between theory and experiment suggest

that attractive interactions between polyphosphoinositides, possibly mediated by

hydrogen bonding, can lessen the effect of electrostatic repulsions.

Uncharged chaotropic agents (e.g. trehalose, urea, temperature) that disrupt water

structure and the ability of water to mediate intermolecular hydrogen bonding were also

shown to specifically expand PIP2 monolayers, highlighting the importance of hydrogen

bonding or lipid headgroup hydration in maintaining the physical state of PIP2 in planar

systems, and precluding a strictly electrostatic explanation of the observed results. These

results suggest a combination of water-mediated hydrogen bonding and headgroup

repulsion in determining the organization of PIP2, and the specificity of these

observations for PIP2 over other anionic and inositol-based lipids suggests that PI(4,5)P2

may have unique ability to form hydrogen-bonded networks as a mechanism for its

structural and functional sequestration.

2.1 - Experimental Design and Methods

15

2.1.1 - Lipids and reagents

Natural lipids (bovine liver L-α-phosphatidylinositol, porcine brain L-α-

phosphatidylinositol-4-phosphate, porcine brain L-α-phosphatidylserine, and porcine

brain L-α- phosphatidylinositol-4,5-bisphosphate) were purchased as 1 mg/ml solutions

(chloroform:methanol:water 20:9:1 for PPI’s ; chloroform for PS) from Avanti

(Alabaster, AL) and stored at -20oC. Synthetic PIP2 analogs (dioleoyl

phosphatidylinositol (x,y) bisphosphate) were purchased as dried 0.1 mg aliquots,

dissolved in the supplied solvent (C:M:W) and stored at -20oC. The concentrations of the

lipid solutions were confirmed initially with phosphate analysis following acid digestion

of organic components83 and subsequently by comparing to the measured area per lipid

molecule. Subphase reagents HEPES, EDTA, D-trehalose, and urea were purchased

from Sigma (St. Louis, MO) and CsCl, NaCl, KCl, LiCl, MgCl2, CaCl2 were purchased

from Fisher (Hampton, NH).

2.1.2 - Pressure-area isotherms

For comparisons between different lipids at pH 7.5, monolayer subphases were prepared

with 10 mM HEPES, 0.1 mM EDTA, pH 7.5 dissolved in 18.2 MΩ ddH2O. For the low

pH experiments, the buffer was 10 mM sodium phosphate. For varying pH experiments,

the buffer was 3.3 mM sodium phosphate, 3.3 mM sodium citrate, and 3.3 mM glycine.

No effect of the various buffers at the same pH and ionic strength was observed. 25-30

mL of subphase solution were filtered through a 0.2 μm syringe filter (Sigma) and

introduced to a MicroTroughX Langmuir trough (Kibron Inc. Helsinki, Finland).

Approximately 7 nmol of lipid was withdrawn through a septum from a container stored

16

at -20oC to prevent solvent evaporation and deposited slowly on the subphase surface.

After a 10 min stabilization of the monolayer, the lipids were compressed at 15

Å2/molecule/min by moving the barriers of the trough using a microstepping motor. The

monolayer surface pressure was monitored with a surface probe using the Wilhelmy

method83 and the FilmWare software package (Kibron). Using the assumption of

complete wetting of the probe by the hydrophilic phase (i.e. zero contact angle, a

reasonable assumption for the hydrophilic platinum probes used in our experiment), the

expression for the measured interfacial tension, σi, is,

,

where F is the downward force on the probe by the surface and P is the well-defined

perimeter of the probe. The surface pressure, π, is then defined as the difference between

the measured tension and the surface tension of pure water, σw,

.

Temperature of the subphase was maintained using a circulating water bath.

2.1.3 - Time-course experiment

Approximately 0.01 nmol of PIP2 was deposited on the interface of 1 mL of filtered

subphase added to a single well of a multiwell plate (Kibron). Lipid was added until the

surface pressure increased to between 15-20 mN/m. The lipid was left to stabilize for

~30 min, until the surface pressure was stable (within 1 mN/m) for several minutes. 50

μL of 5 M NaCl were added to the subphase through an injection port and the change in

surface pressure was measured as a function of time.

17

2.1.4 - Other important experimental considerations

Both the low amount of lipids and the slow deposition rate were critical parameters for

reproducibility of monolayer isotherms, as was the lipid-impermeability of the Teflon-

coated barriers and the purity of the subphase solution. After several months of

experiments, barriers tended to become “leaky”, as defined by the inability to hold the

monolayer at a certain surface pressure without significant continued compression. After

this point, washing, scrubbing, or sonication of the barriers did not seem to improve their

performance, and they were replaced.

Surface active contaminants in the subphase solutions (for example, from inadequate

rinsing after washing glassware with soap or leaching of plastic storage vessels after

long-term storage of corrosive solutions) were a significant source of potential error in

the monolayer experiments. To check for surfactant contamination, the barriers were

compressed to their minimum separation after the trough was filled with subphase, but

prior to lipid deposition. A surface pressure increase (above instrument noise) at

relatively high barrier separations (before the meniscus of the barrier reached the probe)

indicated the presence of a surface active component in the subphase. If such an increase

was observed, the subphase buffer was not used for experiments.

Between each experiment, the barriers and trough were thoroughly cleaned 2x with

ddH2O and 100% ethanol, while the probe was flamed using a Bunsen burner or butane

torch. Monolayers of pure PIP2 could not be compressed past ~37 mN/m in our

experiments because the Teflon coated barriers of the microtrough wetted at high surface

PIP2 concentrations; hence the collapse pressure of the PIP2 monolayers could not be

18

measured, but its lower bound is at least 37 mN/m. It was observed that PIP2, while

well-soluble at room temperature in the CHCl3:MeOH mixture provided by the

manufacturer, formed a visible flocculent upon freezing. For correct lipid concentration,

it was important to ensure complete solvation of the lipid prior to use.

2.2 – Experimental results – Combined electrostatics and hydrogen bonding

determine intermolecular interactions in PIP2 monolayers

2.2.1 - Phase behavior of pure, natural PIP2

The relationship between the surface pressure (π) and molecular area of pure naturally-

derived PIP2 was investigated by compressing monolayers of PIP2 from 250 to 50

Å2/molecule and observing the effect of compression on the surface pressure of the

interface. Average isotherms for 10 separate trials are shown in Fig. 2.1a. As expected

from the known composition of the acyl chains of pure PIP2 (~50% unsaturated, 33%

arachidonic acid), these isotherms show a smooth, monotonic increase in surface pressure

as the molecular area is decreased. No phase transitions were observed for monolayers of

PIP2 under any of the conditions used in these experiments. The average area of PIP 2 at a

surface pressure corresponding to physiological conditions (~30 mN/m84) was 73.1 ± 3.0

Å2/molecule, somewhat larger than published values for SAPC (65 Å2)12, which is to be

expected from the added bulk of the sugar headgroup and electrostatic repulsions.

Despite the size and relatively high charge density of the PIP2 headgroup at physiological

pH, this molecule readily forms tightly compressed monolayers, as opposed to collapsing

into aqueous micellar structures at higher surface pressures. Hysteresis of the

monolayers due to loss of lipids through barrier leakage or monolayer collapse was

19

negligible under all conditions, and similar to control lipids such as SOPC (data not

shown).

2.2.2 - Expanding effect of increased ionic strength on monolayers of PIP2

To investigate the effect of ionic strength on the behavior of PIP2 monolayers, π-A

isotherms were taken with varying concentrations of NaCl in the subphase. Addition of

NaCl significantly expanded the monolayers at all surface pressures above 5 mN/m (Fig.

2.1a). This response was also observed upon addition of NaCl to the subphase of a

preformed PIP2 monolayer. At constant molecular area, the surface pressure increased

after addition of 250 mM NaCl with a magnitude commensurate to that observed in the

isotherm experiments, on a diffusion-limited time scale (Fig. 1a inset). At π = 30 mN/m,

the area per PIP2 molecule was increased by 13% to 82.5 Å2/molecule (Fig. 2.1b).

20

Figure 2-3. Expanding effect of NaCl on PIP2 monolayers.(A) π-A isotherms with 0 mM (squares) and 250 mM NaCl (triangles); (inset) change in surface pressure at

constant area/molecule upon subphase injection of 250 mM NaCl (at time = 0). (B) Area per molecule at 30

mN/m at pH 1.8 (n=7) and pH 7.4 (n=5). (C) Dose response to subphase NaCl. Error bars are average ± SE

at n=5, except where indicated. All data are L-α PIP2 on HEPES buffered subphase, pH 7.4 (unless

indicated), 30oC.

21

Quantification of the dose response of this effect reveals that the effect saturates at

approximately 200 mM NaCl and shows significant variation within the range of

physiologically-relevant salt concentrations (Fig. 2.1c).

To test for the possibility of an electrostatic mechanism (e.g. counter-ion cloud repulsion)

causing the monolayer expansion, the effect of 250 mM NaCl was measured on another

charged lipid, L-α PS, using the same conditions as employed in the PIP2 experiments.

Monolayers of PS were not affected in the same way as those of PIP2, instead showing a

very slight contraction in response to increased subphase ionic strength (Fig. 2a).

To determine whether the PIP2-specific expansion resulted primarily from the bulky

inositol ring, and at the same time control for acyl chain composition, the pressure-area

isotherms were repeated with phosphatidyl inositol 4-phosphate (L-α PI(4)P) and

phosphatidyl inositol (L-α PI). Because these molecules are precursors for enzymatic

PIP2 production in cells, they have similar or identical fatty acid compositions as PIP2,

and only differ in the degree of phosphate substitution on the inositol ring. In monolayer

experiments, neither of these lipids showed a significant expansion in response to

increased concentration of NaCl, although the monophosphate PI(4)P exhibited the same

trend as the bisphosphate PIP2, suggesting a similar, but much smaller effect (Fig. 2b).

These data suggest that the mechanism involved in NaCl-induced expansion of PIP2

monolayers is specific to PIP2 over other anionic, as well as other inositol-based lipids.

22

Figure 2-4. Specificity of salt-expanding effect to PIP2.Area per molecule of (A) L-α PIP2 and L-α PS; and (B) L-α PIP2, L-α PI(4)P and L-α PI on HEPES-

buffered subphase, pH 7.4, 30oC at π = 30mN/m. Mean ± SE, n=4.

23

In addition to the specificity of the expanding effect of NaCl on PIP2 compared to other

anionic phospholipids, the effect is also PIP2 isomer dependent. Quantification of the

molecular areas of synthetic PIP2 analogs substituted at different positions on the inositol

ring (3 and 5, 4 and 5, 3 and 4) shows that not only are the molecular areas dependent on

the positions of the phosphate, but also that the magnitude of the NaCl-induced expansion

is affected by the placement of the phosphomonoesters in the three different isomers (Fig.

2.5a). Direct comparison of this expansion reveals the greatest difference between 0 and

250 mM NaCl for PI(3,5)P2 (~22 Å), followed by PI(4,5)P2 (11 Å2) and PI(3,4)P2 (5 Å2),

and that the differences between PIP2 isomers are statistically highly significant

(p<0.001).

2.2.3 - Effects of different counterions

To determine the ion specificity of the expanding effect of monovalent salts on PIP2

monolayers, the effects of other cationic counterions were tested. At 250 mM, all

monovalent cations tested (Na+, K+, Li+, Cs+) showed similar, statistically significant

expansion of the PIP2 monolayers, with the magnitude of the effect directly related to the

charge density of the ion, i.e. Li+ > Na+ > K+ ~ Cs+ (Fig. 2.3a). The charge-density

dependence observed here differs from that reported for salt-induced expansion of less

highly charged anionic phospholipid monolayers, where either no cation dependence or

the opposite trend was observed14. The magnitude of the expansion of PIP2, in contrast to

PG14, by the different cations appears to be directly related to the Hofmeister series

describing the chaotropic nature of the ion (reviewed in85). This result suggests that in

addition to effects on headgroup protonation, these ions may also disrupt the structure of

multi-molecular water-mediated hydrogen-bonded networks within the monolayer. 24

Divalent counterions have a very different effect on PIP2 compared to monovalent salts.

Both CaCl2 and MgCl2 had a large compressing effect on pure PIP2 monolayers (Fig.

2.3b). The representative isotherms in Fig. 3b highlight these differences, both in the

area per PIP2 at π = 30mN/m and at lower surface pressures. The inset shows a

quantification of the compressing effect of divalent cations and demonstrates that PIP2

monolayers with 250 mM Ca2+ and Mg2+ were compressed by 15% and 9% over control,

respectively. These results are consistent with the previously observed ability of Ca2+ to

act as a PIP2 crosslinker by binding and dehydrating multiple phosphates with high

affinity86, 87, neutralizing their charges, and bridging headgroups to form tightly

compressed monolayers88, even at low surface pressure.

25

Figure 2-5. Effects of various counterions.(A) Area per molecule at π = 30 mN/m of L-α PIP2 on HEPES-buffered subphase with 250 mM salt; Mean

± SE, n=5. (B) π-Area isotherms of L-α PIP2 HEPES-buffered subphase, pH 7.4, 30oC (solid line) and

same conditions plus 250 mM CaCl2 (dashed line); (inset) quantification of the effects of 250 mM CaCl2

and MgCl2; mean ± SE, n=4.

26

2.2.4 - Expanding effect of non-ionic chaotropes and temperature

To test the hypothesis that monovalent salts disrupt attractive hydrogen bonding

interactions among PIP2 headgroups that partially overcome the electrostatic repulsion

expected from high headgroup charge density, several non-ionic chaotropic factors were

tested for their ability to disrupt these putative networks and induce monolayer

expansion. Urea, a protein denaturant commonly used because of its chaotropic

character, and trehalose, a non-reducing glucose dimer known for its cryoprotective

properties which derive from its ability to disrupt water structure, were tested for their

effect on PIP2 monolayers. Consistent with attractive interactions through hydrogen-

bonding, both non-ionic chaotropes had a strong expanding effect on the monolayers. At

π=30 mN/m, 5 M urea increased the area per PIP2 molecule by almost 25% to 90.9

Å2/molecule, the highest value observed for any of the conditions employed in these

experiments (Fig. 2.4b). Similarly, 5 mM trehalose significantly increased the area of the

PIP2 monolayer by 9%. These effects were specific to PIP2, as neither treatment had a

significant effect on monolayers of PI.

Finally, as confirmation of the hydrogen bonding hypothesis, the temperature-dependent

behavior of PIP2 monolayers was tested. These monolayers showed a very significant

contraction as the temperature of the subphase was decreased from 34 to 17oC, decreasing

the area per molecule by almost 50% (Fig. 2.4a). In contrast, monolayers of PI were

contracted by only ~10% over the same temperature range, consistent with a simple

scaling of pressure with kBT. While some contraction is expected due to the decrease in

kinetic energy of the lipids, the 50% difference observed for PIP2 strongly suggests an

additional mechanism, such as the disruption of a hydrogen bonded network by increased 27

thermal energy of the subphase. Pure PIP2 could not form compressed monolayers at

subphase temperatures below ~15oC, instead exhibiting collapse at relatively low surface

pressures (<10 mN/m; data not shown). This result could be relevant to understanding

temperature-induced changes in cell structure, such as cold activation of platelets, a

process during which changes in PIP2 organization at the plasma membrane trigger actin

assembly89.

28

Figure 2-6. Evidence for water-mediated intermolecular hydrogen bonding.Area per molecule of L-α PIP2 and L-α PI at π = 30 mN/m on HEPES buffered subphase, pH 7.4 (A) in

presence of 5 mM trehalose and 5 M urea; and (B) as a function of the temperature of the subphase (circles

= PIP2; squares = PI).

29

Figure 2-7. PIP2 isomer specificity of subphase NaCl expansion effect.(A) Area per molecule at π = 30 mN/m of DO-PIP2 isomers on HEPES-buffered subphase. Mean ± SE,

n=7. (B) Difference in area per molecule of DO-PIP2 isomers between 250 mM NaCl and no subphase

NaCl. The isomer dependence of the NaCl effect was measured to be significant to p = 0.0001 by two-way

ANOVA. (C) Conceptual cartoon of the intermolecular interactions between PIP2 molecules. In absence

of chaotropic agents (green ellipses), PIP2 molecules form water-mediated hydrogen-bonded networks.

Upon addition of chaotropes, networks are broken, and electrostatic repulsion between charged phosphates

induces expansion of the monolayer.

30

2.3 – Theoretical modeling results – Electrostatic contribution to the surface

pressure of charged monolayers containing polyphosphoinositides

2.3.1 – Experimental justification for electrostatic modeling

Phospholipids in a cell membrane pack to a density corresponding to an area per

molecule of 40-70 Å2, equivalent to a lipid monolayer with surface pressure of

approximately 30 mN/m 84. At this density the spacing between charges for univalent

phospholipids is slightly greater than the Bjerrum length (lB), the distance at which

electrostatic energies are equal to the thermal energy kBT (approximately 7.1 Å). As a

result, the lipids can be approximated as independent point charges that create a

significant field orthogonal to the membrane surface, but only modest repulsive

interactions within the plane of the membrane. However, when the valence is greater than

2, as it is for PIP2, the most common PPI and one with great biological importance, the

charge spacing becomes less than lB leading to significant electrostatic interactions within

the plane of the membrane and effects not seen in less highly charged membranes. The

magnitude of the electrostatic contribution is shown in Figure 2-6, which compares area-

pressure isotherms of the anionic lipids phosphatidyserine (PS), phosphatidylinositol (PI)

and its mono- and di-phosphorylated derivatives PIP and PIP2. All of these lipids are

natural products containing mainly stearoyl and arachidonoyl moieties at their SN1 and

SN2 positions, respectively. The unsaturated acyl chains prevent any observable phase

transitions from a liquid-condensed phase to a gel phase, hence the differences between

the isotherms are the direct effect of the increased charge on PIP and PIP2, which at

neutral pH bear slightly more than 2 and 3 charges, respectively. The increased surface

31

pressure of PIP2 monolayers is not screened out by increasing subphase monovalent salt.

Figure 2-6 shows that whereas increasing ionic strength from 10 mM to 250 mM has no

significant effect on monolayers of PS or PI, it increases the pressure of PIP2 monolayers

over a broad range of molecular areas.

Figure 2-8. Experimental justification for electrostatic model.Surface pressure ()-molecular area () isotherms of naturally-derived anionic phospholipids (PS, PI, PIP,

and PIP2) on a buffered subphase with 10 mM (a) and 250 mM (b) subphase NaCl. Isotherms comparing

the effect of low (open circles) and high (closed circles) subphase ionic strength on monovalent (PI) (c) and

multivalent (PIP2) (d) acidic phospholipids. All isotherms shown are representative of the average of 5-8

isotherms per condition.

32

2.3.2 - Theoretical model of electrostatic contribution to surface pressure

A continuum model can be applied to calculate the electrostatic component of surface

pressure when the distance between charged lipids is greater than ZlB, where Z is the

number of charges per lipid head group. In the case of a PIP2 monolayer, where Z is

about 3-4, ZlB corresponds to a surface area of ~160 Å2 per molecule. Fig. 2-6d shows

that at molecular areas higher than ZlB, where electrostatics are not expected to contribute

significantly to lateral pressure, the surface pressure of PIP2 does not depend strongly on

ionic strength. This estimate is also in agreement with the experimental observation (Fig.

2-6c) that for PI, which has only one charged group (ZlB ~ 50 Å2), the surface pressure

does not show any significant dependence on ionic strength down to 50 Å2/molecule. PIP

which has 2 negatively charged groups does show a small influence of ionic strength

below ~80-100 Å2. If the distance between charged groups is less than defined by the

Bjerrum length, then collective effects described by a Gouy-Chapman approach become

important to consider. In this case we describe the monolayer as a charged surface with

surface charge density (. The Poisson-Boltzmann equation for electric potential (

distribution near this charged surface reads

(1)

where e is the elementary charge, n0 is the number density of univalent electrolytes, and

is the dielectric permittivity of water. Solution of the Poisson-Boltzmann equation using

the boundary condition at a charged interface

(2)

33

is well known90. The surface pressure is calculated by evaluation of the variation of the

thermodynamic potential ( due to introducing a charged surface in an electrolyte

solution with fixed chemical potentials of ions, which can be found from the charging

theorem91, 92:

(3)

After taking into account Eqs.(1,2) and integrating, the change of the thermodynamic

potential due to a charged surface can be expressed as

(4)

This relation is identical to that used in93 for consideration of charge-reversal instability in

mixed bilayer vesicles. The electrostatic part of the surface pressure can be calculated by

differentiating (4) with respect to the surface area of a monolayer S at fixed charge

(variation of the surface charge with surface area is important experimentally and is

accounted for by dissociation-association equilibria, as discussed later). Direct

calculation of the derivative from relation (4) and the solution of the Poisson-

Boltzmann equation (1) requires several steps. An efficient way to carry it out is to use

the identity

(5)

which follows from the Poisson-Boltzmann equation, differentiating it by S, multiplying

by and integrating over the volume occupied by electrolyte. As a result we arrive at the

following relation

34

(6)

where the first term in relation (6) comes from the fact that the thermodynamic potential

(4) is proportional to the area of the monolayer. Since , from the

boundary condition (2), we have

and for the surface pressure we obtain the following relation

Since

and

it follows from the solution of the Poisson-Boltzmann equation, where

is the Debye screening length, that after integration we obtain

(7)

Eq. 7 can be rewritten in a more convenient form as follows

35

(8)

In the limit of large absolute values of the surface potential (0), relation (8) reduces to a

very simple form:

(9)

Relation (9) shows that for a highly charged monolayer the electrostatic contribution to

the surface pressure is equal to twice the kinetic pressure of a 2D gas, although its

physical meaning is different. The surface potential of a monolayer according to relation

(2) is found from the expression

(10)

According to (8), the electrostatic contribution to the surface pressure decreases with

diminution of the absolute value of the surface potential. According to relation (10) this

should take place if the ionic strength of the solution increases while the charge of the

monolayer remains constant.

