Lipid Bilayers As Surface Functionalizations for Planar and Nanoparticle Biosensors

Lipid Bilayers as Surface Functionalizations for Planar and Nanoparticle Biosensors

Shell Y. Ip

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Chemistry University of Toronto

Lipid Bilayers as Surface Functionalizations for Planar and Nanoparticle Biosensors

Shell Y. Ip

Doctor of Philosophy

Department of Chemistry University of Toronto

2010 Abstract

Many biological processes, pathogens, and pharmaceuticals act upon, cellular membranes. Accordingly, cell membrane mimics are attractive targets for biosensing, with research, pathology, and pharmacology applications. Lipid bilayers represent a versatile sensor functionalization platform providing antifouling properties, and many receptor integration options, uniquely including transmembrane proteins. Bilayer-coated sensors enable the kinetic characterization of membrane/analyte interactions. Addressed theoretically and experimentally is the self-assembly of model membranes on plasmonic sensors. Two categories of plasmonic sensors are studied in two parts. Part I aims to deposit raft-forming bilayers on planar nanoaperture arrays suitable for multiplexing and device integration. By vesicle fusion, planar bilayers are self-assembled on thiol-acid modified flame-annealed gold without the need for specific lipid head-group requirements. Identification of coexisting lipid phases is accomplished by AFM imaging and force spectroscopy mapping. These methods are successfully extended to metallic, plasmon-active nanohole arrays, nanoslit arrays and annular aperture arrays, with coexisting phases observed among the holes. Vis-NIR transmission spectra of the arrays are measured before and after deposition, indicating bilayer detection. Finally, the extraction of membrane proteins from cell cultures and incorporation into model supported bilayers is demonstrated. These natural membrane proteins potentially act as lipid-bound surface receptors. Part II aims to encapsulate in model lipid bilayers, metallic nanoparticles, which are used as probes in surface enhanced Raman spectroscopy. Three strategies of encapsulating particles, and incorporating Raman-active dyes are demonstrated, each using a different dye: malachite green, rhodamine-PE, and Tryptophan. Dye incorporation is verified by SERS and the bilayer is ii visualized and measured by TEM, with support from DLS and UV-Vis spectroscopy. In both parts, lipid-coated sensors are successfully fabricated and characterized. These results represent important and novel solutions to the functionalization of plasmonic surfaces with biologically relevant cell membrane mimics.

iii

Dedicated to my Mother and Father, Mimi and Ken Ip

Acknowledgments

It is with profound pleasure, humility, and sincerity that I thank those whose help and support have made this endeavour possible. First and foremost, I would like to thank my supervisor, Prof. Gilbert C. Walker for his mentorship, patience, and unvarying support. Though he has undoubtedly provided invaluable scholarly direction throughout the years, it is his good nature, and social acumen that have been most appreciated. I also sincerely thank the members of my committee, Prof. Ulrich J. Krull, and Prof. J. Stewart Aitchison, who, with their enthusiasm, wisdom and expertise, have both contributed significantly to our discussions. Also, it has been my privilege to work with friends and collaborators Matthew J. Kofke, and Prof. David H. Waldeck. Mr. Kofke has skillfully, and reliably provided many FIB samples in timely fashion. Our visits and cooperative efforts in experiments have been grueling at times, but inspired and stimulating. I am grateful to my friend and labmate, Christina MacLaughlin, for her skill and marathon-honed determination, but most of all her spirit, which have made the nanoparticle experiments possible. I also sincerely thank Dr. Nikhil Gunari, not only for his contributions to the TEM imaging of the nanoparticles, but also for years of mentorship, and encouragement. Also to my friend and colleague James K. Li, I extend my gratitude, for sharing with me, his coding expertise and force map analysis code, which forms the framework for the code used within.

Several others have made instrumental contributions to this work including my friend and colleague Ruby May A. Sullan, who has been an invaluable partner in many helpful and thoughtful discussions of lipid bilayers. Always willing to share her knowledge and experience, she has had a tremendous influence on the lipid techniques employed. I extend my gratitude also to Natalie Tam in from Gang Zheng’s group at the University Health Network, for the preparation and culturing of cells used in these experiments, not to mention her friendly, generous support and expert advice on cells. I also gratefully acknowledge the helpful discussions of surface plasmons and near-field optics I’ve shared with my friend and colleague Daniel Lamont.