However, in reality, the charge of the monolayer depends on the dissociation-association

equilibrium of the ionic groups of the lipid. In the negatively charged monolayer

considered here, a decrease of the surface potential increases the electrochemical

potential of the charged lipoid head groups by –e(0), and the dissociation-association

equilibrium is shifted. In the condition of equilibrium

36

the charging of the monolayer can be accounted for by introducing an effective

equilibrium constant pKe, whose value depends on surface potential according to the

relation . The degree of deprotonation dP of the lipid head

group is then

(11)

If the absolute value of the surface potential goes down, the effective pKe goes down as

well and the head group becomes more deprotonated. This effect of adding more charged

groups to the surface can increase the surface pressure (6). If the lipid has several ionic

groups which can be deprotonated, then the degree of deprotonation of each of them

should be calculated according to the relation (11). As a result we have the following set

of equations for the calculation of the electrostatic surface pressure for n lipid headgroups

as a function of the ionic strength of the solution

(12)

(13)

(14)

37

where is the surface area per lipid molecule. The set of Eqs.(12-14) allows one to draw

several important conclusions. The dependence of the electrostatic surface pressure on

the surface area per molecule of a lipid like PIP2 with 5 ionizable groups and bare pK

values of 2, 3, 4, 7, and 7.7 (first ionization pKs from11; second ionization pKs estimated

from pKs of phosphatidic acid94) is shown in Fig. 2-7a for several pH values at low (c=10

mM) and high (c=250 mM) salt concentrations. The theoretical curves have several

interesting features, which qualitatively correspond to the experimental data for PIP2 in

Fig. 2-6d. In agreement with the experimental data, the surface pressure of compressed

monolayers is higher at higher ionic strength, which initially appears counterintuitive due

to the increase of screening with ionic strength. The increased surface pressure is due to

the increased charge density of the monolayer at higher ionic strength, since high ionic

strength diminishes the pKe value, consistent with previous observations and modeling of

less charged amphiphile monolayers13, 14, 16, 17. This expansion due to charging of the

monolayer with increased subphase ionic strength becomes prevalent when the surface

area per lipid molecule reaches a minimal threshold, as seen from the crossing of

theoretical isotherms for low and high salt concentrations (Fig. 2-7a), in agreement with

experimental data (Fig. 2-6d). The importance of lipid headgroup deprotonation in the

behavior of isotherms is illustrated in Fig. 2-7b, where theoretical isotherms for low and

high salt concentrations are shown both with and without accounting for the shift of

effective pK values as a function of the potential of the monolayer. From these plots, it is

clear that if the charge per lipid molecule is fixed, the pressure of the monolayer

decreases with increasing ionic strength due to screening. The dependence of the

effective pK value on the area per lipid molecule, calculated according to Eqs.(12, 13) for

38

pH = 7.5 at low (10 mM) and high (250 mM) ionic strengths, is shown in Fig. 2-7c. Due

to the decrease in pKe values at higher ionic strength, the charge per lipid molecule

increases, which leads to increased surface pressure, as remarked above. Eqs. (12-14)

also predict another important property of the system considered here. Calculated at a

fixed electrostatic pressure, the dependence of the area per lipid molecule () on the

charge of that lipid is linear with a slope of 2kBT/s as found from relation (9).

This theoretical model was developed in collaboration with Dr. Andrejs Cebers, a visiting

professor and collaborator from the Physics Department at the University of Latvia.

39

Figure 2-9. Electrostatic model results.(a) Calculated isotherms of electrostatic surface pressure at low (c = 10 mM; dashed line) and high (c = 250

mM ; solid line) subphase ionic strength at pH=5, 7.5, 12. (b) Isotherms of electrostatic pressure with and

without accounting for the dependence of effective pK values on the potential of the monolayer. c = 10 mM

(dashed line) and c = 250 mM (solid line) accounting for pK shift, c = 10 mM (large dots) and c = 250 mM

(small dots) without accounting for pK shift. (c) Effective pK values as a function of surface area per lipid

molecule () at pH=7.5 and c = 10 mM (dashed line) or c = 250 mM (solid line).

40

2.3.3 – Comparison of model with experimental results

The theory derived above makes several predictions that are verified by experimental

data. The linear dependence of the area per lipid molecule on its charge was tested by

varying the pH of the aqueous subphase and measuring pressure-area relationships

(specifically, area/molecule at S = 30 mN/m). Fig. 2-8a shows the change in

area/molecule of PIP2 monolayers as a function of pH at high and low ionic strength, and

reveals expansion of the monolayer both by increasing ionic strength and increasing pH,

both of which lead to increased deprotonation of PIP2. Eqs. 12-14 allow calculation of

the net charge on PIP2 at each value of pH and S, and the molecular area as a function of

net charge is shown in Fig. 2-8b. The data at high ionic strength are well fit by a linear

relationship over the entire range of charge above -1. The proportionality constant

derived from the fit is less than, but within a factor of two from, the simple prediction of

2kBT=S expected from Eq. 9.

41

Figure 2-10. Area per charge of PIP2.(a) Isobaric (S = 30 mN/m) area/molecule () as a function of measured pH of pure, naturally-derived

PIP2 on a buffered subphase with added 10 mM (open circles) or 250 mM (closed circles) NaCl. Points

shown are the average ± standard deviation for three trials. (b) Measured area/molecule of PIP2 as a

function of charge/molecule calculated from Eq. 11. The slope of the linear least-squares fit suggests that

each charge contributes at additional 10 Å2 to the experimental molecular area of PIP2. This observation is

much small than the ~25 Å2/charge predicted by the purely electrostatic model, confirming the

incompleteness of the purely electrostatic model neglecting an attractive component.

42

A more detailed comparison of theory and data is shown in Figure 2-9, which compares

the area-pressure isotherms of PIP2 at three different pH values where significant

differences in charge density are expected. At pH 12, where the charge separation is well

below the Bjerrum length and near the minimum value for PIP2, changing the ionic

strength from 10 mM to 250 mM has a small effect over the measurable range of

area/molecule (Fig. 2-9c). The theoretical curves are similar in shape and magnitude to

the experimental data as both theory and experiment show a crossing of the curves at a

critical area/molecule where the screening effect of salt on electrostatic repulsions begins

to dominate the pKe lowering effect important at lower molecular areas. At pH 7.5,

theory predicts that the crossover occurs in a more expanded monolayer (near 160 Å2,

Fig. 2-9c), in excellent agreement with the experimental result (Fig. 2-9a). The measured

differences in pressure of very expanded monolayers are small, but statistically

significant (inset Fig. 2-9a). At pH 1.8, where the charge on PIP2 is near -1, theory

predicts a very small effect of electrostatic repulsion on the surface pressure, and the

experimentally measured pressure is indistinguishable from zero at areas above 150

Å2/molecule. The theoretical curves with no adjustable parameters shown here

qualitatively agree with the experimental data, suggesting that, unlike PS or PI,

polyphosphoinositide membrane surface pressures are strongly affected by electrostatic

effects under physiological conditions. However, quantitative differences between theory

and experiment suggest the limits of this purely electrostatic model. The theory is not

expected to be valid at small molecular areas where steric interactions become

significant; correspondingly, below 60 Å2 the experimentally measured pressures are

systematically larger than theoretical prediction. In contrast, at higher molecular areas

43

(>120 Å2) and pH values, this electrostatic theory predicts significantly larger pressures

than are measured (Fig. 2-9a-d) and a steeper dependence of molecular area on charge

than is observed (Fig. 2-8b). The lower pressures measured experimentally compared to

the predicted purely electrostatic contribution to lateral pressure suggest that attractive

interactions counter the electrostatic repulsions measured and modeled in this study. A

likely mechanism of attractive interactions is hydrogen-bonding between lipid

headgroups, as suggested in charged phosphatidic acid95 and zwitterionic phosphatidic

ethanolamine membranes96, and recently confirmed in bilayers containing

phosphoinositides55, 97.

The modeling, data analysis, and writing of the theory section of this chapter was

performed in close collaboration with both Dr. Andrejs Cebers and the advisor for this

thesis work, Dr. Paul Janmey.

44

Figure 2-11. Comparison of model with experimental isotherms.Measured isotherms of pure, naturally-derived PIP2 on buffered subphase with added 10 mM (open circles)

and 250 mM (closed circles) NaCl at pH 7.5 (a), pH 12 (c), and pH 1.8 (e). Isotherms shown are

representative of 3 trials/condition. (a - inset) quantification of difference between low and high salt buffer

of expanded (300 Å2/molecule) and compressed (80 Å2/molecule) monolayers of PIP2. Theoretically

calculated isotherms for PIP2 at pH 7.5 (b), pH 12 (d), and pH 1.8 (f) with 10 mM (thin line) and 250 mM

(bold line) subphase ionic strength.

45

2.4 - Results discussion, caveats and significance

Polyphosphoinositides are well characterized as important signaling intermediates, but

much more is known about the genetic regulation and expression of the enzymes that

produce or degrade these lipids than about the physical chemistry that determines these

lipids’ distributions within the plasma membrane or their trafficking between different

cellular compartments. Because of their large negative charge, it appears generally

accepted that these lipids display only mutually repulsive interactions within the plane of

the bilayer that keep them dispersed unless they are complexed to specific proteins50, 53, 98.

Some lines of evidence suggest that PPIs are strongly sequestered under conditions that

produce detergent insoluble lipid fractions (often taken as evidence of PPIs’ localization

to lipid rafts45), while other studies using fluorescence energy transfer methods provide

evidence that hydrogen bonding might stabilize PPI-rich clusters54, 55. In this context, the

present results provide quantitative estimates of the magnitude of electrostatic

interactions among PPIs and show that attractive interactions, mediated by hydrogen

bonding, significantly counterbalance the electrostatic repulsions.

A feature of pressure-area isotherms of PIP2 that is well explained by purely electrostatic

mechanisms is the general effect of monovalent ions on surface pressures. While the

expanding effect of monovalent salt in the subphase of PIP2 monolayers may seem

inconsistent with electrostatic repulsions between the headgroups (subphase ions might

be expected to shield the anionic headgroups and allow tighter packing 99, 100), monolayer

expansion by subphase cations results from the dependence of the phosphomonoester

ionization potential on ionic strength, previously shown for monolayers of phosphatidic

46

acid13. This effect has been shown to be important in regulating the gel-liquid transition

temperature of charged monolayers15, although the measured magnitude of the expansion

effect of subphase salts with other anionic lipids is much smaller than the expansion

observed here with PIP214.

The purely electrostatic contribution to the surface pressure of PIP2 monolayers was

determined by modeling the system as a uniformly distributed plane of ionizable groups,

the charge density of which is a function of both the pKa’s of the ionizable groups and

the ionic strength of the subphase solution. The surface pressure due to electrostatic

repulsion, calculated by differentiating the thermodynamic potential with respect to the

surface area corresponds qualitatively with some of the observed experimental results.

The high pressure observed with expanded monolayers (up to 150 Å2/molecule) at neutral

pH can be explained by the repulsion of the highly charged headgroups. Additionally,

both the crossing over between isotherms with low and high ionic strength and the

expansion of the monolayer due to high ionic strength were confirmed with the

electrostatic model at neutral pH (Fig. 2-1a and Fig. 2-9b). However, many of the

experimentally observed results are not compatible with a purely electrostatic treatment.

Specifically, the varying effects of different monovalent ions cannot be accounted for

entirely by changes in subphase ionic strength. Both the PIP2 isomer specificity of the

NaCl-induced monolayer expansion and the effects of uncharged chaotropes and

temperature also point to a more complex molecular mechanism than the strictly

electrostatic subphase ionic strength modulation of apparent headgroup pKa.

Additionally, the expanding effect of subphase salt at pH 1.8 (Fig. 2-1b) is inconsistent

with the model which predicts no electrostatic effects under conditions where all

47

phosphomonoesters are protonated (Fig. 2-9f). Finally, in nearly all cases, the

experimentally determined surface pressure of PIP2 is significantly lower than predicted

from a conservative estimate for the purely electrostatic effect.

The results of the experiments described above highlight the importance of attractive

interactions, probably mediated by hydrogen bonding, that significantly counter the

repulsive electrostatic interactions between PIP2 lipids in planar systems. These

attractive interactions can be disrupted by chaotropic factors such as monovalent ions,

trehalose, or urea. These findings are summarized in a qualitative model presented in

Fig. 2-5c. In absence of disrupting agents, several PIP2 molecules are shown as

interacting through a water-mediated hydrogen bonded network. When either ionic

factors that disrupt water-PIP2 interactions or non-ionic chaotropes are present, hydrogen

bonding is disrupted and electrostatic repulsion causes an increase in molecular area.

This model is supported by the magnitude of the expanding effect of monovalent cations

on pure PIP2 monolayers, as well the effects of urea and trehalose (strong non-ionic

chaotropes). The calculated energy difference between the proposed hydrogen-bonded

state and the chaotrope-disrupted expanded state (for 250 mM LiCl: ΔArea = 17.8

Å2/molecule at 35 mN/m = ~6 kJ/mol) is commensurate with the loss of approximately

one hydrogen bond per PIP2 molecule. The possibility of intermolecular hydrogen

bonding between PIP2 headgroups in mixed lipid systems has been shown both

experimentally54, 55 and in simulations97, and the data presented here confirm that

possibility through experiments showing hydrogen bonding to be an important factor in

intermolecular PIP2 interactions.

48

The effect of temperature on PIP2 monolayers also suggests important non-electrostatic

interactions among these lipids. The striking decrease in surface pressure with decreased

temperature is far greater than observed with other charged fluid phase lipids, and does

not scale simply with thermal energy. Indeed monolayers of pure PIP2 are significantly

less stable at room temperature than at 37oC, and cannot form below 15oC. The collapse

of PIP2 monolayers at low temperature may be related to the hypothetical clustering of

PPIs at low temperature thought to trigger cold-activation of platelets and possibly other

biological functions89.

An alternative explanation to electrostatics and hydrogen bonding for the observed

effects of subphase salts involves the intercalation of the monovalent salts into the plane

of the anionic headgroups to form a network lattice between the phosphates and cations.

This explanation appears less likely since the expansion is greatest with the smallest,

most electropositive ion (Li+) and decreases with ion radius (Fig. 2-3a). Also, while the

formation of a rippled phase in the absence of salts could produce a more compressed

monolayer, a phase transition from the liquid phase to the rippled phase was not observed

with any of the isotherms (Fig. 2-1a). Additionally, the ripple phase would only be likely

to form at high surface pressures, whereas the differences between the high and low salt

states are apparent at pressure as low as 5 mN/m (Fig. 2-1a).

Two pieces of evidence argue for the importance of water in maintaining this network, as

opposed to hydrogen bonding directly between adjacent PIP2 molecules. The non-ionic

solutes urea and trehalose, which are not expected to interact with phosphate groups,

have a strong expanding effect on PIP2 monolayers, likely as a result of their disruption

of water structure and subsequent disturbance of the hydrogen-bonded network (Fig. 2-49

4a). Second, the significant reduction of the area per molecule of PIP2 induced by

divalent cations (Ca2+ and Mg2+) confirms their ability to bridge neighboring lipids with

resulting dehydration of the interface, and suggests that although the PIP2 monolayers

maintain a compressed state through their ability to hydrogen bond, they are not as tightly

compressed as when directly crosslinked by divalent cations (Fig. 2-3b).

Many experiments suggest that there are at least two distinct modes of interaction for the

many cellular binding partners of PIP2. Some proteins (e.g. those containing PH

domains) have a specific binding site for individual PIP2 molecules101-103, whereas others

contain unstructured polybasic domains thought to bind several PIP2 molecules

simultaneously through non-specific electrostatic attraction (e.g. MARCKS53, 104). It

seems reasonable to consider the possibility that a cell could regulate PIP2-mediated

signaling by influencing the balance between hydrogen-bonding and electrostatic

repulsion, thereby moderating the pools of PIP2 available for single-lipid binding protein

domains versus those that bind multi-molecular assemblies.

The biological relevance of this work derives from experimentation on lipids in a planar

context, which is the appropriate configuration for PIP2, as a cytoplasmic leaflet plasma

membrane constituent. Additionally, the surface pressure, temperature, pH and ionic

conditions have been chosen to properly reflect physiological conditions. However, the

major caveat of this work is that while these experiments are on pure PIP2 monolayers,

the physiological concentration of PIP2 in resting cells is quite low (~1% of the total

lipids). This fact does not invalidate the relevance of the results presented in this chapter

because of 1 - physiological fluctuations in local PIP2 concentration due to enzymatic

production, sequestration, and self-aggregation; and 2 – importance of intermolecular 50

interactions. Although simple diffusive arguments suggest that lipids cannot be strongly

concentrated through enzymatic production (the rate of generation is lower than the rate

at which the generated molecules diffuse away) , these arguments do not account for

intermolecular forces, like the attractive hydrogen bonding suggested by the work in this

chapter, that may influence this balance towards aggregation. However, even if very high

PIP2 concentrations (like those considered here) never exist in a cell, intermolecular

attraction between neighboring like molecules are important to consider, especially

considering the multitude of distinct PIP2 binding proteins, and the possibility of

differential binding based on aggregate size. Additionally, results presented in the

following chapter will confirm the relevance of the conclusions in this chapter with

experiments on PIP2 organization in mixed lipid systems, where PIP2 is the minority

component.

A shortcoming of the monolayer results presented in this chapter is that there is no

information in this experimental setup regarding the size, structure or abundance of the

proposed hydrogen-bonded PIP2 clusters. This shortcoming is partly addressed in the

following chapter with experiments using neutron scattering to measure lateral

heterogeneity in mixed lipid vesicles containing PIP2 that give the first suggestion of the

possible size of the putative PIP2 domains.

Parts of this chapter have been adapted from accepted manuscripts to Biophysical Journal

and the Journal of the American Chemical Society with copyright permission from the

publishers pending.

51

Chapter 3 – Domain formation, lateral segregation and line

tension in mixed lipid systems

To address one of the major limitations of the previous chapter’s results on the behavior

of pure naturally-derived PIP2 in planar systems, namely that such a high concentration

of PIP2 is not likely to exist in a cell, experiments in this chapter evaluate PIP2

localization in mixed lipid systems, including Large Unilamellar Vesicles (LUVs) and

mixed lipid monolayers. Initially, lateral heterogeneity is evaluated by neutron scattering

and Fluorescence Resonance Energy Transfer to determine whether vesicles containing

small fractions of PIP2 in a background of zwitterionic phospholipids (phosphatidyl

choline – PC) form locally-enriched domains or distributed homogeneously around the

surface of the vesicle. FRET data suggest that lateral inhomogeneity does exist in this

system, and that it is dependent on ionic factors, consistent with previous results54, 55.

Novel neutron scattering examination of the same system similarly suggests the presence

of PIP2 domains at physiologically-relevant conditions, while features in the scattering

data from these vesicles suggested an initial estimate for the size of the putative PIP2

clusters suggested in the previous chapter.

Additional experiments presented in this chapter will evaluate lateral heterogeneity in

more complex lipid mixtures, namely monolayers including cholesterol, which is shown

to induce large-scale phase separation at physiological concentrations, consistent with

previous results (see Section 1.3.1). Fluorescent microscopic observation of these phase-

separated monolayers was used to determine the effect of PIP2 on the formation of

cholesterol-dependent Lo domains, as well as localization and segregation of PIP2 52

induced by these domains. Surprisingly, inclusion of PIP2 in these lipid mixtures did not

have a measurable effect on the presence, properties or stability of the cholesterol-

induced domains. However, the presence of the domains did induce the segregation of

PIP2 away from the cholesterol-rich Lo phase and concentration in the Ld phase.

Finally, the results of a microscopic study measuring line tension in mixed lipid

monolayers are presented. As in the PIP2-containing monolayers, domain formation is

induced by inclusion of cholesterol in the lipid mixtures, and the line tension () and

dipole density differences () between demixed fluid phases of monolayers comprised of

dimyristoylphosphatidylcholine (DMPC) and dihydrocholesterol (DChol) are

investigated by measuring the two-dimensional thermal fluctuations of domain

boundaries visualized by the inclusion of a fluorescent tracer lipid. Employing an

extensive data set, the surface pressure dependence of and is determined at three

different monolayer compositions (30%, 35%, and 40% DChol). Both parameters are

found to decrease with a power law dependence as the surface pressure approached the

phase transition pressure (t), in agreement with previous measurements. Additionally,

photobleaching effects and domain size influence were quantified and found to be small

but significant. Finally, comparing the domain-promoting effect of with the domain-

disrupting effect of showed that there is a composition-independent correlation between

these parameters and the reduced surface pressure (t-). This result suggests a universal

relationship between these parameters in the absence of perturbing factors, which could

be used to identify the presence and activity of line-active compounds.

3.1 - Experimental Design and Methods

53

3.1.1 - Lipids and reagents

Lipids were purchased as solutions (chloroform:methanol:water 20:9:1 for PIP2 ;

chloroform for other lipids) from Avanti and stored at -20oC. The concentrations of the

lipid solutions were confirmed initially with phosphate analysis following acid digestion

of organic components83 and subsequently by comparing to the measured area per lipid

molecule. Subphase and solution reagents were purchased from Sigma and Fisher.

3.1.2 – Monolayer imaging

Lipid monolayers were prepared as described in Section 2.1.2, with the addition of 0.1%

fluorescent lipid (usually rhodamine-dioleoyl phosphatidylethanolamine - rhoPE) for

fluorescent visualization. Lipids were mixed prior to interface deposition to ensure

proper mixing; this step was important for uniform and reproducible monolayers. The

monolayers were imaged on a Leica microscope using the appropriate filter cubes.

Significant evidence from past work (cited in Section 1.3.1), as well as experiments in

this chapter (Fig. 3-3a), shows that mixtures of phospholids and cholesterol form

immiscible liquid-liquid domains which separate through a significant range of surface

pressures and subphase conditions. The localization of PIP2 was imaged with respect to

a known marker for the Ld phase (rhoPE) in these liquid-liquid coexistence monolayers

using three distinct PIP2 fluorescent markers. These were: 1) a fluorescently-labeled PH

domain from PLCδ1; 2) a fluorescent gelsolin-derived PIP2-binding peptide; and 3) a

long-chain fluorescent analog of PIP2. cDNA of a GST chimera of the PH domain from

rhPLCδ1 (gift from Dr. Tobias Baumgart) was expressed in XL1-blue E. Coli and

purified using glutathione-functionalized Sepharose beads using a standard molecular

54

biology protocol (Sigma). The purified GST-PH was fluorescently-labeled by reaction

with fluorescein isothiocyanate (FITC) for 30 minutes at room temperature. The progress

of the reaction was carefully monitored because FITC reacts with solvent-accessible

primary amines on the surface of proteins, typically lysines and arginines, the same

residues known to make up the PIP2 binding pocket of PH-PLC 103. For this reason, the

labeling ratio of FITC:PH domain was minimized to maintain PIP2-binding activity.

In addition to fluorescently-labeled GST-PH chimeras, a PIP2-binding-peptide (PBP10)

synthesized using the known sequence of the PIP2 binding site of gelsolin functionalized

with a rhodamine B fluorescent group 105 and available in the Janmey lab will be used.

This reagent is known to bind PIP2 in cells, and its fluorescence makes it a logical

candidate for the proposed experiments.

Finally, lipid mixtures were doped with a small proportion (0.5%) of a fluorescently-

modified synthetic PIP2 analog (NBD-PIP2) to determine the localization of this analog

in relation to the phases formed by the phospholipid-cholesterol mixtures.

3.1.3 – Neutron scattering and FRET in LUVs

Because the putative domains/clusters of PIP2 discussed in the Chapter 2 are far below

the optical resolution limit, to complement the monolayer imaging studies described

above, two techniques with nanoscopic spatial resolution were used to study the lateral

organization of PIP2 in mixtures with an unsaturated phosphotidylcholine (SOPC) in

large unilamellar vesicles (LUVs): Fluorescence Resonance Energy Transfer (FRET)

and neutron scattering.

3.1.3a – Production of LUVs

55

LUVs were produced by a standard protocol - briefly, PIP2 and SOPC in organic solvents

were mixed followed by evaporation of the solvent, first under a stream of pure N2,

followed by vacuum drying (2 hr – overnight) to ensure complete solvent evaporation

(necessary for vesicle reproducibility). After hydration of the lipids with aqueous buffer

(150 mM NaCl, 10 mM HEPES, 0.1 mM EDTA, pH 7.4) to form a 0.3 mM lipid

solution, the aqueous lipid mixture was freeze-thawed five times by freezing on dry ice

followed by thawing at 37oC to break multilamellar structures and remove defects that

arise during rapid hydration. Finally, the lipids were homogenized by extrusion through

a filter with well-defined pores to produce monodisperse vesicles of prescribed diameter

(100 nm for FRET ; 400 nm for neutron scattering).

3.1.3b – FRET experiments

Fluorescent analogs of PIP2 (0.6%, Bodipy-TR PIP2, Molecular Probes) and SOPC

(0.7%, NBD-PC, Molecular Probes) were incorporated into these vesicles by addition of

these analogs into the vesicle solution followed by a 30 minute incubation, which has

been shown to be an efficient method for incorporation of these fluorescent lipids into the

outer leaflet of the vesicles 54 (the leaflet into which the analogs are incorporated can be

tested by quenching the fluorescence with a membrane impermeable agent). FRET was

measured by exciting the donor fluorophore at the NBD excitation maximum (450 nm)

and measuring the resulting emission spectrum from 520 nm to 700 nm. FRET ratio was

defined as the ratio of the intensity at the maximum of the acceptor emission peak (630

nm) and the intensity at the maximum of the donor emission peak (540 nm). A high

FRET ratio means that the fluorescent lipids are in close contact with each other,

suggesting mixing between the SOPC and PIP2, whereas decreased FRET means a 56

separation of Bodipy-PIP2 from NBD-SOPC suggesting the formation of laterally

segregated domains of PIP2 and SOPC.

3.1.3c - Neutron scattering

To complement the FRET experiments, which rely on fluorescent lipid analogs, domain

formation in LUVs was also assayed without fluorescence using neutron scattering. The

vesicles were prepared in the way described for the FRET experiment with the exception

that total lipid concentration was 6 mg/mL and the vesicle diameter was 400 nm. These

changes were made because of the low efficiency of neutron scattering and the difficulty

of extruding very high lipid concentrations through small pores, respectively. In neutron

scattering, contrast is provided by the relative stability of excited nuclei of different

atoms, and conveniently for biological samples, one of the greatest isotopic contrasts is

provided by hydrogen and deuterium. Although deuteration of the acyl chains of lipids

may have an effect on the physical properties of lipids, similar LUV experiments with

deuterated and protonated PC have shown identical IR spectra and phase behavior

suggesting that deuteration of lipid components would not be expected to induce

nonspecific phase separation (personal communication Dr. Arne Gericke). In the these

experiments, 10% hydrogenated PIP2 was mixed with 90% deuterated PC, and the

scattering intensity of these vesicles was measured as a function of different buffer

mixtures of D2O/H2O (following an established experimental protocol 106). Scattering

experiments were done on the Small Angle Neutron Scattering (SANS) machine at the

Intense Pulsed Neutron Source (IPNS) at Argonne National Laboratory (ANL) with

instrumental assistance by Jyotsana Lal and Ed Lang. Scattering was performed at 50oC,

above the melting temperature of the saturated components (~ 42oC for DPPC).57

By varying D2O/H2O ratio, the contrast between the sample and the buffer is varied, until

zero scattering is achieved when the scattering line density of the buffer matches that of

the sample. However, if the sample is non-uniform, i.e. if there are parts of the vesicle

that are enriched in protonated PIP2 while others are enriched in deuterated PC, then

those two parts of the vesicle can not be matched by a single D2O/H2O ratio, and instead

of zero scattering, there will be a minimum in scattering between the match point of one

of the domains and that of the other. Using this technique not only the presence of

domains in PIP2/PC vesicles, but also potentially their size, can be determined because of

the Angstrom-scale resolution of neutron scattering experiments.

3.1.3 – Edge fluctuation of liquid-liquid domains

Mixed lipid monolayers were produced as above from dimyristoylphosphatidylcholine

(DMPC), dihydrocholesterol (DChol), and rhodamine-labeled

dioleoylphosphatidylethanolamine (rhoPE). Experiments were conducted at room

temperature (~25oC) with a subphase composed of phosphate buffered saline (PBS: 7.5

mM phosphate, 140 mM NaCl) at pH 7.4 with 5 mM dithiothreitol (Sigma) to prevent

oxidation of lipids.

Following deposition of lipids on the interface, the height of the surface was adjusted by

withdrawing subphase from beyond the monolayer barriers until the monolayer could be

visualized with an inverted microscope (IX81; Olympus, Center Valley, PA) with a 60X

1.1 numerical aperture long-working distance water immersion objective with coverslip

correction (Olympus), a Texas Red filter cube (Chroma, Rockingham, VT), and a back-

illuminated electron multiplying charge-coupled device (EM-CCD) camera (ImageEM;

58

Hamamatsu, Bridgewater, NJ). Images were taken at a pixel resolution of 256x256 and

an exposure time of 0.016 seconds/frame, yielding an average frame rate of 60 fps,

accounting for the finite readout time. The pixel edge size was set at 0.264 μm, close to

the optical point spread function width of the microscope.

The monolayer was first compressed quickly (25 Å2/molecule/minute) through the

transition pressure while the surface pressure was monitored as above (Sec. 2.1.2). The

transition pressure was recorded as the last pressure at which inhomogeneity could still

be observed under our experimental conditions. After allowing for 5 minutes of

stabilization, the surface pressure was reduced at 5 Å2/molecule/minute to the highest

pressure where stable domains could be observed. For each film pressure, several >2000-

frame movies were obtained while slow monolayer flow during the imaging process was

compensated for by manual translation of the microscope stage supporting the monolayer

trough.

Extensive details about the image analysis and theoretical considerations for this

experiment are included in Sec. 3.2.4 below.

3.2 – Experimental results – Lipid segregation and domain formation in

mixed lipid systems

3.2.1 – FRET detection of PIP2 demixing in LUVs

In 100 nm vesicles comprised of 90% SOPC and 10% L-α PIP2, lateral separation of

PIP2 from the bulk component was detected by measuring FRET between fluorescent

PIP2 (Bodipy TR-PI (4,5)P2) and PC (NBD-PC). In this experiment, low FRET

suggested spatial separation of fluorescent PIP2 from fluorescent PC, hence formation of

59

PIP2 domains. The pH dependence of FRET in LUVs containing PIP2 is shown in Fig.

3-1. The relatively high acceptor peak (540 nm) and low FRET ratio (2.57) at

physiological pH in the absence of calcium (Fig. 3-1a – pH 7.4) suggests a significant

separation of the NBD-PC from its fluorescence acceptor (Bodipy-PIP2), suggesting the

presence of Bodipy-PIP2 enriched domains. Further, the pH dependence of this

phenomenon suggests that deprotonation of at least one of the monoester phosphates is

necessary for this clustering to occur, consistent with previously published results (Fig. 3-

1a and55). The inclusion of divalent calcium in the extravesicular media also increased

FRET efficiency slightly from 2.57 to 3.57 (Fig. 3-1b), suggesting that in two-component

vesicles, the presence of cations induces greater accessibility between NBD-PC and

Bodipy-PIP2. This result could either be the consequence of domain disruption by

calcium, or the formation of smaller, more dispersed cluster of PIP2 that would interact

more readily with the PC medium.

3.2.2 – Neutron scattering observation of PIP2 demixing and domains size

To validate the results from the FRET study suggesting PIP2 domains in mixed PIP2/PC

LUVs, neutron scattering was performed from similar vesicles to determine whether

domains could be observed with a measurement modality not requiring fluorescent

tracers. Specifically, the vesicles were assayed for lateral inhomogeneity using the

solvent contrast variation technique described in Sec. 3.1.3c and previous work 106.

Briefly, solvent neutron scattering length density (SLD) was varied by increasing the

proportion of D2O (deuterated water) versus H2O in the solvent. In the case of uniform

samples (e.g. laterally homogeneous vesicles), the scattering resulting from the interface

between solvent and sample will decrease as the SLD of the solvent approaches that of 60

the sample, eventually reaching zero scattering as sample and solvent achieve identical

SLD’s. This scenario was observed when neutron scattering spectra were taken with

varying concentrations of D2O/H2O from vesicles containing 90% d62-DPPC (DPPC

where a deuterium replaces every hydrogen on the acyl chains) and 10% protonated

DPPC, as seen in Fig. 3-2a below. As the sample-solvent contrast approaches zero, so

does the scattering intensity (expressed as the square root of the scattering intensity

linearly extrapolated to zero scattering vector (q) as described106), until no scattering is

observed at 80% D2O (the SLD of 80% D2O is consistent with the expected SLD of a

90% d62-DPPC sample). In contrast, lateral inhomogeneous samples with domains of

different SLDs would not be expected to achieve zero scattering since the solvent would

either match the background or the domain, but never both at the same time. This

situation is observed in the case of PIP2 containing vesicles, both with naturally derived

D-myo PIP2 and DP-PI(3,4)P2. The saturated synthetic PIP2 was included to ensure that

the observed demixing was not a consequence of acyl chain mismatch (although this

would be unexpected since all experiments were performed above the chain melting

temperatures of the lipids). For both PIP2 containing samples, non-zero minima in

scattering are observed suggesting the presence of PIP2-depedent lateral inhomogeneity

(Fig. 3-2a).

Observation of the superimposed scattering spectra from these samples leads to an

unexpected finding, namely significant scattering intensity at q = 0.05-0.15 Å -1 only

observed in samples which were designated as domain-containing by the contrast

matching experiments (Fig. 3-2b). Converting to real space, this observation suggests

inhomogeneous features of 30-90 nm in the demixed vesicles, which corresponds to a

61

domain size of 4-12 molecules (using the ~70 Å2 molecular area for PIP2 found in

Chapter 2), giving the first estimate for the size of the putative PIP2 hydrogen-bonded

clusters.

62

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

480 530 580 630 680Wavelength (nm)

Norm

alize

d inte

nsity

pH 7.4pH 3

Ratio = 6.77

Ratio = 2.57

A

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

480 530 580 630 680Wavelength (nm)

Relat

ive F

luore

scen

ce

[Ca]=0mM[Ca]=2mM

Ratio = 2.57

Ratio = 3.57

B

Figure 3-12. FRET in LUVs.FRET from NBD-PC to Bodipy-PIP2 in 100nm LUVs (90%PC/10%PIP2) as a function of (a) pH and (b)

[Ca2+]. The ratios are quantification of normalized fluorescence intensity of the acceptor peak (625 nm)

and the donor peak (540 nm). A high FRET ratio suggests proximity between the donor and acceptor

components (ie mixing of the NBD-PC and Bodipy-PIP2) while a low FRET ratio would be expected of

vesicles with lipid separation.

63

0

0.2

0.4

0.6

0.8

1

1.2

1.4

60 80 100% D2O

sqrt

(I @

q=0

)

DPPCD-myo PIP2DP-PIP2

A

0.01

0.1

1

10

100

0.001 0.01 0.1 1Normalized q (1/A)

Nor

mal

ized

scat

teri

ng in

tens

ity Demixed (with PIP2)

Mixed (no PIP2)

B

Figure 3-13. Neutron scattering from LUVs.(A) Square root of interpolated scattering intensity at scattering vector equals zero (q=0) as a function of

percentage of D2O versus H2O in the buffer. All samples are 90% deuterated DPPC with the remaining

10% protonated DPPC (circles), naturally-derived PIP2 (squares), or synthetic saturated acyl chain PIP2

(triangles). The non-zero minimum in the PIP2-containing samples suggests the presence of lateral

inhomogeneities. (B) Scattering intensity as a function of scattering vector (q) for the sample containing

10% DP-PIP2 (closed squares) and the sample containing 10% DPPC (open circles). The unexpected

scattering at q = 0.05-0.15 suggests features of 30-90 nm.

64

3.2.3 – PIP2 segregation in cholesterol-containing monolayers

To evaluate the localization of PIP2 in a more complex, and cell-relevant mixture lipid

system, monolayers of ternary mixtures of L-α PIP2, unsaturated phosphotidyl choline

(SOPC), and cholesterol were microscopically evaluated. Liquid-liquid phase separation

was observed in these mixtures at almost all monolayer compositions, consistent with

previously published results76, 107, 108. These domains are characterized as liquid-liquid

because of their circular shape, edge fluctuation, and lateral diffusivity. Domains were

visualized by doping the lipid mixture with 0.1% rhoPE which partitions to the liquid-

disordered (Ld) phase, as opposed to the cholesterol-rich Lo phase. The shape, size, and

lateral and size distribution of these domains were highly variable, and inclusion of PIP2

in the monolayer did not seem to affect either the presence of domains, or any of these

superficial characteristics (Fig. 3-3a). Upon compression of the monolayer, the

monolayers undergo a demixed-mixed transition from the domains state shown in Fig. 3-

3a to a single homogeneously-fluorescent phase. The pressure at which the transition

occurs is highly dependent on the cholesterol content of the monolayer, as shown for a

composition 5:1 SOPC:PIP2 (Fig. 3-3b). The profile of this cholesterol-dependence of

the transition pressure was nearly identical to those of phospholipids/cholesterol

monolayers without PIP2109, suggesting that PIP2 did not have a significant effect on the

stability of these domains, which was somewhat surprising due to its highly-charged

nature. Confirming this lack of PIP2-dependence, variance of PIP2 concentration at a

constant cholesterol fraction did not affect the mixed-demixed transition pressure in these

monolayers (Fig. 3-3b inset).

65

Although the presence of PIP2 did not affect liquid-liquid domain formation/stability in

monolayer mixtures of PIP2, SOPC, and cholesterol, PIP2 was laterally segregated by

these domains to the Ld phase of the monolayers (Fig. 3-4). Three distinct markers were

used to image PIP2 localization in demixed monolayers in relation to the Ld phase

markers rhodamine-PE and NBD-PC. Fluorescently labeled PIP2 (NBD-PIP2, Fig 3-4a),

the PIP2-binding gelsolin-derived peptide PBP10 (Fig. 3-4b), and a fluorescently labeled

PH domain of PLC-δ (Fig. 3-4c) all co-localized with rho-PE and NBD-PC, strongly

suggesting the exclusion of PIP2 from the cholesterol rich Lo phase and its concentration

in the Ld phase.

66

Figure 3-14. Liquid-liquid domain formation in mixed lipid monolayers.(A) Micrographs of monolayers of L-α PIP2/SOPC/cholesterol with 0.1 mol% rhodamine-PE to provide

contrast between phases. (B) Domain-to-homogeneous transition pressure as a function of cholesterol mol

% (at constant PC:PIP2 = 5:1) and (inset) PIP2 mol% (at constant cholesterol=30%).

Figure 3-15. Co-localization of PIP2 and Ld phase.Monolayers of 10% PIP2, 50% Cholesterol, and 40% SOPC co-stained with PIP2 markers (A-C) against

markers of the liquid disordered phase (rhodamine SOPE in D and F; NBD-SOPC in E). All images were

taken at 10x magnification except B and E which are 60x to provide greater detail.

67

3.3 – Experimental results – Line tension in cholesterol-DMPC monolayers

In recent years, fluid-fluid phase coexistence in lipid model systems has been the subject

of significant attention due to its proposed relationship to the phenomenon of

physiological lipid phase separation in general, specifically in the context of membrane

rafts. Of particular interest in the study of fluid-fluid phase coexistence is the

characterization of transitions from two observable immiscible phases to a single

homogeneous phase. These mixing/demixing transitions have been shown to be

functions of lipid composition110, temperature111-113, surface pressure114, 115, and degree of

crosslinking116, while compositional fluctuations on length scales below optical resolution

are also beginning to be understood117-120. Recent observation of qualitatively similar

liquid-liquid phase separation in plasma membrane-derived giant vesicles (GPMVs) has

further underlined the potential biological relevance of these model membrane

findings121.

Of particular interest is the question of how domain size is regulated122. In lipid bilayer

membranes, the only known driving force for domain coarsening is line tension at the

phase boundary. Quantification of this parameter is helpful not only for understanding

domain coarsening kinetics and thermodynamics, but also to elucidate three-dimensional

modulation of membrane shape – both in model membranes111, 123, 124 and possibly

extended to biologically relevant membrane shape transitions related to membrane

trafficking125. Additionally, variation of the interfacial line tension by membrane minority

components would suggest line active species that function as domain stabilizers or

68

disruptors in model systems, and possibly as “membrane raft” regulators in plasma

membranes of living cells.

Line tension at liquid domain boundaries has previously been examined both

theoretically and experimentally in monolayers126-129 and in bilayers111, 124, 130, 131. We have

recently obtained extremely small line tensions in fluctuating lipid bilayer domains of

giant unilamellar vesicles130. Giant vesicles, typically in the size range of a few dozen

micrometers in diameter, pose significant challenges to the accurate analysis of

experimental domain undulations due to the spherical geometry that is imaged in the

planar focal plane130. Lipid monolayers, however, are not limited in lateral dimensions

and their optically flat surface is advantageous for extended flicker spectroscopy studies.

In monolayers of 30% cholesterol and 70% dimyristoylphosphatidylcholine (DMPC),

line tension was estimated by Benvegnu and McConnell through the relaxation rate

following mechanical deformation of bolas-shaped domains to the energy minimizing

circular domain shape126 (also see Refs.132-134). This study demonstrated monolayer

domain line tension to vary by two orders of magnitude (from ~0.1 pN to more than 10

pN) depending on the monolayer surface pressure. A potential limitation of this early

work was the need for several secondary parameters and simplifying assumptions in

order to analyze the experimental data to yield line tensions126. McConnell et al. also

estimated the dipole density difference between coexisting phases from the diffusional

mobility of electrostatically trapped domains135.

Fourier power spectra of thermal domain boundary fluctuations, observable at relatively

small line tension, have first been published by Seul and Sammon136. Goldstein and

Jackson (GJ) then adapted a theory previously developed for magnetic films with phase 69

coexistence137 to the ultrathin film limit of lipid monolayers with dipolar interactions127.

The GJ theory relates the competing effects of interfacial line tension and dipolar

repulsion to thermal domain boundary in-plane undulations, and allowed analysis of the

data of Ref.136 to yield line tension and dipolar density difference close to the critical

pressure, for a single pressure and composition127. Stottrup et al. recently extended these

early measurements129; their analysis, however, neglected the potentially important

dipolar interaction which can modulate the power spectra of domain edge fluctuations. In

fluctuating lipid bilayer domains, dipolar contributions to the fluctuation spectra are not

discernable, as expected from the screening effect due to the existence of an additional

aqueous half-space130. However, the data presented here, in combination with the GJ

analysis, not only show that in monolayers dipolar interactions significantly modify

fluctuation spectra, but also demonstrate that dipolar interactions can be accurately

quantified by flicker spectroscopy.

Here, these findings are extended by applying the GJ theory to the analysis of a large data

set of time-lapse images of demixed monolayers to accurately quantify both the line

tension and the dipole density difference between the two coexisting liquid phases as a

function of surface pressure without external perturbation. Using this approach, we find

excellent agreement with published values for at 30% Dchol; we also determine these

parameters in mixtures with 35% and 40% Dchol. We furthermore obtain critical

exponents for as the surface pressure approaches the critical pressure where phase

coexistence disappears, and again find good agreement with predicted values. To our

knowledge, this is the first concurrent quantification of both the line tension and dipole

70

density differences in coexisting fluid phases in lipid monolayers as a function of surface

pressure.

3.3.1 - Capillary Wave Theory

The GJ theory relates fluctuation mode amplitudes ζn associated with mode number, n, to

the line tension, γ, and the dipole density difference, μ127. These two parameters define the

dimensionless Bond number NB = 2μ2/γ that characterizes the relative importance of

dipolar and phase boundary energies. To quadratic order in mode amplitudes (i.e. for

small elongations), the energy E of the monolayer domain is

(1)

where E0 and R0 are the energy and the radius of a non-fluctuating domain, respectively.

The mode number-dependent quantity βn is a function of NB, R0, and the thickness of the

domain h:

(2)

Note that the form of βn in Refs127, 138 contains a sign error (R. Goldstein, personal

communication). The radius R0 is related to the domain area A by R0 = (A/π)1/2. The

dependence of fit results on domain thickness h has been discussed and found to be

small127. In all subsequent analysis below we assume h = 1 nm127. In the case of negligible

dipolar interactions, NB = 0 and βn = 1/(n2 – 1), which we previously showed to be an

accurate description of lipid bilayer domain fluctuation spectra, where long range dipolar

interactions are screened130. For the case of Nb = 0, the ratios of averaged mode powers,

71

i.e. second mode <C2> divided by higher modes <Cn> = < >, will yield a straight line

with slope 1/3 when plotted against n2-1 127, 130. If NB > 0, however, the same plot will

show deviations from the 1/3 slope which increase with NB, as we experimentally

demonstrate in Figures 3-5a, 3-5d and 3-5e. The dependence of NB values on film

pressure for the composition of 30% Dchol is shown in Figure 3-5e. As expected126, 127, NB

increases as the film pressure approaches the critical pressure (of 10.1 mN/m for our

system).

Equation (2) defines conditions for the stability of the circular ground state shape towards

transitions to ground states of different symmetry127, 139-141. With increasing Bond number,

the first unstable mode is found for n = 2, and from (2), the critical Bond number below

which the circular shape is stable is obtained127

(3)

Note that this critical Bond number is dependent on domain radius, whereas NB is not.

Alternatively, for fixed Bond number, Eqn 3 can be used to define a critical radius above

which circular domains are instable127, 139-141. All fluctuation spectra examined in the

present contribution were obtained from domains with NB values below the critical Bond

number defined by (3). Figure 3-5e shows for the example of 30 mol% Dchol that all NB

values remained below the range of critical Bond numbers for the experimental domains

considered for this composition. For the condition NB < thermal equipartitioning

yields the following expression for the mode amplitudes 127:

(4)

72

where kB is Boltzmann’s constant.

3.3.2 – Data analysis

All image processing was performed using MATLAB (The Mathworks, Natick, MA).

Our code allowed selection of individual domains in multi-domain frames with a click of

a mouse; the tracing routine automatically centered and cropped to the neighborhood

around the same identified domain in all subsequent frames, thus allowing tracking and

localization frame by frame. The original gray-scale images were converted to binary via

thresholding, and the domain boundary for each image was determined from the binary

frames and parameterized as the radius function r(θ), where θ is the polar angle. Image

frames where the area of the domain changed by more than 3% were discarded. Area

changes of those magnitudes were attributed to motion blur, departure of the domain

from the field-of-view due to flow or diffusion, or other imaging artifacts. From the

domain area, A, the equivalent radius R0 = (A/π)1/2 was obtained. R0 therefore refers to the

radius of a perfect (i.e. non-fluctuating) domain with equal area.

Trace analysis was performed as previously described127, 130, 136, 142. The mode powers

<Cn> (in units of μm2) were determined through fast Fourier transform (FFT) of the

individual traces on a frame-by-frame basis and then frame averaged for each domain.

Specifically, the radial deviation Δr(θ) = r(θ) - <r(θ)>, where <r(θ)> is the average

radius, was Fourier transformed. Note that this average radius is not the same as the

equivalent radius R0 if the domain is fluctuating130. We have previously discussed the

influence of total frame number included in subsequent analysis130. The determination of

dipolar effects requires us to obtain average mode powers with high statistical

73

significance. We therefore determined mode powers from averaging FFT data for 1000

frames per domain. An example of averaged mode powers as a function of mode number

is given in Figure 3-5a. Note that, contrary to bilayer spectra, a significant upward

deviation from the 1/3 slope discussed above is observed, confirming the findings of

Goldstein and Jackson127. The set of unitless ratios βn/β2 was then fit to the experimentally

determined mode power ratios <C2>/<Cn> by varying the single fit parameter NB (Figure

3-5a). This procedure was repeated for all mode sets [n] = [2,…, nmax], where nmax ranged

from 3 to 25, resulting in 23 potentially different values of NB. These NB values were

plotted as a function of nmax, (see Figure 3-5b) and the largest mode set prior to a drop-off

in the magnitude of NB was taken to include all resolvable modes for that domain (Figure

3-5b). This mode set [n]* = [2,…,n*max] and its corresponding NB value were used in all

subsequent analyses. The range of modes included for analysis in the example of Figure

3-5 is indicated by filled symbols in Figures 3-5a and 3-5c, as opposed to open ones that

indicate excluded data points, and is further indicated by an arrow in Figure 3-5b. The

rationale for this analysis procedure is the fact that mode powers associated with higher

mode numbers n > n*max will be increasingly distorted through the effects of image

pixelization, optical resolution limit, and averaging of domain motion due to finite frame

acquisition times130. Additionally, mode sets that are too small do not contain enough data

points to yield sufficiently accurate NB values (Figure 3-5b, open symbols on the lefthand

side). The maximum number of resolvable modes depends on the size of fluctuation

amplitudes.

With both [n]* and NB determined for each individual domain, the mode powers <Cn>

were plotted against 1/βn (see Figure 3-5c) and a least-squares linear fit to the form y =

74

mx, with a slope, m, equal to kTR0/πγ (Equation 4), yielded the line tension. Small values

of 1/βn, referring to large mode numbers (n > n*max) showed a progressive upward

deviation from the expected linear relation (see open symbols in the inset of Figure 3-5c)

and were not included in the analysis. Finally, the dipole density difference μ was

obtained from the Bond number NB via NB = 2μ2/γ. Approximately 10 domains at each

pressure and composition were analyzed to obtain average values of γ and μ2 (typically

for mode power analysis of the first 1000 frames in an image sequence only, except

where mentioned below).

75

Figure 3-16. Analysis of mode power fluctuation spectra at 30 mol% Dchol.(A) FFT-determined mode power ratios <C2>/<Cn> for a single domain of radius 7.8 μm at a pressure of

8.25 mN/m at the critical composition. Closed circles are values of <C2>/<Cn> where n is an element in

the mode set [n]* included in analysis, while open circles have n values larger than n*max. The dashed line

indicates a slope of 1/3. The solid line represents the βn/β2 fit to the optimal mode set [n]*, while the dotted

lines represent changes in the fit parameter NB by ± 5% to indicate fit quality. (B) Fit parameter NB of the

same domain as in (A) as a function of mode set considered, as defined by [n] = [2,…,nmax]. The largest

mode set referring to n*max is indicated by an arrow. Closed triangles represent mode sets large enough to

minimize noise, but that also exclude higher modes distorted by aliasing and other distorting effects. (C)

Mode powers, <Cn>, of the same domain as in A) are plotted against 1/βn using the NB value from the fit

shown in A). Excellent agreement to a linear fit is seen for the modes included in the set [n]* (closed circles

directly related to those in A), but deviation from linearity is found at higher modes (open circles), which

are excluded from analysis. The slope of the linear plot is proportional to line tension (see text). (D) FFT-

determined mode power ratios <C2>/<Cn> for three individual domains at pressures of 6.5, 8, and 9 mN/m

76

at the critical composition. Deviation from the 1/3 slope line (dashed) is seen to increase as the pressure

approaches the critical pressure. (E) Average values of NB as a function of surface pressure for the critical

composition. Also plotted are the 2nd mode critical Bond numbers, N*B (2) (see Eqn 3), for domains of radii

equal to 16 (dashed line), 9 (dotted line), and 6 μm (solid line), referring to the largest, the average, and

smallest domain size included in analysis, respectively. NB is seen to increase with pressure, but remains

below N*B (2) for all domains considered.

77

3.3.3 – Line tension and dipole density results

Mixed monolayer membranes of DMPC and DChol at the critical composition (70%

DMPC / 30% DChol), were imaged at a range of surface pressures decreasing from the

critical pressure (10.1 mN/m) down to a pressure of 6.5 mN/m. Our critical pressure is in

good agreement with published values143. Image frames were analyzed to extract the bare

line tension and dipole density difference as a function of surface pressure as

described above. Both and decreased from 0.64 to 0.22 pN and 0.68 to 0.44 D/100

Å2, respectively, as the surface pressure was increased from 6.5 to 9 mN/m (Figure 3-6a).

The pressure dependence of line tension and dipole density difference, near the critical

point, can be expressed as a function of the critical exponents d and , respectively126, 144.

These relations are , and , where m and n are adjustable parameters and

the reduced surface pressure r c - , i.e. r is the deviation of the film pressure from

the critical pressure πc. For the critical composition, the exponents for the dependence of

and 2 on reduced surface pressure were 0.9 ± 0.22 and 0.35 ± 0.09, respectively (see

Figure 3-7c). These values compare favorably with those assumed by Benvegnu and

McConnell (1.0 and 0.33, respectively). The largest contribution to the uncertainty in our

critical exponents stems from the uncertainty of the measurement of c (± 0.5 mN).

The effective line tensions obtained from the relation eff = – 2 measured here are in

excellent agreement with those previously derived from the recovery of domain shape

following external distortion (see Figure 3-6b and Ref.126). Note however, that the

discussion in Ref.126 indicates a small uncertainty in their measured values due to

78

approximations inherent in their analysis approach. Hence the remarkable agreement of

our findings with those of Ref.126 could be somewhat fortuitous.

Additionally, and measured by Goldstein and Jackson from a preliminary data set

near the critical pressure match the trends observed with our data (quantitative

differences could be due to their data being taken closer to the critical point127). A recent

capillary wave theory quantification of eff129 measured a very similar line tension at high

surface pressure (8.3 mN/m), although a significantly different surface pressure

dependence (see Figure 3-6b) of was observed, associated with a critical line tension

exponent different from the value of d 1.

Measurement of 2000 consecutive frames for each domain allowed us to quantify the

effect of photo-induced oxidation on the measured parameters of and . Both

parameters were reduced by a small and statistically insignificant amount when

calculated from the third and fourth sets of 500 frames (<10 secs) of the sequences

compared to the first or second (Figure 3-6c). This reduction was associated with

significant photobleaching and might be suggestive of photo-induced generation of line

active oxidation products. This observation is in accordance with the observed reduction

of line tension by photo-induced production of cholestenone in a similar mixed

monolayer145. Note, however, that since both and are affected by photobleaching

products, these appear to have an effect not only on phase boundary properties, but also

on the bulk properties of the coexisting domains.

The large number of domains analyzed in this report allowed quantification of the

dependence of line tension on domain size. Surprisingly, a correlation was found between

79

the radius and the and values obtained for each domain, with a slight reduction in

both parameters with increasing domain size (Figure 3-6d). Although the data from a

single pressure were quite spread, normalizing all data sets to a single surface pressure

using the critical exponents showed the dependence to be significant to a p-value of 0.05

for and 0.06 for .

80

Figure 3-17. Line tension and dipole density differences at critical composition.(A) Log of (filled diamonds) and 2 (open circles) as a function of log of reduced surface pressure (c-)

at 70% DMPC / 30% DChol. Points represent the average and standard deviation of 9-13 domains per

pressure. (B) Effective line tension γeff = γ – μ2 as a function of surface pressure derived using our results

(open circles) compared with previously published data (filled diamonds – Ref. 126; filled squares – Ref. 129).

(C) Normalized (black) and 2 (lined) and standard deviation calculated from 500 frame sequences of 10

domains at = 9 mN/m. Values for each domain were normalized to and 2 calculated using the first 500

frames of that domain. (D) (filled diamonds) and 2 (open circles) as a function of domain radius at =

9 (top left), 8.75 (top right), 8.5 (bottom left), and 8.25 (bottom right) mN/m.

81

In addition to quantification of critical exponents for line tension and dipole density

difference at the critical composition (30% Dchol), these parameters were measured at

35% and 40% DChol, compositions that lie in the same binary miscibility gap as the 30%

sample, and domain edge fluctuations were observed for these additional mixtures as

expected from the monolayer phase diagram146. Although a trend of and decreasing

with surface pressure was observed for all compositions, and all data sets were fit well by

power laws, the quantitative relationships between these parameters were not identical

(Figure 3-7a and b). The critical exponents at 35% DChol were 1.2 for and 0.6 for 2,

significantly higher than those for either 30% or 40% (Figure 3-7c). Since these critical

exponents are strongly dependent on exact quantification of the transition pressure of the

monolayer, it is difficult to confidently ascribe a trend to these observations.

An interesting observation is that, despite the composition dependence of the power

law exponents for the relationships between or and r, there appears to be a similar

relationship between the ratio of these parameters ( or NB) and reduced pressure for

all three compositions tested (Figure 3-7d). This result suggests the possibility of a

composition-independent (but pressure-dependent) relationship between these

parameters, and further, that deviation from this behavior could be an indicator for the

presence/efficacy of line active components. However, as above, these relationships are

strongly dependent on transition pressure, and therefore subject to the same potential

error.

82

The experiments, analysis, and results reporting of Section 3-3 regarding the

measurement of line tension and dipole density differences using capillary wave theory

was done in collaboration with Michael Heinrich and Hannah Gelman under the

advisement of Dr. Tobias Baumgart.

83

Figure 3-18. Composition dependence of line tension and dipole density difference.(A) Same as Figure 2A at 65% DMPC / 35% DChol. Points represent the average ± standard deviation of

6-10 domains per pressure. (B) Same as Figs. 2A and 3A at 60% DMPC / 40% DChol. Points represent

the average ± standard deviation of 10-14 domains per pressure. (C) Exponents for (black) and 2

(lined) as a function of monolayer composition. Exponents calculated from slopes of linear fits to data in

Figs. 2A, 3A, and 3B. Error bars reflect the 25% uncertainty in exponents resulting from a 0.5 mN/m error

in transition pressure. (D) Average Bond number as a function of reduced surface pressure for all three

compositions (30% DChol – open circles ; 35% DChol – filled squares ; 40% DChol – filled diamonds)

with a linear fit to all data.

84

3.4 - Results discussion, caveats and significance

Building on the results of Chapter 2 showing the interplay between hydrogen-bonded

attraction and electrostatic repulsion in governing the lateral organization of PIP2 in a

pure model system, in this chapter PIP2 localization and organization was investigated in

mixed lipid systems where PIP2 was the minority component, as physiologically-

appropriate. The results of these studies confirm the conclusions from Chapter 2, namely

that local enrichment of PIP2 occurs in mixed monolayer and bilayer systems, with and

without the presence of cholesterol. Neutron scattering and FRET experiments with

bimolecular LUVs of PC (1:10) suggest the presence of 3-10 molecule PIP2

nanodomains, not only confirming the possibility of hydrogen bonded PIP2 domains

suggested in Chapter 2 and previous studies54, 55, but also providing the first evidence

regarding the possible size of these putative domains. In addition to the hydrogen-

bonded domains suggested by experiments with 2-component bilayers, more complex

monolayer studies suggest the enrichment of PIP2 in micron-scale cholesterol-dependent

domains. A significant caveat of the FRET and monolayer fluorescence experiments

described in this aim is the reliance on fluorescent lipid analogs as PIP2 tracers.

Although the use of these analogs is widespread in the study of PIP2 behavior in vitro

and in cellular experiments, simple consideration of the relative sizes of native PIP2

(1100 Da) and a fluorescent group such as 7-nitrobenz-2-oxa-1,3-diazole (NBD, MW =

200 Da) suggests that these large fluorescent groups are likely to interfere with the

biophysical properties of the native molecules, since these are inherently dependent on

the size and shape of the lipids. Additionally, since fluorescent molecules are typically

comprised of fused aromatic rings, their resulting hydrophobicity confines them to the 85

lipidic membranes, further increasing the likelihood for their interference with the

interpretations of the experiments described in this chapter. This problem has been

addressed here by employing several imaging techniques in parallel to analyze the same

phenomenon. Although each technique has unique flaws, the agreement of results across

imaging modalities confirms the validity of my results.

An additional point of concern for the monolayer domains experiments showing PIP2-

enriched domain formation is that the presence of cholesterol-dependent domains is

surface pressure dependent, with domains dissolving at pressures below those estimated

to have physiological relevance, an observation that somewhat undermines the

physiological relevance of these findings. This concern is addressed by consideration of

the specific lipids used in our system. Domains like the ones observed have been studied

extensively by several groups, with similar results regarding the dissolution of domains at

low surface pressures. However, it was also found that this dissolution was highly

dependent on the lipid species used, specifically the degree of saturation of the

phospho/sphingolipids110. It has been shown that domains very similar to the ones we

have observed can persist to higher surface pressures, both in monolayers as well as

bilayer systems where the lipid packing density is inherently in the range of

physiologically-relevant systems. Additionally, the results of experiments presented in

the following chapter will attempt to connect the findings from these simpler model

systems with a much more complex lipid mixture where qualitatively similar phase

coexistence is observed.

An unanswered question in all of these experiments, and an inherent flaw of all in vitro

lipid experiments, is the effect of curvature on the physical behavior of these systems, 86

more specifically, how the curvature of these systems relates to the physiologically

relevant curvature of membranes in the cell. Although in vivo membranes are usually

thought of as flat, planar bilayers, there are places where the curvature is high enough to

be relevant on a molecular scale, for example in the folds of the Golgi cisternae147.

Additionally, it has been observed by electron microscopy the plasma membrane is not

flat, but instead highly wrinkled, and therefore curved on a variety of scales. Although

our results show good agreement between curved bilayer and flat monolayer models, the

full effect of curvature on PIP2 domain formation has not been investigated.

3.4.1 – Line tension discussion and significance

The major advance of the analysis and results presented in Section 3.3 is the utilization of

a large data set in the framework of the Goldstein and Jackson theory to independently

measure dipole density difference and line tension in coexisting liquid phases to

determine critical exponents for the variation of these parameters as a function of surface

pressure. The data included at least 1000 image frames per domain, 10 domains at each

pressure, 6-7 surface pressures per composition, and three different compositions. This

data volume allowed both confident determinations of average and values at each

pressure, as well as quantification of the domain-to-domain variability in these

measurements. The magnitudes of the measured parameters agree remarkably well with

previously published values126, especially considering the difference in the approaches

used to derive them. It is interesting to note that the two-dimensional Ising model predicts

a value of 1/8 for the exponent, 144. Hirshfeld and Seul115, however, found the shape of

their mixed monolayer phase coexistence boundary to be in better agreement with an

exponent of 1/3, the Ising model exponent for three dimensions144, and Hagen and 87

McConnell obtained = 0.25 ± 0.7 from a set of different mixtures148. As discussed in

Ref.148, despite the molecular thickness of monolayer films, intermolecular interactions in

lipid monolayer are not truly two-dimensional, which may explain the deviation from the

2D Ising model expectation.

An important distinction between our data and previously published results is the

quantification of bare vs. effective line tension. Previous estimates of line tension in

monolayers by physical perturbations126, 128 have measured eff, which includes both the

bare line tension () quantified here and the dipolar repulsive effects: eff = . For

our measurements, dipolar contributions to effective line tension were 7-9%, increasing

slightly with film pressure for all compositions. This similar relationship between and

at the three measured compositions is interesting to note because it suggests a possible

interdependence of these parameters in the absence of line active components. It seems

plausible that line active components could affect without modifying thereby

changing the interdependence of these parameters.

The photobleaching effect observed in our experiments was not unexpected, since

previous work cited a line tension reducing effect of cholestenone produced by the photo-

induced oxidation of cholesterol145. Although dihydrocholesterol was used in our studies

to prevent oxidative effects145, this cholesterol analog can be oxidized to cholestanone149

with a similar structure to cholestenone, which could have the same line active properties.

We included 5 mM DTT in the subphase to minimize photobleaching with the aim at

maximizing the number of image frames that could be analyzed. Without DTT addition,

the illuminated area of the monolayer was bleached within 15 s (referring to less than

1000 frames) to the point that boundary tracing by thresholding became inaccurate. It is 88

interesting to note that photobleaching products appear to have the opposite effect (i.e.

lead to an increase in line tension) in bilayer vesicles, as recently found by flicker

spectroscopy in GUVs 130.

A further point of note is the composition dependence of the measured parameters and

. The exponents for the relationship of these parameters to film pressure for all

compositions were similar to the 1.0 for and 0.33 for predicted by Benvegnu and

McConnell126, but there was an apparent increase in these exponents at non-critical

concentrations. However, these exponents depend strongly on the experimental value for

the transition pressure, which is observation-dependent, with ~25% variation in

exponents expected with a 0.5 mN/m error in transition pressure. Thus, although we

observed composition-dependent variations in exponents, a systematic trend in the

concentration dependence could not be deduced from our data set.

An unexpected finding enabled by the large data set used in these experiments was the

domain size dependence of line tension and dipole density difference. Equation 3 predicts

domain shape instability for domain radii that approach a critical value. It is possible that,

as experimental domain radii get closer to the critical radius for the first instable mode,

the capillary wave theory of GJ becomes increasingly inaccurate since fluctuations may

transiently probe unstable regimes near the shape transition. This hypothesis awaits more

systematic experimental investigation.

Finally, although this analysis of the physical properties of liquid-liquid monolayer

domains did not include PIP2 in the lipid mixtures, the next step in these studies is to

evaluate the “line activity” of various membrane-associated additives/contaminants. One

89

of the most exciting candidates for being a biologically-important line active component

is PIP2, with the reasons for its candidacy described in Sec. 6-3 below.

Section 4.3.1 was written in collaboration with Michael Heinrich and Dr. Tobias

Baumgart. Parts of this chapter were adapted from work accepted for publication by the

Journal of Physical Chemistry with permission from the editor pending.

90

Chapter 4 – Bridging membrane raft model systems:

Cholesterol-dependent phase separation in Giant Plasma

Membrane Vesicles

Extensive investigation into biological lipid phase separation has confirmed the

importance of lateral compositional heterogeneity in determining the function of the

plasma membrane. This investigation has been approached from two directions: (1)

biophysical characterization of mixing phenomena in purified lipid mixtures; and (2)

biochemical characterization of cell membrane heterogeneity exploiting phase detergent

resistance differences. Although both techniques provide certain insights, their biological

relevance has been disputed due to the lack of compositional complexity, and the

requirement for low temperature and detergent-mediated cell lysis, respectively. The

recent discovery of phase separation in cell-derived Giant Plasma Membrane Vesicles

(GPMVs) has introduced the possibility of investigating lipid phase separation in a

system with appropriate biological complexity, without the requirement of membrane

solubilization by detergent. Here, GPMVs were used to investigate the cholesterol

dependence of phase behavior, specifically Lo/Ld phase abundance, temperature

dependence of phase coexistence, and phase diffusivities. In agreement with purified

lipid mixtures, cholesterol depletion reduces the Lo phase fraction, and vice versa. In

addition, the Lo phase is the majority phase in untreated vesicles, confirming recent

findings in live cells. Additionally, cholesterol level determines the temperature-

91

dependence of phase separation in a way consistent with simple lipid mixtures, and

correlates strongly with the presence of a detergent-insoluble membrane fraction in cell

lysates. Finally, fluorescence correlation spectroscopy reveals two distinctly diffusing

populations in phase-separated vesicles whose diffusivities correspond well to Lo/Ld

diffusivities in model liposomes and live cells.

4.1 - Justification for GPMV experiments

Several lines of evidence, including the identification2, 3 and characterization (reviewed

in4) of detergent resistant membrane fractions, anomalous diffusion of membrane bound

tracers8, and nanoscale aggregation of fluorescent proteins and markers7, among others5, 6,

have led to the hypothesis of cholesterol and sphingomyelin (SPM) enriched “membrane

rafts” in the plasma membrane9. These putative “rafts”, existing as stable and insoluble

domains within the bulk membrane, have been suggested to contain a variety of GPI-

linked, transmembrane, and peripheral proteins60-68, leading to their proposed role as

platforms for the organization and concentration of signaling components4. Model

systems experiments, using mixtures of synthetic lipids in monolayers76, supported

bilayers79, and giant vesicles77, 150-152, have reproduced and extensively characterized phase

demixing in mixtures of cholesterol and various phospholipids (reviewed in108) . A

consistent result across all model systems is that inclusion of cholesterol into lipid

mixtures often results in liquid-liquid phase separation into a liquid-ordered (Lo) and a

liquid-disordered phase (Ld or Lα)70, 108. The Lo phase is characterized by conformational

lipid ordering resembling that of crystalline or gel phases75, but distinguished from those

92

by a high degree of rotational and translation lipid mobility characteristic of the Lα

phase74.

Although these model system experiments have successfully recapitulated cholesterol-

and SPM-enriched phase separation, they have offered no conclusive evidence that Ld-Lo

phase immiscibility is physiologically related to the raft hypothesis. This is due in part to

the fact that the model systems employed cannot replicate, conceptually or technically,

the tremendous complexity of the plasma membrane, both in the heterogeneity of lipid

species and the inclusion of a large number and variety of membrane associated proteins

that would be expected to affect the thermodynamics of lipid-mediated demixing.

However, recent experiments using Giant Plasma Membrane Vesicles (GPMVs), cell-

derived liposomes which maintain the lipid80 and protein81 diversity of the plasma bilayer,

have shown temperature dependent liquid-liquid phase separation, similar to that

observed in model systems82. This phase separation was found to segregate known

protein and lipid markers of “lipid rafts”, as well as other physiologically important

proteins, providing a convincing link between Ld-Lo phase separation in model systems

and the “lipid raft” hypothesis in cellular plasma membranes.

4.2 - Experimental Design and Methods

4.2.1 – Cell culture and treatment

NIH-3T3 fibroblasts were cultured at 37oC in 5% CO2 in DMEM (Sigma) supplemented

with 10% Calf Serum (Gibco). For cholesterol depletion, cells at 70% confluence were

treated with 5 mM methyl-beta cyclodextrin (MBCD, Sigma) dissolved in serum-free

DMEM for 1 hr at 37oC. For cholesterol loading, the treatment was the same except the

93

MBCD was pre-loaded with cholesterol by overnight incubation with 160 g of

cholesterol (Avanti). For sphingomyelin depletion, cells were treated for 10’ with 50

mU/mL exogenous sphingomyelinase (SMase, Bacillus aureus, Sigma) in serum-free

DMEM at 37oC (the time was carefully controlled to minimize cellular effects resulting

from ceramide production).

4.2.2 – GPMV isolation and visualization

GPMVs were isolated using the PFA/DTT method as previously82 and labeled with

rhodamine B 1,2-dioleoyl phosphatidylethanolamine (rhoPE, Avanti) and naphthopyrene

(nap, Sigma) by incubation at room temp for 15 mins with 2.5 g/mL rhoPE and/or 10

g/mL nap. A chamber was created by making a square of silicon grease (Dow Corning)

on a BSA-coated coverslip, into the middle of which 20 L of labeled GPMV suspension

was deposited, followed by sealing of the chamber with another coverslip. The vesicles

were fluorescently visualized using an inverted microscope (Leica) equipped with

appropriate filter sets (red fluorescence for rhoPE ; green fluorescence for nap).

Temperature was controlled using a Peltier temperature control stage (TS-4, Physitemp).

Percent surface area covered by the Ld phase was quantified by calculating the surface

area of the spherical cap (SAcap) covered by the bright phase by:

where hcap is defined as the height of the spherical cap and equivalent to:

if Ld is the minority phase and,

if Ld is the majority phase.

94

rvesicle and rcap are defined as the radii of the vesicle and cap, respectively, and were

measured using ImageJ software (NIH).

4.2.3 – Cholesterol mol fraction quantification

Cells were grown to 70% confluence in T150 flasks (Corning Inc), treated with chl-free

or chl-loaded MBCD, and GPMVs were isolated as above (the large number of cells was

necessary to get detectable signal from the phosphate assay). The vesicle suspension was

then extracted using the Folch method153 with 3.75 mL of CHCl3:MeOH (1:2) per 1 mL

of suspension overnight at room temp. Phase separation was then induced by adding 1.25

mL ddH2O and 1.25 mL of CHCl3. The samples were then centrifuged for 10’ at 2000xg,

the top (aqueous) phase was aspirated, and the bottom (organic, lipid rich) phase was

saved. 10% of the resulting organic phase was analyzed for cholesterol concentration as

follows: the extracted lipid solution was dried under N2, rehydrated with water, vortexed

briefly, then sonicated for 30’ at room temp to produce small vesicles composed of the

extracted lipid components. These vesicles were then analyzed for their cholesterol

content using a fluorimetric enzymatic kit (Amplex Red Cholesterol Assay Kit,

Invitrogen) following the manufacturer’s instructions. The remaining organic phase was

used to quantify the phospholipid concentration using the colorimeteric inorganic

phosphate assay83. The results from these assays were then combined to determine the

relative fraction of cholesterol and phosphate, with the simplifying assumptions that the

only extracted components are cholesterol and phospholipids, and that there was one

phosphate/lipid.

4.2.3 – Quantification of SMase treatment

95

Fibroblasts were grown to 10% confluence, treated with SMase as above, washed with

phosphate buffered saline, then treated with trypsin (Gibco) to remove from the dish.

The cell suspension was then extracted using the Folch method, as above. Extracted

lipids in CHCl3 were spotted onto TLC plates along with several SM standards using

CHCl3:MeOH:Acetic acid:H2O (50:37.5:3.5:2) as the elution liquid. SM concentration

was quantified by densitometric analysis of a digital image of the TLC plate after staining

with iodine vapor using MultiGauge V3.0 image analysis software (Fujifilm).

4.2.4 – Fluorescence correlation spectroscopy

Fluorescence correlation spectroscopy was performed on GPMVs as previously described

for Giant Unilamellar Vesicles154-156. Briefly, the vesicles were labeled as described

above, with the exception that the final labeling concentration was 1 nM rhoPE (this

concentration was an important parameter for getting good signal). 515-nm laser light

was introduced into the aperture of a high numerical aperture objective (Nikon, Plan Apo,

60x, NA=1.3) through the epifluorescence port of an inverted microscope (Nikon). The

confocal volume was calibrated by measuring the correlation from free diffusion of a

known dye solution (rhodamine 6G, DT = 2.8x10-6 cm2/sec). GPMVs were then placed in

a chamber (as above), located using phase contrast and positioned such that the middle of

vesicle was superimposed on the focal laser spot. The focus was then adjusted such that

the laser spot was focused on the top of the membrane (measurements taken from the

bottom and side membranes did not yield significantly different results). The

fluorescence signal was then detected for 30 secs/measurement with an avalanche

photodiode and correlated online using a correlator card. 7-10 vesicles per condition

were measured at various focal planes with 15-20 measurements/vesicle. Repeat 96

measurements taken at the same spot were very repeatable and gave nearly identical

results.

Correlation curves (G() vs. ) seconds were fit from = 1e-4 to 5 with a two-component

two-dimensional free diffusion equation where one of the components was always the

unincorporated dye diffusing much faster than those in the vesicle ( free ~ 150-400

m2/sec).

where N is the average number of fluorescent molecules in the confocal area, D is the

diffusion time, and Cbgd is a constant defining the background correlation. The diffusion

coefficients were then defined as:

, where is the radius of the confocal volume defined by the free dye

calibration (for these experiments = 0.6 m). A small number of curves at 10oC (<5%)

were not fit well by the two-component model and required a three-component fit (circles

in Fig. 4-5a), and these were interpreted as capturing tracers in both phases during the 30

secs of acquisition.

4.2.5 – Detergent resistant membrane quantification

Detergent resistant membranes were isolated on a discontinuous sucrose gradient as

previously described157. Briefly, NIH-3T3 fibroblasts were grown to 70% confluence in

10 cm dishes (Fisher). The cells were then harvested by trypsinization, and the trypsin

was inactivated using soybean trypsin inhibitor (Sigma). The cells were then washed 2x

97

in TNE buffer (25 mM Tris-HCl, 150 mM NaCl, 5 mM EDTA, pH 7.4). After the

second wash, the cells were resuspended in TNE supplemented with 1% Triton X-100

(Sigma) and lysed for 30 mins in a temperature-controlled water bath. Following lysis,

the cells were homogenized while immersed in a temperature-controlled bath by shearing

through a 25-gauge needle (20 strokes). 1 mL of lysate was then mixed with 2 mL of

56% sucrose to make 3 mL of 40% sucrose lysate solution, which was overlayed with 7

mL of 35% sucrose, followed by 2 mL of 5% sucrose. This gradient was then

centrifuged at 270,000xg for 18 hr. 1 mL fractions were then analyzed for their

cholesterol content as above. Detergent resistant fractions were defined as fractions 1-3

from the top of the column in all samples except those extracted at 10oC. At this

temperature (as well as at 18oC) a significant amount of cholesterol was found in the

intermediate fractions 4-7 (Figure 4-6), likely reflecting the presence of a transition state

between detergent-labile and detergent-resistant membranes. For DRM quantification of

the 10oC samples, the detergent-resistant fractions were defined as those lighter than the

fraction in which no cholesterol was observed (fractions 6, 7, and 7 for the three 10oC

samples, fraction 7 in the sample shown in Fig. 4-6).

There is some evidence that cholesterol may be enriched in the detergent-resistant

fractions9, which suggests that estimating DRM fraction by cholesterol quantification

may overstate the abundance of the detergent-resistant phase. This is not a significant

concern because not only would this error be systematic and not affect the results shown

in Fig. 4-4a, but also a recent study suggests that the DRM phase may have equal

cholesterol concentration to the detergent-labile phase158.

98

4.3 - Experimental results - Cholesterol-dependent phase separation in

GPMVs

4.3.1 – Lo phase comprises majority of GPMV surface area

GPMVs derived from adherent NIH 3T3 fibroblasts were stained with rhoPE, a Ld phase

tracer, and observed by fluorescence to quantify their relative abundance of Ld and Lo

phase. Representative pictures in Fig. 4-1a and the quantification in Fig. 4-1e show that

more than 70% of the surface area of these cell-derived vesicles is comprised of the Lo

(raft) phase. This finding is inconsistent with the prevailing view of liquid-ordered

domains as isolated and inabundant lipid rafts, instead suggesting the possibility that the

plasma membrane has the potential to exist as a majority liquid-ordered continuum

interrupted by liquid-disordered domains.

99

Figure 4-19. Cellular cholesterol level determines Lo/Ld ratioRepresentative images of rhoPE-stained GPMVs from untreated cells (A), cholesterol-depleted cells (B),

and cholesterol-loaded cells (C). Quantification of fraction of phase separated vesicles (D) and relative

surface area covered by the Ld phase in phase separated vesicles (E). Error bars are standard deviations

from 35-50 vesicles measured per condition; results are representative of three different experiments. Both 100

treatments suggest that lower cholesterol level in cells leads to more Lo phase in GPMVs and that the Lo

phase in control cells is the majority phase. Scale bars are 5 m.

101

4.3.2 – Cholesterol depletion/loading affects phase separation and Lo phase fraction

Compared to detergent-labile cell membranes, detergent resistant membrane fractions are

enriched with cholesterol and sphingomyelin, leading to the hypothesis that lipid rafts are

liquid-ordered membrane structures enriched in, and possibly dependent on, the presence

of these plasma membrane lipids. We have modulated the levels of cholesterol and

sphingomyelin in cells prior to GPMVs isolation, as well as in isolated GPMVs, to

determine whether changes in the abundance of these putative lipid raft components

would affect the formation and relative abundance of the two liquid phases in these

vesicles.

Depletion of cellular cholesterol by treatment with 5 mM MBCD decreased the

cholesterol mol fraction by almost 30% in GPMVs derived from those cells and resulted

in significant changes to their phase behavior. The fraction of vesicles that showed

detectable microscopic phase separation at 10oC decreased from essentially 100% to less

than 80% (Fig. 4-1d). Additionally, the relative abundance of the liquid-disordered (non-

raft) phase was nearly doubled, increasing from 28% to more than 65% of the surface

area of the vesicles (Fig. 4-1e).

Conversely, loading cells with cholesterol had the opposite effect on their plasma

membrane derived vesicles. Cells were loaded by treatment with cholesterol-saturated

MBCD, resulting in an increase of GPMV cholesterol mol fraction from 49% to 58%.

While this cholesterol loading did not affect the demixed vesicle fraction (Fig. 4-1d),

there was a significant increase in the abundance of the Lo (raft) phase, from 72% to 95%,

resulting in these GPMVs appearing nearly dark with very small, bright, Ld patches (Fig.

102

4-1c). Similar results were observed when isolated GPMVs were cholesterol loaded

(Figs. 4-2b and 4-2d).

Although depletion of plasma membrane cholesterol resulted in drastic changes in the

phase abundance of GPMVs derived from those cells, cellular mechanisms for replacing

membrane cholesterol prevented depletion below ~35 mol%. To determine the effect of

more extensive cholesterol depletion, the compensatory mechanisms were circumvented

by direct MBCD treatment of the vesicles following their isolation from cells. This

treatment had an interesting and unexpected effect. At the 40-60x magnification used for

imaging all other conditions, these vesicles appeared small and uniformly bright.

However, further magnification (100x) revealed non-circular, jagged, and ribbon-like

domains (Fig. 4-2c) similar in morphology to gel phase domains observed in cholesterol-

free GUVs where demixing was the consequence of acyl chain length differences

between the component phospholipids156. To our knowledge, this is the first microscopic

observation of gel-liquid phase coexistence in a mixture with the complexity of a cellular

membrane.

103

Figure 4-20. Direct cholesterol modulation in GPMVs.Representative images of rhoPE-stained GPMVs from untreated cells without treatment (A), treated with 5

mM cholesterol-saturated MBCD for 1 hr (B), and treated with 5 mM cholesterol-free MBCD for 1 hr (C).

Quantification of relative surface area covered by the Ld phase in control and cholesterol-loaded vesicles

(D). The cholesterol-depleted group was not quantified because gel-liquid area fractions could not be

quantified using the methods described. Error bars are standard deviations from 35-50 vesicles measured

per condition; results are representative of three different experiments. Direct cholesterol modulation in

GPMVs confirms the cell results, and additionally suggests that complete cholesterol depletion induces gel-

fluid phase coexistence in GPMVs. Scale bars are 5 m.

104

4.3.3 – Sphingomyelin depletion has no effect on Lo phase

In contrast to the significant effects of modulating cholesterol levels on the relative

abundance of the Lo to the Ld phase, depletion of sphingomyelin had little to no

observable effect. Treatment with 50 mU/mL of exogenous sphingomyelinase (SMase)

for 10 mins resulted in ~50% reduction in cellular SM, as quantified by TLC (data not

shown). However, this treatment resulted in no significant difference in Lo/Ld fraction in

vesicles derived from these cells (Fig. 4-3a and 4-3b). Similarly, direct treatment of

GPMVs with the same SMase concentration resulted in no effects of phase separation at

10oC or relative Lo/Ld abundance (Fig. 4-3c and 4-3d). These results suggest that while

cholesterol is a critical determinant of raft phase abundance and demixing in GPMVs,

sphingomyelin is not.

105

Figure 4-21. Sphingomyelin depletion has no effect on GPMV phase behavior.Quantification of phase separated vesicle fraction (A and C) and relative surface area covered by the Ld

phase (B and D) in vesicles from sphingomyelin-depleted cells (A and B) and in vesicles from untreated

cells that were SM-depleted following isolation (C and D). Error bars are standard deviations from 35-50

vesicles measured per condition; results are representative of three different experiments.

106

4.3.4 – Cholesterol level determine phase separation temperature

Cholesterol mol fraction has been shown to be an important parameter determining phase

coexistence in numerous simplified lipid model systems76, 77, 150, prompting the hypothesis

that the same effect may be observed in the complex lipid and protein mixture present in

GPMVs. Cellular cholesterol levels were manipulated as above and the temperature-

dependent phase separation of GPMVs derived from those cells was measured. As

expected82, for all cholesterol levels, GPMVs were fluorescently uniform at high

temperatures (>40oC) and macroscopically phase separated below ~15oC (Fig. 4-4b).

Analysis of phase separation at intermediate temperatures revealed that the fraction of

phase separated GPMVs was sigmoidally dependent on temperature, suggesting either

multiple distinct first-order phase transitions, or a higher-order transition. While any

single vesicle undergoes transition at a particular temperature, the average transition

temperature occurred over a temperature range (5-10oC), probably reflecting the

cholesterol heterogeneity in individual GPMVs. Vesicles undergoing phase transition

exhibited a variety of domain morphologies, but could be grouped into two general types

of transitions: 1 – a spinodal or striping domain dissipation consistent with a critical

phase transition159 (Fig. 4-4c1 and 4-4c2); and 2 – domain edge fluctuation away from a

circular morphology followed by gradual dissipation or melting of the domains (Fig. 4-

4c3 and 4-4c4). These transitions were very similar to those observed in simple 3-

component mixtures of phospholipids and cholesterol77.

Comparing the temperature-dependence of GPMVs from cells with varying cholesterol

levels revealed an exciting result (Fig 4-4b). Loading cells with cholesterol (from 45% to

50% cholesterol) induced a decrease in the average GPMV transition temperature from 107

24 to 21oC, while depleting cellular cholesterol increased this temperature to 32oC.

Further, this depletion also resulted in a significant fraction (15%) of vesicles that were

phase separated at the physiological temperature of 37oC, suggesting the possibility of

cholesterol-dependent phase separation in complex membranes at physiological

conditions. This observation corresponds well to the previously observed phase

separation induced by cholesterol depletion of live cells at physiological temperature160.

4.3.5 – Correlation between GPMV phase separation and presence of DRM

To determine the relationship between Lo-Ld phase coexistence in cell-derived plasma

membrane vesicles and the presence of a low density membrane fraction in detergent-

lysed whole cells, the temperature dependence of these two distinct membrane

phenomena was investigated. As noted above, the temperature profile of phase

coexistence in GPMVs from untreated cells followed a relatively abrupt transition from

essentially entirely phase-separated to microscopically uniform vesicles between 20-

25oC. Interestingly, quantification of the temperature-dependent abundance of the mass

of cholesterol in detergent resistant membrane fractions yielded a very similar

temperature profile, with the detergent resistant fractions making up 20-25% of the total

cholesterol mass below the phase transition temperature of the GPMVs, but only <10%

above the transition temperature (Fig. 4-4a). This unexpected correlation suggests that

detergent insolubility and the existence of a Lo phase in these complex mixtures are

related phenomena, and that both may be related to the existence of “membrane rafts”.

108

Figure 4-22. Cholesterol determines temperature-dependent phase separation.(A) Correlation between the temperature dependence of phase separation in GPMVs from untreated cells

(red points; red line is a sigmoidal fit to the data) and the relative mass of the detergent resistant membrane

fraction from untreated cells (striped bars; error bars are standard deviations from three experiments). (B)

Temperature-dependence of phase separation in GPMVs isolated from untreated cells (45 mol% chol -

black diamonds), cholesterol-depleted cells (33 mol% chol - red circles), and cholesterol-loaded cells (50

mol% chol - blue squares). Temperature profiles are representative of three repeats. Pictures are GPMVs

from untreated cells stained with rhoPE and naphthopyrene at 10oC (left) and 37oC (right). (C)

109

Representative images of GPMVs undergoing phase transitions from phase separated at low temperatures

to uniform at high temperatures. All images taken between 20-22oC. Images (1) and (2) show spinodal

decomposition while images (3) and (4) show edge rippling and gradual domain fingering and melting.

Scale bars are 5 m. Work in Fig 4-4c was done in collaboration with Jon Madara.

4.3.6 – Lipid tracer diffusivity is 3x slower in Lo phase than Ld phase

Translational and rotational diffusivity differences are one of the distinguishing

characteristics of ordered versus disordered fluid phases in model lipid vesicles 75.

Additionally, lipid and protein diffusivity are major determinants of the cellular

distribution and corresponding functions of plasma membrane components. As such,

lipid diffusivity was quantified in both phase separated and uniform GPMVs by

Fluorescence Correlation Spectroscopy (FCS) measurement of a fluorescent tracer lipid

incorporated into the vesicles. At a temperature at which GPMVs are known to separate

into two liquid phases (10oC), we observed two distinct types of correlation curves, one

type that were fit well by two-component simple diffusion models (Fig. 4-5a - diamonds

and squares) and a second best fit by models with three diffusive components (Fig. 4-5a -

circles). We interpreted the two components in the former curves to correspond to a

combination of tracers incorporated into vesicles and unincorporated freely diffusing dye

molecules. The latter curves likely correspond to an averaged correlation of tracers in the

two phases (in addition to the unincorporated dye), and were quite rare (~5% of all

measurements). The histogram in Fig. 5b shows two distinct distributions of diffusion

coefficients calculated from exponential fits to correlation data at 10oC fitted to normal

distributions which correspond to average diffusivities of 1.0 and 4.0 m2/sec. The

correlation data at 37oC suggest a single population of diffusion coefficients whose mean

was roughly equivalent to the faster diffusing component at 10oC (Fig. 4-5c). 110

Figure 4-23. Phase separation induces two distinct populations of diffusivities. (A) Representative correlation curves ( G() vs. ) of rhoPE diffusing in GPMVs isolated from untreated

cells. All curves taken at 10oC where all vesicles are phase-separated. Curves represent the slowly

diffusing population (squares ; D = 1 m2/sec), the quickly diffusing population (diamonds ; D = 5

m2/sec), and a curve that includes tracers from both populations (circles ; D = 0.5 and 7.5 m2/sec). Fits

are simple two-component diffusion models (three-component for the curve that includes a fast and slow

component) where one of the components is the unincorporated dye (D ~ 200 m2/sec). (B and C)

Histograms of diffusion coefficients obtained by fitting to correlation data taken from phase-separated

vesicles (10oC - B) and microscopically uniform vesicles (37oC – C) showing a single diffusing population

of tracers in uniform vesicles and two distinct populations in phase-separated vesicles (bold lines are

111

Gaussian fits to all data while thin lines in (B) show the component Gaussians). Histograms are from 68-81

measurements on 7-9 vesicles/condition. FCS experiments performed and analyzed in collaboration with

Dr. Pramit Choudhourie under the advisement of Dr. Feng Gai.

112

Figure 4-24. Triton-extracted cholesterol distribution in sucrose gradients.

Temperature-dependent percent mass of cholesterol in the fractions of a sucrose step gradient where the

first fraction is the lightest and the twelfth is the densest. Representative profiles from three trials are

shown for 4oC (black squares), 10oC (red diamonds), 18oC (blue triangles), 27oC (green diamonds), 30oC

(purple circles), and 37oC (black open squares).

113

4.4 - Results discussion, caveats and significance

The data presented here provide a bridge between results of simplified model systems

regarding the phase behavior of lipid membranes and the much more complex protein

and lipid mixtures of cellular plasma bilayers, in addition to novel findings regarding the

effects of cholesterol and sphingomyelin on phase behavior in cell-derived model

systems. The cholesterol depletion data of Figs. 4-1 and 4-2 are consistent with both

monolayer and bilayer experiments that have shown cholesterol-dependent formation of a

liquid-ordered phase, and the abolition or reduction of that phase when cholesterol was

depleted76, 77, 109, 150. Additionally, the induction of a non-liquid gel phase by wholesale

depletion of cholesterol (Fig. 4-2c) is consistent with Lα/so separation in GUVs absent of

cholesterol156 as well as diffusivity measurements in cholesterol-depleted live cells161.

Finally, the cholesterol mol fraction dependence of the phase separated-to-mixed

transition temperature (Fig. 4-4b) agrees strongly with the same dependence measured in

model liposomes77. This agreement is particularly striking in that not only the trends, but

also the quantitative values for the transition temperatures, seem to agree strongly

between three-component GUVs and the tremendously complex cell-derived vesicles.

On the opposite extreme of complexity from 2- and 3-component lipid models of

membrane rafts are the experiments that provided the original basis for the raft

hypothesis – isolations that showed the existence of detergent-resistant membrane

fractions. These experiments also suggest the existence, or at least possibility, of

biophysically and biochemically distinct membrane phases in whole-cell lysates, however

due to the requirement for low temperature and detergent treatment, their relevance to

114

physiological phase separation has been intensely scrutinized162. The strong correlation

between the existence of a separated Lo phase in GPMVs (and simpler lipid mixtures) and

the presence of a detergent resistant membrane fraction (shown in Fig. 4-4a) suggests that

these may be distinct observations of the same phenomenon, namely lipid fluid phase

coexistence in a biologically relevant context.

A striking initial finding was that the Lo or raft phase in vesicles derived from untreated

cells was the dominant phase comprising more than 60% of the surface area of the

GPMVs. This finding is inconsistent with the prevailing view of membrane rafts as

small, isolated domains in the bulk liquid disordered membrane, instead suggesting a

continuous raft phase. Although the detergent resistant part of a typical membrane

preparation is a small fraction of the total lipids3, both detergent extraction and low

temperatures could induce artifacts into these experiments that are avoided with GPMVs.

It may be argued that the GPMV preparation induces its own set of artifacts, however

detailed spectroscopic characterization of individual lipid species derived using this

method revealed compositions consistent with those expected from plasma membrane

and intermediate between a detergent resistant raft fraction and a whole cell lipid prep

(that would include cholesterol-poor organelle membranes)80. Additionally, the idea of a

percolating raft phase is consistent with recent measurements of diffusivity of raft and

non-raft protein and lipid markers163, electron-spin resonance in live cells164, and single

cell detergent extractions160.

A surprising result from this work was the lack of dependence of GPMV phase fraction

on the abundance of cellular sphingomyelin. Along with cholesterol and certain types of

protein species, detergent resistant membrane fractions are known to be enriched in 115

sphingolipids, most prominently sphingomyelin. This fact led to the hypothesis,

supported by experimental data165, 166, that membrane rafts were dependent on a specific

association between cholesterol and ceramide derivatives. However, it is possible that

SM is enriched in DRMs because of the length and saturation of its acyl chains, which

tend to be longer and more saturated compared to glycerol-based lipids167, rather than a

function of the sphingosyl backbone or phosphocholine headgroup. This possibility is

likely in light of our data, since hydrolysis by SMase would affect the polar part of the

SM while leaving the hydrophobic, cholesterol-interfacing chains unperturbed.

The FCS measurement of lipid diffusivity in GPMVs agree surprisingly well with

previous measurements in both cells and model systems. The three-fold difference in

lateral mobility between the two phases is almost exactly the same as was measured by

fluorescence recovery in DMPC-cholesterol bilayers at physiological temperature75.

Additionally, the magnitudes of diffusion coefficients measured in Lo and Ld phases are

very close to the diffusion coefficients measured here, strongly suggesting that the 1.0

m2/sec component corresponds to the Lo phase of our GPMVs while the faster

component is likely the Ld phase. The diffusivity differences and magnitudes measured

here correspond very well to those measured by FCS for Ld and Lo phase markers in raft-

composition GUVs154 as well as to small-scale diffusivities recently measured in live cells

by optical tweezers168, underlining the remarkable agreement not just in phase separation,

but also in the properties of those phases, between live cells, GPMVs, and purified lipid

systems. The single diffusion constant at 37oC confirms the presence of a continuous Ld

phase above the phase separation transition temperature in these vesicles.

116

The dependence of raft phase abundance and separation in GPMVs on the presence and

levels of membrane cholesterol suggests interesting questions regarding the involvement

of raft-dependent processes in the pathobiology of hypercholesterolemia. While most

cells have finely tuned mechanisms for regulating plasma membrane cholesterol (SREBP

review), conditions such as heightened plasma cholesterol or increased intracellular

cholesterol esters (as in atherosclerotic foam cells) could induce long-term changes in

cellular cholesterol levels. Based on our results, varying cholesterol levels can regulate

the presence and abundance of a demixed raft phase, which has been implicated in

adaptive immune system signaling through T-cell receptors169. This raft-dependent

immune signaling could be responsible for the variety of physiological changes

concomitant with hypercholesteromia, including the currently idiopathic induction of

local inflammatory responses at atherosclerotic lesions.

The figures and writing in this chapter were adapted from a submitted manuscript with

copyright permission pending.

117

Chapter 5 – Conclusions

With the discovery and characterization of the cellular importance of phosphoinositides,

gangliosides, lysophospholipids, arachadonic acid, and membrane rafts, the physiological

importance of lipidic species has become fully appreciated. However, currently, much

more is known about the structure and biological effectors and interactions of these

molecules than about the physical structures and organization that is are inherent from

their amphiphilicity and essential to their biological functions.

In this dissertation, the lateral organization of lipids in various biologically-relevant

model systems was examined with the purpose of addressing the hypothesis that lateral

lipid organization in planar model systems can be affected by variation of

physiologically-relevant factors and that this variation can impact the biologically-

relevant properties of the component lipids. This hypothesis was investigated in model

systems of increasing complexity, initially determining the intermolecular interaction of

PIP2 in a pure monolayer, followed by determining PIP2 organization and line tensions

in mixed lipid systems, and eventually exploring phase separation and coexistence in a

mixture featuring the immense lipid and protein diversity of the intact plasma membrane.

The results of these studies are described in detail in the previous chapters, and the

conclusions arising from those results are briefly described below.

5.1 – A combination of electrostatics and hydrogen bonding determine PIP2

organization

118

Experiments on monolayers of pure, naturally-derived PI(4,5)P2 suggest that contrary to

the expectation that electrostatic repulsion, due to the high valence and molecular

packing of these lipids, would dominate the intermolecular interactions in this system,

hydrogen bonding plays an important role in regulating the lateral organization of these

lipids. The initial evidence for this hypothesis was the expanding effect of subphase

salts, a result inconsistent with electrostatic shielding of charged groups. Although this

ionic strength effect is partly explained by a decrease in surface electrostatic potential

(modeled in Section 2.3), the varying effects of different monocationic salts, the

correlation of the magnitude of expansion with the Hoffmeister series, as well as the

variance of the results with other PIP2 isomers precluded a purely electrostatic

explanation for the observed phenomena. Finally, experiments with non-ionic hydrogen-

bond disrupting factors including urea, trehalose, and high temperature confirmed the

relevance of attraction through intermolecular, water-mediated hydrogen bonding in

considering the interactions of PIP2 in planar systems. Collectively, these results suggest

that regulation of PIP2 intermolecular attraction may be an important mechanism

underlying the specificity PIP2’s interaction with its many binding partners and

conferring the necessary promiscuity that is the hallmark of the unique functionality of

this lipid.

5.2 – Theoretical modeling of the electrostatic contribution to surface

pressure of charged monolayers

Unlike univalent charged phospholipids for which charge separation is greater than the

Bjerrum length and electrostatic contributions to monolayer surface pressures are

119

negligible compared to steric and dipolar effects, multivalent polyphosphoinositides form

monolayers in which electrostatic contributions are significant. A theoretical model is

developed to calculate the magnitude of electrostatic contributions to surface pressures

for any lipid with known valence and pKa values. Electrostatic contributions lead to

significant surface pressures at molecular areas of more than 4 times the steric size of the

lipid, and electrostatic screening by increased salt concentration leads to two opposing

effects, increased deprotonation and increased surface pressures for compressed

monolayers, and decreased repulsions and subsequent lower pressures for highly

expanded monolayers. These effects are significant at physiological conditions and

perhaps play a role in the unique functionality of polyphosphoinositides in the structure

and activity of cell membranes that cannot be reproduced by more abundant, but

univalent, anionic lipids such as phosphatidylserine. The value of the theoretical analysis

presented here is both its ability to predict many of the observed phenomena, confirming

the importance of electrostatics in determining membrane organization, as well as in the

prediction of an attractive interaction that would confirm existing experimental results.

5.3 – PIP2 domain formation and segregation in mixed lipid systems

To address the concern from the pure lipid experiments regarding the physiological

relevance of experiments with pure PIP2, domain formation was investigated with mixed

lipid systems including PIP2. Two distinct nanoscopic methods were used to assay for

lateral inhomogeneity in sub-micron 2-component bilayer vesicles comprised of mixtures

of PC and PIP2. FRET experiments confirmed previous results that significant

segregation of PIP2 away from PC exists in these vesicles, and that this segregation is

120

dependent on pH and calcium concentration. Additionally, the unexpected reduction in

segregation with decreasing pH (where electrostatic repulsion should have been reduced)

confirmed hydrogen bonding between neighboring charged groups as a possible

mechanism for interaction. These results were verified without the requirement of

fluorescent tracers by neutron scattering studies from bilayer vesicles. Domain formation

was confirmed by a solvent contrast matching experiment that showed that PIP2-

containing vesicles (in contrast to those without PIP2) could not be matched by a single

buffer, strongly suggesting the presence of lateral inhomogeneity. Additionally, analysis

of the scattering spectra allowed for the estimation of the size of these putative domains

as 3-12 molecules in diameter.

Adding another level of complexity, the localization of PIP2 in mixed monolayers

containing cholesterol was assayed. As shown in extensive previous research,

cholesterol-containing lipid systems are known to phase separate at a variety of different

conditions into two immiscible liquid phase termed the liquid ordered and liquid

disordered phase. Similar phase separation was observed upon inclusion of PIP2 into

these mixtures. The surface pressure dependence of phase coexistence was measured and

determined to be a strong function of cholesterol concentration, as expected from

previous experiments, but not dependent on PIP2 fraction. This result was surprising in

light of the unusual physicochemical properties of PIP2, and showed that the cholesterol-

induced packing of phospholipids into Lo domains was energetically dominant over the

electrostatics of PIP2. Despite the lack of effect of PIP2 on cholesterol-induced domains,

these domains did affect PIP2 localization, in that fluorescent visualization of PIP2 with

three distinct markers showed that it strongly partitioned into the more disordered,

121

phospholipid-enriched Ld phase. This finding suggests that the previously shown

presence of PIP2 in detergent-resistant membrane fractions is likely due to some specific

cellular mechanism (e.g. the binding of PIP2 by a raft-partitioning protein) rather than a

function of PIP2’s preference for ordered lipid environments.

5.4 – Line tension and dipole density differences in cholesterol-containing

monolayers

By examining thermally induced fluctuations of domain boundaries in mixed monolayers

of DMPC and DChol, the line tension and dipole density differences between coexisting

fluid monolayer domains were independently quantified. These parameters were

dependent on a reduced pressure, expressed as the difference between film pressure and

the miscibility transition pressure, with the exponents closely matching those predicted

by previous estimates. Both parameters were weakly dependent on photobleaching and

domain radius, in accordance with previously published results. Finally, quantification of

the relationship between line tension and dipole density difference at three different

monolayer compositions suggests that this relation has a characteristic magnitude in the

absence of line active compounds, and that modulation of this relationship may be

indicative of line activity. These results suggest that the method of flicker spectroscopy

could contribute to the identification of line active biologically relevant components, of

which PIP2 is a likely candidate.

5.5 – Cholesterol-dependent phase separation in GPMVs

Giant plasma membrane vesicles (GPMVs) are a novel model system combining the

experimental convenience of synthetic lipid vesicles with the biological complexity of the

122

intact plasma membrane. In chapter 4, the GPMV system was used to characterize the

cholesterol and sphingomyelin dependence of the observed liquid-liquid phase

separation, and found that cholesterol is a critical component determining the coexistence

of two liquid phases in this system. Quantification of the proportion of Lo/Ld phase in the

GPMVs suggests that far from being Lo “rafts” in a “sea” of Ld bulk membrane, the Lo

phase makes up the majority of the GPMVs, and that the cholesterol content of GPMVs

significantly alters the relative abundance of the two liquid phases. Additionally, a strong

correlation between phase separation in GPMVs and the presence of a detergent-resistant

membrane fraction in cell lysates suggested that detergent resistance and lipid phase

separation may be related processes, and that both may correspond to an aspect of the

“membrane raft” hypothesis. It was also demonstrated that the cholesterol content of the

plasma membrane derived vesicles affects the temperature stability of liquid-liquid phase

coexistence, a finding that is consistent with model system studies and which can further

the understanding of the role of cholesterol in the existence/stability of domains in the

plasma membrane. Finally, quantification of the diffusivity of fluorescent tracers in

GPMVs by fluorescence correlation spectroscopy (FCS) showed that in GPMVs that do

not phase separate, there is a single observed diffusion time for the fluorescent tracer,

whereas in phase-separated GPMVs there are two distinct diffusivities (one

approximately 4x slower than the other), likely corresponding to the diffusivity difference

between the Lo and Ld phase.

123

Chapter 6 – Future Directions

6.1 – Calcium-induced mesoscopic domains in PIP2-containing monolayers

In collaboration with David Christian and Dr. Dennis Discher, microscopic domains have

recently been observed in 2-component monolayers of SOPC and PIP2. When divalent

calcium ions (in the form of CaCl2) are added underneath monolayers of 25% L-α PIP2

and 75% SOPC doped with 0.5% NBD-PIP2, small, diffraction-limited, bright spots are

observed upon a formerly uniform background (Fig. 6-1). Concomitant with the

appearance of these domains is a marked pressure drop, consistent with the observed

contraction of the monolayer observed in Fig. 2-3 above. This pressure drop is [Ca2+]

dependent and saturable (Fig. 6-1), and saturates at a magnitude commensurate with the

calcium-induced contraction observed in Fig. 2-3 (i.e. at p = 20 mN/m, where the

difference in pressure is ~20 mN/m in pure PIP2 monolayers, the pressure drop is

~5mN/m with 25% PIP2 in the monolayer). Both effects are specific for the presence of

PIP2, as monolayers of SOPC doped with fluorescent PC did not show a pressure drop of

fluorescent non-uniformity. Additionally, both the calcium-induced monolayer

contraction and microscopically-observable domain formation were pH-dependent, with

the total contraction decreasing as the pH was decreased, and domain formation only

being observed when non-negligible contraction was observed (Fig. 6-1). Finally,

quantification of the dose-dependence of the calcium-contraction effect suggests that the

Kd for calcium binding to these PIP2 containing monolayer is approximately 3 M, a

concentration that is very relevant to cellular processes. These results are an exciting

new direction for the study of biologically-relevant PIP2 lateral organization. Divalent 124

calcium is an import regulator and intermediate signaling moiety in numerous cellular

processes, and these results suggest that part of its functionality is related to the ability to

bind, concentrate, and/or PIP2.

125

Figure 6-25. Calcium-induced domain formation in PIP2 containing monolayers.

Total pressure drop as a function of subphase Ca2+ concentration at pH 7.5 (diamonds), pH 6 (circles), pH

4.5 (squares) and pH 3 (triangles). Error bars are standard deviations from 3 repeats. Black/green symbols

indicate conditions at which lateral inhomogeneity/domain formation was observed whereas all green

symbols indicate the lack of microscopically observable inhomogeneity. For all experiments, monolayers

were 25% L-α PIP2, 75% SOPC, with 0.5% NBD-PIP2 as the fluorescent marker. Results suggest that a

calcium-induced pressure drop is coincident with the visualization of mesoscale bright domains. Scale bar

is 20 m.

These preliminary results prompt further investigation into this phenomenon.

Specifically, the effects of other subphase ions will need to be assayed. Since PIP2

resides primarily in the cytoplasmic leaflet of the plasma bilayer, a relevant divalent

cation to consider in this context is Mg2+, which is much more abundant and could

potentially affect PIP2 in a similar manner to Ca2+. Additionally, the dependence of the

calcium contraction and domain formation effect on subphase ionic strength will need to

126

be determined, as current experiments have been conducted at a non-physiological ionic

strength (~5 mM).

An interesting observation with this system has been the heterogeneity in domain sizes

and shapes. At various times, in addition to small bright dots, non-circular “ribbon-like”

domains have been observed, as well as large micron-size circular domains. It will be

interesting to investigate the factors that determine the shape, size, and appearance of

these domains (e.g. time, rate of Ca2+ addition, pH, total calcium, PIP2 mol%).

Additionally, it will be interesting to determine whether the phenomena observed are

sensitive to the proportion of PIP2 in the monolayer.

Finally, it is possible that other polycationic factors may be able to induce the same

domain formation as has been observed here for divalent calcium. Namely, it would be

interesting to determine whether polycations (such as spermine or putrescine) as well as

polybasic peptides (such as the MARCKS polybasic domains) or whole proteins (such as

the PIP2 binding exocyst protein Exo70) are able to induce the same effects.

6.2 – Continuation of neutron scattering experiments

The neutron scattering results presented in Section 3.2.2 provide an exciting new

direction in the study of lateral heterogeneity in bilayer vesicles, specifically the assaying

of domain formation without the requirement of a fluorescent tracer that may confound

the interpretation of the results. However, the results presented are somewhat

preliminary, and further, more complete experimentation will need to be done before

definitive conclusions can be drawn. The specific drawback of the results presented here

is that domain formation was observed with either DP-PI(3,4)P2, which is not the isomer

127

of wide interest, or L-α PIP2 in a matrix of DPPC. In the case of L-α PIP2, although the

experiments were performed above the gel-liquid transition of the DPPC, differences in

acyl chain composition may account for the observed inhomogeneity. The next step in

these experiments would be to repeat the experiments presented in Section 3.2.2 with DP-

PI(4,5)P2 to determine whether the same demixing would be observed. Additionally, the

significance of the “bump” between q = 0.05 and 0.15 Å-1 will need to be confirmed with

more control experiments with domain-forming and non-domain mixtures (possibly acyl

chain mismatched lipids like DMPC and DSPC for a mixture that is known to phase

separate).

6.3 – Monolayer behavior of PIP3

The novelty of many of the results presented in this dissertation derives from the fact that

these are the first measurements of PIP2 in a monolayer system, likely because

conventional wisdom suggested that this lipid was too polar to withstand compression in

a planar context (which indeed it is in bilayers, forming large micelles). The same has

been assumed about the enzymatic product of PIP2 phosphorylation, phosphotidylinositol

(3,4,5) trisphosphate (PIP3), which in itself is an important and unique signaling

intermediate. It would be interesting to repeat many of the experiments described in this

work with PIP3 to determine whether the same intermolecular forces are relevant in

regulating the organization of this important lipid.

6.4 – Effect of PIP2 on line tension in cholesterol-containing monolayers

As mentioned, the technique to quantify line tension described in Sec. 3-3 appears to be

particular amenable to the identification and quantification of line activity of minority

128

components in lipid monolayers. The presence of such a line active component could act

as a two-dimensional “emulsifier”, stabilizing the interface between the two dissimilar

fluid phases and possibly introducing new and unexpected effects. Speculation about the

molecular identity of such a line active component suggests that, just as lipids act as

organic-aqueous emulsifiers, a line active component would need to interact with both

phases by having one portion preferring a disordered environment and the other

preferring the ordered phase. One such candidate molecule is PIP2, whose natural acyl

chains are typically composed of one highly unsaturated fatty acid (arachadonic acid -

20:4) and one fully saturated fatty acid. Therefore, it will be very interesting the quantify

the line tension and dipole density difference of cholesterol-induced domains with the

presence of PIP2 to determine whether the biological importance of this lipid extends to

modulation of macroscopic phase separation.

6.5 – Influence of lipid composition perturbation on demixing in GPMVs

In Chapter 4, it was shown that modulation of cellular cholesterol had a great effect of the

behavior, stability, and abundance of two coexisting liquid phases in cell-derived Giant

Plasma Membrane Vesicles. Since cellular cholesterol was controlled in these studies

using a non-physiological chemical treatment (i.e. MBCD) it would be interesting to

extend these studies to more physiological methods of varying cholesterol, such as

cholesterol-carrying lipoproteins (i.e. LDL, VLDL, HDL) or pharmacological factors that

interfere with cholesterol synthesis (e.g. statins). Additionally, these methods could be

used to study whether there are cell-type dependent, as well as cell-cycle dependent,

129

differences in plasma membrane cholesterol, and whether these would influence phase

separation in GPMVs derived under those conditions.

In addition to modulating cholesterol levels, it would be appropriate to examine whether

variance of other lipid component fractions would affect GPMV phase behavior.

Although the enzymatic degradation of sphingomyelin seemed to have no effect on phase

behavior, an interesting hypothesis to test would be that the presence of long-chain and/or

saturated fatty acids is integral to Lo phase formation/abundance. Pharmacological as

well as cellular methods (e.g. siRNA) exist that could be used to interfere with the

biosynthesis of such fatty acids to determine the cellular effects as well as GPMV phase

behavior of such perturbations.

6.6 – Protein sorting in membrane rafts

Despite extensive characterization of protein content in membrane rafts (through

detergent resistant membrane fractions), an open question remains regarding the

structural and physical factors determining protein partitioning between raft and non-raft

phases in the plasma membrane. Although the prevailing view is that partitioning

between membrane phases depends on the length of the transmembrane (TM) helix, this

view is supported solely by the fact that Lo regions of model bilayers are thicker than Ld

ones, not by any direct evidence from protein-based experiments. As raft association has

been shown to be a major determinant of protein localization, interaction, and ultimately

function, the lack of a determining mechanism for protein association with rafts remains a

glaring hole in the understanding and manipulation of raft-based phenomena. This

shortcoming could be addressed by investigating the structural determinants of protein

130

sorting into membrane rafts by systematic mutation of various domains of the raft-

partitioning proteins versus non-raft proteins. These mutants could then be assayed for

their raft-partitioning on the basis of several distinct definition of raft-preference,

including Lo phase localization in GPMVs, presence in detergent resistant fraction in cold

detergent lysates, and measurement of their diffusion characteristics using FCS or optical

tweezers.

131

References

1. Singer, S. J. & Nicholson, G. L. The fluid mosaic model of the structure of cell membranes.

Science 175, 720-31 (1972).

2. Brown, D. A. & Rose, J. K. Sorting of GPI-anchored proteins to glycolipid-enriched

membrane subdomains during transport to the apical cell surface. Cell 68, 533-44 (1992).

3. Fiedler, K., Kobayashi, T., Kurzchalia, T. V. & Simons, K. Glycosphingolipid-enriched,

detergent-insoluble complexes in protein sorting in epithelial cells. Biochemistry 32, 6365-73

(1993).

4. Simons, K. & Toomre, D. Lipid rafts and signal transduction. Nat Rev Mol Cell Biol 1, 31-9

(2000).

5. Harder, T., Scheiffele, P., Verkade, P. & Simons, K. Lipid domain structure of the plasma

membrane revealed by patching of membrane components. J Cell Biol 141, 929-42 (1998).

6. Wilson, B. S., Pfeiffer, J. R. & Oliver, J. M. Observing FcepsilonRI signaling from the inside

of the mast cell membrane. J Cell Biol 149, 1131-42 (2000).

7. Varma, R. & Mayor, S. GPI-anchored proteins are organized in submicron domains at the

cell surface. Nature 394, 798-801 (1998).

8. Schutz, G. J., Kada, G., Pastushenko, V. P. & Schindler, H. Properties of lipid microdomains

in a muscle cell membrane visualized by single molecule microscopy. Embo J 19, 892-901

(2000).

9. Simons, K. & Ikonen, E. Functional rafts in cell membranes. Nature 387, 569-72 (1997).

10. Toner, M., Vaio, G., McLaughlin, A. & McLaughlin, S. Adsorption of cations to

phosphatidylinositol 4,5-bisphosphate. Biochemistry 27, 7435-43 (1988).

11. van Paridon, P. A., de Kruijff, B., Ouwerkerk, R. & Wirtz, K. W. Polyphosphoinositides

undergo charge neutralization in the physiological pH range: a 31P-NMR study. Biochim

Biophys Acta 877, 216-9 (1986).

132

12. Brockman, H. L., Applegate, K. R., Momsen, M. M., King, W. C. & Glomset, J. A. Packing

and electrostatic behavior of sn-2-docosahexaenoyl and -arachidonoyl phosphoglycerides.

Biophys. J. 85, 2384-96 (2003).

13. Trauble, H. in Structure of Biological Membranes (eds. Abrahamsson, S. & Pascher, I.) 509-

550 (Plenum Press, New York

London, 1977).

14. Sacre, M. M. & Tocanne, J. F. Importance of glycerol and fatty acid residues on the ionic

properties of phosphatidylglycerols at the air-water interface. Chem. Phys. Lipids 18, 334-54

(1977).

15. Trauble, H. & Eibl, H. Electrostatic effects on lipid phase transitions: membrane structure

and ionic environment. Proc. Natl. Acad. Sci. U S A 71, 214-9 (1974).

16. Helm, C. A., Laxhuber, L., Lösche, M. & Möhwald, H. Electrostatic interactions in

phospholipid membranes I: Influence of monovalent ions. Colloid & Polymer Science 264, 46

(1986).

17. Thuren, T. et al. Evidence for the control of the action of phospholipases A by the physical

state of the substrate. Biochemistry 23, 5129-5134 (1984).

18. Lawrence, T., Willoughby, D. A. & Gilroy, D. W. Anti-inflammatory lipid mediators and

insights into the resolution of inflammation. Nat Rev Immunol 2, 787-95 (2002).

19. Esposti, M. D. Lipids, cardiolipin and apoptosis: a greasy licence to kill. Cell Death Differ 9,

234-6 (2002).

20. Yin, H. L. & Janmey, P. A. Phosphoinositide regulation of the actin cytoskeleton. Annu. Rev.

Physiol. 65, 761-89 (2003).

21. Eling, T. E. & Glasgow, W. C. Cellular proliferation and lipid metabolism: importance of

lipoxygenases in modulating epidermal growth factor-dependent mitogenesis. Cancer

Metastasis Rev 13, 397-410 (1994).

22. Czech, M. P. PIP2 and PIP3: Complex Roles at the Cell Surface. Cell 100, 603 (2000).

133

23. McLaughlin, S., Wang, J., Gambhir, A. & Murray, D. PIP(2) and proteins: interactions,

organization, and information flow. Annu. Rev. Biophys. Biomol. Struct. 31, 151-75 (2002).

24. Ferrell, J. E., Jr. & Huestis, W. H. Phosphoinositide metabolism and the morphology of

human erythrocytes. J. Cell Biol. 98, 1992 (1984).

25. Rameh, L. E. & Cantley, L. C. The Role of Phosphoinositide 3-Kinase Lipid Products in Cell

Function. Journal of Biological Chemistry 274, 8347 (1999).

26. Berridge, M. J. & Irvine, R. F. Inositol trisphosphate, a novel second messenger in cellular

signal transduction. Nature 312, 315 (1984).

27. Janmey, P. A. & Lindberg, U. Cytoskeletal regulation: rich in lipids. Nat. Rev. Mol. Cell

Biol. 5, 658-66 (2004).

28. Gilmore, A. P. & Burridge, K. Regulation of vinculin binding to talin and actin by

phosphatidyl-inositol-4-5-bisphosphate. Nature 381, 531-5 (1996).

29. Janmey, P. A., Iida, K., Yin, H. L. & Stossel, T. P. Polyphosphoinositide micelles and

polyphosphoinositide-containing vesicles dissociate endogenous gelsolin-actin complexes and

promote actin assembly from the fast-growing end of actin filaments blocked by gelsolin.

Journal of Biological Chemistry 262, 12228 (1987).

30. Lanier, L. M. & Gertler, F. B. Actin cytoskeleton: thinking globally, actin' locally. Curr Biol

10, R655-7 (2000).

31. Rohatgi, R., Ho, H.-y. H. & Kirschner, M. W. Mechanism of N-WASP Activation by CDC42

and Phosphatidylinositol 4,5-bisphosphate. The Journal of Cell Biology 150, 1299 (2000).

32. Rohatgi, R., Nollau, P., Ho, H.-Y. H., Kirschner, M. W. & Mayer, B. J. Nck and

Phosphatidylinositol 4,5-Bisphosphate Synergistically Activate Actin Polymerization

through the N-WASP-Arp2/3 Pathway. Journal of Biological Chemistry 276, 26448 (2001).

33. Martin, T. F. J. PI(4,5)P2 regulation of surface membrane traffic. Curr. Op. Cell Biol. 13,

493 (2001).

134

34. Varnai, P. et al. Inositol lipid binding and membrane localization of isolated pleckstrin

homology (PH) domains. Studies on the PH domains of phospholipase C delta 1 and p130. J.

Biol. Chem. 277, 27412-22 (2002).

35. Hilgemann, D. W., Feng, S. & Nasuhoglu, C. The complex and intriguing lives of PIP2 with

ion channels and transporters. Sci. STKE 2001, RE19 (2001).

36. Cremona, O. & De Camilli, P. Phosphoinositides in membrane traffic at the synapse. J. Cell

Sci. 114, 1041-52 (2001).

37. Tolias, K. & Carpenter, C. in Biology of Phosphoinositides (ed. Cockroft, S.) (Oxford

University Press, Oxford, 2000).

38. Stephens, L., McGregor, A. & Hawkins, P. in Biology of Phosphoinositides (ed. Cockroft, S.)

(Oxford University Press, Oxford, 2000).

39. Woscholski, R. & Parker, P. in Biology of Phosphoinositides (ed. Cockroft, S.) (Oxford

University Press, Oxford, 2000).

40. Lemmon, M. & Ferguson, K. in Biology of Phosphoinositides (ed. Cockroft, S.) (Oxford

University Press, Oxford, 2000).

41. Hamada, K., Shimizu, T., Matsui, T., Tsukita, S. & Hakoshima, T. Structural basis of the

membrane-targeting and unmasking mechanisms of the radixin FERM domain. Embo J 19,

4449-62 (2000).

42. Boguslavsky, V. et al. Effect of monolayer surface pressure on the activities of

phosphoinositide-specific phospholipase C-beta 1, -gamma 1, and -delta 1. Biochemistry 33,

3032-7 (1994).

43. James, S. R., Paterson, A., Harden, T. K., Demel, R. A. & Downes, C. P. Dependence of the

activity of phospholipase C beta on surface pressure and surface composition in

phospholipid monolayers and its implications for their regulation. Biochemistry 36, 848-55

(1997).

135

44. Laux, T. et al. GAP43, MARCKS, and CAP23 modulate PI(4,5)P(2) at plasmalemmal rafts,

and regulate cell cortex actin dynamics through a common mechanism. J. Cell Biol. 149,

1455-72 (2000).

45. Pike, L. J. & Miller, J. M. Cholesterol depletion delocalizes phosphatidylinositol

bisphosphate and inhibits hormone-stimulated phosphatidylinositol turnover. J. Biol. Chem.

273, 22298-304 (1998).

46. Huang, S. et al. Phosphatidylinositol-4,5-Bisphosphate-Rich Plasma Membrane Patches

Organize Active Zones of Endocytosis and Ruffling in Cultured Adipocytes. Mol. Cell. Biol.

24, 9102 (2004).

47. Tran, D. et al. Cellular distribution of polyphosphoinositides in rat hepatocytes. Cell Signal.

5, 565-81 (1993).

48. van Rheenen, J., Achame, E. M., Janssen, H., Calafat, J. & Jalink, K. PIP2 signaling in lipid

domains: a critical re-evaluation. EMBO J. 24, 1664-73 (2005).

49. van Rheenen, J. & Jalink, K. Agonist-induced PIP(2) hydrolysis inhibits cortical actin

dynamics: regulation at a global but not at a micrometer scale. Mol. Biol. Cell 13, 3257-67

(2002).

50. Wang, J. et al. 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, 34401-12 (2002).

51. Wang, J., Gambhir, A., McLaughlin, S. & Murray, D. A computational model for the

electrostatic sequestration of PI(4,5)P2 by membrane-adsorbed basic peptides. Biophys. J.

86, 1969-86 (2004).

52. Zhang, W., Crocker, E., McLaughlin, S. & Smith, S. O. Binding of peptides with basic and

aromatic residues to bilayer membranes: phenylalanine in the myristoylated alanine-rich C

kinase substrate effector domain penetrates into the hydrophobic core of the bilayer. J. Biol.

Chem. 278, 21459-66 (2003).

136

53. Gambhir, A. et al. Electrostatic sequestration of PIP2 on phospholipid membranes by

basic/aromatic regions of proteins. Biophys. J. 86, 2188-207 (2004).

54. Redfern, D. A. & Gericke, A. Domain formation in phosphatidylinositol

monophosphate/phosphatidylcholine mixed vesicles. Biophys. J. 86, 2980-92 (2004).

55. Redfern, D. A. & Gericke, A. pH-dependent domain formation in phosphatidylinositol

polyphosphate/phosphatidylcholine mixed vesicles. J. Lipid Res. 46, 504-15 (2005).

56. Kooijman, E. E. et al. What makes the bioactive lipids phosphatidic acid and

lysophosphatidic acid so special? Biochemistry 44, 17007-15 (2005).

57. Holopainen, J. M., Subramanian, M. & Kinnunen, P. K. Sphingomyelinase induces lipid

microdomain formation in a fluid phosphatidylcholine/sphingomyelin membrane.

Biochemistry 37, 17562-70 (1998).

58. Niemela, P. S., Hyvonen, M. T. & Vattulainen, I. Influence of chain length and unsaturation

on sphingomyelin bilayers. Biophys. J. 90, 851-63 (2006).

59. Pike, L. J. Rafts defined: a report on the Keystone Symposium on Lipid Rafts and Cell

Function. J Lipid Res 47, 1597-8 (2006).

60. Mukherjee, S. & Maxfield, F. R. Role of membrane organization and membrane domains in

endocytic lipid trafficking. Traffic 1, 203-11 (2000).

61. Parton, R. G. & Lindsay, M. Exploitation of major histocompatibility complex class I

molecules and caveolae by simian virus 40. Immunol Rev 168, 23-31 (1999).

62. Manes, S. et al. Membrane raft microdomains mediate lateral assemblies required for HIV-1

infection. EMBO Rep 1, 190-6 (2000).

63. Sheets, E. D., Holowka, D. & Baird, B. Membrane organization in immunoglobulin E

receptor signaling. Curr Opin Chem Biol 3, 95-9 (1999).

64. Janes, P. W., Ley, S. C., Magee, A. I. & Kabouridis, P. S. The role of lipid rafts in T cell

antigen receptor (TCR) signalling. Semin Immunol 12, 23-34 (2000).

65. Gupta, N. & DeFranco, A. L. Lipid rafts and B cell signaling. Semin Cell Dev Biol 18, 616-26

(2007).

137

66. Brown, K. D. & Hansen, M. R. Lipid Raft Localization of ErbB2 in Vestibular Schwannoma

and Schwann Cells. Otol Neurotol 29, 79-85 (2008).

67. Incardona, J. P. & Eaton, S. Cholesterol in signal transduction. Curr Opin Cell Biol 12, 193-

203 (2000).

68. Wary, K. K., Mariotti, A., Zurzolo, C. & Giancotti, F. G. A requirement for caveolin-1 and

associated kinase Fyn in integrin signaling and anchorage-dependent cell growth. Cell 94,

625-34 (1998).

69. Manes, S., del Real, G. & Martinez, A. C. Pathogens: raft hijackers. Nat Rev Immunol 3,

557-68 (2003).

70. Simons, K. & Vaz, W. L. Model systems, lipid rafts, and cell membranes. Annu Rev Biophys

Biomol Struct 33, 269-95 (2004).

71. Eroglu, C., Brugger, B., Wieland, F. & Sinning, I. Glutamate-binding affinity of Drosophila

metabotropic glutamate receptor is modulated by association with lipid rafts. Proc Natl

Acad Sci U S A 100, 10219-24 (2003).

72. Schroeder, R. J., Ahmed, S. N., Zhu, Y., London, E. & Brown, D. A. Cholesterol and

sphingolipid enhance the Triton X-100 insolubility of glycosylphosphatidylinositol-anchored

proteins by promoting the formation of detergent-insoluble ordered membrane domains. J

Biol Chem 273, 1150-7 (1998).

73. Brown, D. A. & London, E. Functions of lipid rafts in biological membranes. Annu Rev Cell

Dev Biol 14, 111-36 (1998).

74. Gally, H. U., Seelig, A. & Seelig, J. Cholesterol-induced rod-like motion of fatty acyl chains

in lipid bilayers a deuterium magnetic resonance study. Hoppe Seylers Z Physiol Chem 357,

1447-50 (1976).

75. Almeida, P. F., Vaz, W. L. & Thompson, T. E. Percolation and diffusion in three-component

lipid bilayers: effect of cholesterol on an equimolar mixture of two phosphatidylcholines.

Biophys J 64, 399-412 (1993).

138

76. Keller, S. L., Anderson, T. G. & McConnell, H. M. Miscibility Critical Pressures in

Monolayers of Ternary Lipid Mixtures. Biophysical Journal 79, 2033 (2000).

77. Veatch, S. L. & Keller, S. L. Separation of liquid phases in giant vesicles of ternary mixtures

of phospholipids and cholesterol. Biophys J 85, 3074-83 (2003).

78. Mukherjee, S. & Maxfield, F. R. Membrane domains. Annu Rev Cell Dev Biol 20, 839-66

(2004).

79. Dietrich, C., Volovyk, Z. N., Levi, M., Thompson, N. L. & Jacobson, K. Partitioning of Thy-

1, GM1, and cross-linked phospholipid analogs into lipid rafts reconstituted in supported

model membrane monolayers. Proc Natl Acad Sci U S A 98, 10642-7 (2001).

80. Fridriksson, E. K. et al. Quantitative analysis of phospholipids in functionally important

membrane domains from RBL-2H3 mast cells using tandem high-resolution mass

spectrometry. Biochemistry 38, 8056-63 (1999).

81. Scott, R. E., Perkins, R. G., Zschunke, M. A., Hoerl, B. J. & Maercklein, P. B. Plasma

membrane vesiculation in 3T3 and SV3T3 cells. I. Morphological and biochemical

characterization. J Cell Sci 35, 229-43 (1979).

82. Baumgart, T. et al. Large-scale fluid/fluid phase separation of proteins and lipids in giant

plasma membrane vesicles. Proc Natl Acad Sci U S A 104, 3165-70 (2007).

83. Kates, M. Techniques of Lipidology (Elsevier Science Publishers B.V, Amsterdam, 1986).

84. Demel, R. A., Geurts van Kessel, W. S., Zwaal, R. F., Roelofsen, B. & van Deenen, L. L.

Relation between various phospholipase actions on human red cell membranes and the

interfacial phospholipid pressure in monolayers. Biochim. Biophys. Acta 406, 97-107 (1975).

85. Cacace, M., Landau, E. & Ramsden, J. The Hofmeister series: salt and solvent effects on

interfacial phenomena. Quar. Rev. Biophys. 30, 241-277 (1997).

86. Flanagan, L. A. et al. The structure of divalent cation-induced aggregates of PIP2 and their

alteration by gelsolin and tau. Biophys. J. 73, 1440-7 (1997).

87. Shah, D. O. & Schulman, J. H. Binding of metal ions to monolayers of lecithins,

plasmalogen, cardiolipin, and dicetyl phosphate. J. Lipid Res. 6, 341 (1965).

139

88. Papahadjopoulos, D. Surface properties of acidic phospholipids: interaction of monolayers

and hydrated liquid crystals with uni- and bi-valent metal ions. Biochim. Biophys. Acta 163,

240-54 (1968).

89. Hoffmeister, K. M. et al. Mechanisms of Cold-induced Platelet Actin Assembly. J. Biol.

Chem. 276, 24751 (2001).

90. Nelson, P. C., Radosavljevic, M. & Bromberg, S. Biological physics: energy, information, life

(W.H. Freeman, New York, 2003).

91. McCormack, D., Carnie, S. & Chan, D. Calculations of Electric Double-Layer Force and

Interaction Free Energy between Dissimilar Surfaces. J Coll Int Sci 169, 177-96 (1995).

92. Manciu, M. & Ruckenstein, E. Role of the Hydration Force in the Stability of Colloids at

High Ionic Strengths. Langmuir 17, 7061-7070 (2001).

93. Chen, Y. & Nelson, P. Charge-reversal instability in mixed bilayer vesicles. Physical Review

E 62, 2608 (2000).

94. Abramson, M. B., Katzman, R., Wilson, C. E. & Gregor, H. P. Ionic Properties of Aqueous

Dispersions of Phosphatidic Acid. Journal of Biological Chemistry 239, 4066 (1964).

95. Eibl, H. & Woolley, P. Electrostatic interactions at charged lipid membranes. Hydrogen

bonds in lipid membrane surfaces. Biophys Chem 10, 261-71 (1979).

96. Teissie, J., Prats, M., Soucaille, P. & Tocanne, J. F. Evidence for conduction of protons along

the interface between water and a polar lipid monolayer. Proc Natl Acad Sci U S A 82, 3217-

21 (1985).

97. Liepina, I., Czaplewski, C., Janmey, P. & Liwo, A. Molecular dynamics study of a gelsolin-

derived peptide binding to a lipid bilayer containing phosphatidylinositol 4,5-bisphosphate.

Biopolymers 71, 49-70 (2003).

98. Langner, M., Cafiso, D., Marcelja, S. & McLaughlin, S. Electrostatics of phosphoinositide

bilayer membranes. Theoretical and experimental results. Biophys. J. 57, 335-49 (1990).

140

99. Lopez Cascales, J. J. & Garcia de la Torre, J. Effect of lithium and sodium ions on a charged

membrane of dipalmitoylphosphatidylserine: a study by molecular dynamics simulation.

Biochim. Biophys. Acta 1330, 145-56 (1997).

100. Pandit, S. A. & Berkowitz, M. L. Molecular Dynamics Simulation of

Dipalmitoylphosphatidylserine Bilayer with Na+ Counterions. Biophys. J. 82, 1818 (2002).

101. Lemmon, M. A., Ferguson, K. M., O'Brien, R., Sigler, P. B. & Schlessinger, J. Specific and

high-affinity binding of inositol phosphates to an isolated pleckstrin homology domain. Proc.

Natl. Acad. Sci. U S A 92, 10472-6 (1995).

102. Harlan, J. E., Hajduk, P. J., Yoon, H. S. & Fesik, S. W. Pleckstrin homology domains bind to

phosphatidylinositol-4,5-bisphosphate. Nature 371, 168-70 (1994).

103. Harlan, J. E., Yoon, H. S., Hajduk, P. J. & Fesik, S. W. Structural characterization of the

interaction between a pleckstrin homology domain and phosphatidylinositol 4,5-

bisphosphate. Biochemistry 34, 9859-64 (1995).

104. Rauch, M. E., Ferguson, C. G., Prestwich, G. D. & Cafiso, D. S. Myristoylated alanine-rich C

kinase substrate (MARCKS) sequesters spin-labeled phosphatidylinositol 4,5-bisphosphate

in lipid bilayers. J. Biol. Chem. 277, 14068-76 (2002).

105. Cunningham, C. C. et al. Cell permeant polyphosphoinositide-binding peptides that block

cell motility and actin assembly. J Biol Chem 276, 43390-9 (2001).

106. Knoll, W. et al. Small-angle neutron scattering of aqueous dispersions of lipids and lipid

mixtures. A contrast variation study. Journal of Applied Crystallography 14, 191-202 (1981).

107. Rice, P. A. & McConnell, H. M. Critical Shape Transitions of Monolayer Lipid Domains.

Proceedings of the National Academy of Sciences 86, 6445 (1989).

108. McConnell, H. M. & Vrljic, M. Liquid-liquid immiscibility in membranes. Annu Rev

Biophys Biomol Struct 32, 469-92 (2003).

109. Radhakrishnan, A. & McConnell, H. Condensed complexes in vesicles containing cholesterol

and phospholipids. Proceedings of the National Academy of Sciences 102, 12662 (2005).

141

110. Keller S.L., Anderson T.G. & McConnell H.M. Miscibility Critical Pressures in Monolayers

of Ternary Lipid Mixtures. Biophysical Journal 79, 2033 - 2042 (2000).

111. Baumgart, T., Hess, S. T. & Webb, W. W. Imaging coexisting fluid domains in biomembrane

models coupling curvature and line tension. Nature 425, 821 - 824 (2003).

112. Veatch, S. L. & Keller, S. L. Organization in Lipid Membranes Containing Cholesterol.

Physical Review Letters 89, 268101-1 - 4 (2002).

113. Veatch, S. L. & Keller, S. L. Separation of Liquid Phases in Giant Vesicles of Ternary

Mixtures of Phospholipids and Cholesterol. Biophysical Journal 85, 3074 - 3083 (2003).

114. Subramaniam, S. & McConnell, H. Critical Mixing in Monolayer Mixtures of Phospholipid

and Cholesterol. J Phys Chem 91, 1715-1718 (1987).

115. Hirshfeld, C. L. & Seul, M. Critical Mixing in Monomolecular Films - Pressure-Composition

Phase-Diagram of a 2-Dimensional Binary Mixture. Journal De Physique 51, 1537-1552

(1990).

116. Hammond, A. T. et al. Crosslinking a lipid raft component triggers liquid ordered-liquid

disordered phase separation in model plasma membranes. Proceedings of the National

Academy of Sciences of the United States of America 102, 6320-6325 (2005).

117. Veatch S., Soubias O., Keller SL. & Gawrisch K. Critical fluctuations in domain-forming

lipid mixtures. PNAS 104, 17650 - 17655 (2007).

118. Heberle, F. A., Buboltz, J. T., Stringer, D. & Feigenson, G. W. Fluorescence methods to

detect phase boundaries in lipid bilayer mixtures. Biochimica Et Biophysica Acta-Molecular

Cell Research 1746, 186-192 (2005).

119. Silvius, J. R. Fluorescence Energy Transfer Reveals Microdomain Formation at

Physiological Temperatures in Lipid Mixtures Modeling the Outer Leaflet of the Plasma

Membrane. Biophysical Journal 85, 1034 - 1045 (2003).

120. Nielsen L.K., Bjornhom T. & Mouritsen O.G. Critical Phenomena: Fluctuations caught in

the act. Nature 404, 352 - 352 (2000).

142

121. Baumgart T. et al. Large scale fluid/fluid phase separation of proteins and lipds in giant

plasma membrane vesicles. PNAS 104, 3165 (2007).

122. Frolov, V. A. J., Chizmadzhev, Y. A., Cohen, F. S. & Zimmerberg, J. "Entropic traps"in the

kinetics of phase separation in multicomponent membranes stabilize nanodomains.

Biophysical Journal 91, 189-205 (2006).

123. Baumgart, T., Das, S., Webb, W. W. & Jenkins, J. T. Membrane elasticity in giant vesicles

with fluid phase coexistence. Biophysical Journal 89, 1067-1080 (2005).

124. Tian, A., Johnson, C., Wang, W. & Baumgart, T. Line Tension at Fluid Membrane Domain

Boundaries Measured by Micropipette Aspiration. Physical Review Letters 98, 208102

(2007).

125. Liu, J., Kaksonen, M., Drubin, D. G. & Oster, G. Endocytic vesicle scission by lipid phase

boundary forces. Proceedings of the National Academy of Sciences of the United States of

America 103, 10277-10282 (2006).

126. Benvegnu D.J. & McConnell H.M. Line Tension between Liquid Domains in Lipid

Monolayers. Journal of Physical Chemistry 96, 6820 - 6824 (1992).

127. Goldstein, R. E. & Jackson, D. P. Domain Shape Relaxation and the Spectrum of Thermal

Fluctuations in Langmuir Monolayers. Journal of Physical Chemistry 98, 9626-9636 (1994).

128. Wurlitzer, S., Steffen, P. & Fischer, T. M. Line tension of Langmuir monolayer phase

boundaries determined with optical tweezers. Journal of Chemical Physics 112, 5915-5918

(2000).

129. Stottrup, B. L., Heussler, A. M. & Bibelnieks, T. A. Determination of line tension in lipid

monolayers by Fourier analysis of capillary waves. Journal of Physical Chemistry B 111,

11091-11094 (2007).

130. Esposito, C. et al. Flicker spectroscopy of thermal lipid bilayer domain boundary

fluctuations. Biophysical Journal 93, 3169-3181 (2007).

143

131. Blanchette CD, Lin WC, Orme CA, Ratto TV & Longo ML. Using Nucleation Rates to

Determine the Interfacial Line Tension of Symmetric and Asymmetric Lipid Bilayer

Domains. Langmuir 23, 5875 - 5877 (2007).

132. Zou, L., Wang, J., Basnet, P. & Mann, E. K. Line tension and structure of smectic liquid-

crystal multilayers at the air-water interface. Physical Review E 76, - (2007).

133. Wintersmith, J. R. et al. Determination of interphase line tension in Langmuir films.

Physical Review E 75, - (2007).

134. Alexander, J. C. et al. Domain relaxation in Langmuir films. Journal of Fluid Mechanics

571, 191-219 (2007).

135. McConnell, H., Rice, P. & Benvegnu, D. BROWNIAN MOTION OF LIPID DOMAINS IN

ELECTROSTATIC TRAPS IN MONOLAYERS. J Phys Chem 94, 8965-8968 (1990).

136. Seul, M. & Sammon, M. J. Competing Interactions and Domain-Shape Instabilities in a

Monomolecular Film at an Air-Water Interface. Physical Review Letters 64, 1903 - 1906

(1990).

137. Langer S.A., Goldstein R.E. & Jackson D.P. Dynamics of labyrinthine pattern formation in

magnetic films. Physical Review A 46, 4894 - 4904 (1992).

138. Lubensky, D. K. & Goldstein, R. E. Hydrodynamics of monolayer domains at the air-water

interface. Physics of Fluids 8, 843-854 (1996).

139. Lee, K. Y. C. & Mcconnell, H. M. Quantized Symmetry of Liquid Monolayer Domains.

Journal of Physical Chemistry 97, 9532-9539 (1993).

140. Deutch, J. M. & Low, F. E. Theory of Shape Transitions of 2-Dimensional Domains. Journal

of Physical Chemistry 96, 7097-7101 (1992).

141. Mcconnell, H. M. Harmonic Shape Transitions in Lipid Monolayer Domains. Journal of

Physical Chemistry 94, 4728-4731 (1990).

142. Seul, M., Sammon, M. J. & Monar, L. R. Imaging of Fluctuating Domain Shapes - Methods

of Image-Analysis and Their Implementation in a Personal Computing Environment.

Review of Scientific Instruments 62, 784-792 (1991).

144

143. Keller S.L., Radhakrishnan A. & McConnell H.M. Saturated Phospholipids with Hight

Melting Temperatures Form Complexes with Cholesterol in Monolayers. Journal of Physical

Chemistry B 104, 7522 - 7527 (2000).

144. Rowlinson JS. & Widom B. Molecular Theory of Capillarity (Clarendon Press, Oxford,

1982).

145. Benvegnu, D. & McConnell, H. Surface Dipole Densities in Lipid Monolayers. J Phys Chem

97, 6686-6691 (1993).

146. Okonogi, T. M. & McConnell, H. M. Contrast inversion in the epifluorescence of cholesterol-

phospholipid monolayers. Biophysical Journal 86, 880-890 (2004).

147. Sprong, H., van der Sluijs, P. & van Meer, G. How proteins move lipids and lipids move

proteins. Nat Rev Mol Cell Biol 2, 504-13 (2001).

148. Hagen, J. P. & McConnell, H. M. Liquid-liquid immiscibility in lipid monolayers.

Biochimica Et Biophysica Acta-Biomembranes 1329, 7-11 (1997).

149. Grimalt, J. O., Fernandez, P., Bayona, J. M. & Albaiges, J. Assessment of Fecal Sterols and

Ketones as Indicators of Urban Sewage Inputs to Coastal Waters. Environmental Science &

Technology 24, 357-363 (1990).

150. Veatch, S. L. & Keller, S. L. Organization in lipid membranes containing cholesterol. Phys

Rev Lett 89, 268101 (2002).

151. Dietrich, C. et al. Lipid Rafts Reconstituted in Model Membranes. Biophysical Journal 80,

1417 (2001).

152. Baumgart, T., Hess, S. T. & Webb, W. W. Imaging coexisting fluid domains in biomembrane

models coupling curvature and line tension. Nature 425, 821-4 (2003).

153. Folch, J., Lees, M. & Stanley, G. H. S. A SIMPLE METHOD FOR THE ISOLATION AND

PURIFICATION OF TOTAL LIPIDES FROM ANIMAL TISSUES. Journal of Biological

Chemistry 226, 497 (1957).

154. Bacia, K., Scherfeld, D., Kahya, N. & Schwille, P. Fluorescence Correlation Spectroscopy

Relates Rafts in Model and Native Membranes. Biophysical Journal 87, 1034 (2004).

145

155. Korlach, J., Baumgart, T., Webb, W. W. & Feigenson, G. W. Detection of motional

heterogeneities in lipid bilayer membranes by dual probe fluorescence correlation

spectroscopy. Biochimica et Biophysica Acta (BBA) - Biomembranes 1668, 158 (2005).

156. Korlach, J., Schwille, P., Webb, W. W. & Feigenson, G. W. Characterization of lipid bilayer

phases by confocal microscopy and fluorescence correlation spectroscopy. Proceedings of the

National Academy of Sciences 96, 8461 (1999).

157. Lingwood, D. & Simons, K. Detergent resistance as a tool in membrane research. Nat Protoc

2, 2159-65 (2007).

158. Ayuyan, A. G. & Cohen, F. S. Raft composition at physiological temperature and pH in the

absence of detergents. Biophysical Journal, biophysj.107.118596 (2007).

159. Keller, S. L. & McConnell, H. M. Stripe Phases in Lipid Monolayers near a Miscibility

Critical Point. Physical Review Letters 82, 1602 (1999).

160. Hao, M., Mukherjee, S. & Maxfield, F. R. Cholesterol depletion induces large scale domain

segregation in living cell membranes. Proceedings of the National Academy of Sciences 98,

13072 (2001).

161. Nishimura, S. Y., Vrljic, M., Klein, L. O., McConnell, H. M. & Moerner, W. E. Cholesterol

Depletion Induces Solid-like Regions in the Plasma Membrane. Biophysical Journal 90, 927

(2006).

162. Lichtenberg, D., Goni, F. M. & Heerklotz, H. Detergent-resistant membranes should not be

identified with membrane rafts. Trends in Biochemical Sciences 30, 430 (2005).

163. Meder, D., Moreno, M. J., Verkade, P., Vaz, W. L. & Simons, K. Phase coexistence and

connectivity in the apical membrane of polarized epithelial cells. Proc Natl Acad Sci U S A

103, 329-34 (2006).

164. Swamy, M. J. et al. Coexisting Domains in the Plasma Membranes of Live Cells

Characterized by Spin-Label ESR Spectroscopy. Biophysical Journal 90, 4452 (2006).

146

165. Ramstedt, B. & Slotte, J. P. Interaction of Cholesterol with Sphingomyelins and Acyl-Chain-

Matched Phosphatidylcholines: A Comparative Study of the Effect of the Chain Length.

Biophysical Journal 76, 908 (1999).

166. Slotte, J. P. Sphingomyelin-cholesterol interactions in biological and model membranes.

Chemistry and Physics of Lipids 102, 13 (1999).

167. O'Brien, J. S. & Rouser, G. The fatty acid composition of brain sphingolipids:

sphingomyelin, ceramide, cerebroside, and cerebroside sulfate. Journal of Lipid Research 5,

339 (1964).

168. Pralle, A., Keller, P., Florin, E. L., Simons, K. & Horber, J. K. H. Sphingolipid-Cholesterol

Rafts Diffuse as Small Entities in the Plasma Membrane of Mammalian Cells. The Journal

of Cell Biology 148, 997 (2000).

169. Langlet, C., Bernard, A.-M., Drevot, P. & He, H.-T. Membrane rafts and signaling by the

multichain immune recognition receptors. Current Opinion in Immunology 12, 250 (2000).

147