For their mentorship and guidance I thank the Post-Docs in our group past and present; specifically, Dr. Weiqing Shi, Dr. Zhara Fakhraai, and Dr. Shan Zou. In particular, I am grateful to Dr. Zou, now with the NRC, not only for her keen insight and pragmatic experimental v sagacity, but also for her support, encouragement, and continued friendship. As a mentor and friend I also thank Dr. Emanuel Istrate of the Institute of Optical Sciences.

I also thank the members of the Walker group past and present, who I’ve had the privilege to call my friends and colleagues. Additionally I extend my warm gratitude to Mandy Koroniak and Anna Liza Villavelez for their resourcefulness and friendly handling of all things group and department related respectively.

For their use of facilities and equipment, I thank the administrators and staff at University of Pittsburgh Nanoscale Fabrication and Characterization Facility, the Nanofabrication facility at the University of Western Ontario, the Cornel Nanofabrication Facility, The Department of Chemistry’s Center for Nanoscale Imaging, and the labs of Eugenia Kumacheva and Gregory Scholes.

Financial support has come from a variety of sources over the years including the Natural Sciences and Research Council of Canada, the Ontario Graduate Scholarship Program, the Department of Chemistry, and the University of Toronto.

For their unwavering love, support, and patience, with my love and respect I thank my family: My parents Mimi and Ken; My sister June and her partner Ferhan

Last but not least, I genuinely thank Lisa and her family for their support; and especially Lisa for the innumerable ways she has buoyed me, and for her tremendous patience, strength, and perseverance.

Table of Contents

ACKNOWLEDGMENTS ...... V

TABLE OF CONTENTS ...... VII

LIST OF TABLES...... XI

LIST OF FIGURES ...... XII

LIST OF ABBREVIATIONS...... XXIII

PART I. LIPID BILAYER FUNCTIONALIZATION OF PLANAR NANOAPERTURE ARRAYS ...... 1

CHAPTER 1: INTRODUCTION ...... 2

1.1 GENERAL INTRODUCTION ...... 2

1.2 PROJECT DESCRIPTION...... 7

1.3 FUTURE APPLICATIONS...... 10

1.4 CONCLUSIONS ...... 13

1.5 REFERENCES...... 13

CHAPTER 2: BACKGROUND ...... 15

2.1 INTRODUCTION ...... 15

2.2 BRIEF REVIEW OF LITERATURE...... 15 2.2.1 Surface plasmons and SPR sensing...... 16 2.2.2 Surface plasmons on nanophotonic structures...... 20 2.2.3 Exploiting cellular interactions...... 26 2.2.4 Phospholipid chemistry ...... 28 2.2.5 Supported lipid bilayers and methods of preparation...... 33 2.2.6 Summary of Background ...... 36

2.3 GENERAL METHODS...... 38 2.3.1 Focused Ion Beam Milling...... 38 2.3.2 Spectroscopy...... 39 2.3.3 Vesicle fusion ...... 40 2.3.4 Atomic Force Microscopy of Lipid Bilayers...... 42 2.3.5 Ultraflat gold via hydrogen flame annealing...... 43 2.3.6 Self assembled monolayers...... 44

2.4 EARLY RESULTS AND CHALLENGES...... 45 2.4.1 Introduction...... 45 vii

2.4.2 Fabrication methods...... 46 2.4.3 Nanoaperture arrays and optical properties...... 51 2.4.4 Artificial Bilayers on gold ...... 53

2.5 CONCLUSIONS ...... 59

2.6 REFERENCES...... 63

CHAPTER 3: PHASESEGREGATION OF MODEL LIPID BILAYERS ON AU BY AFM IMAGING AND FORCE MAPPING...... 73

3.1 INTRODUCTION ...... 74

3.2 MATERIALS AND METHODS...... 79

3.3 RESULTS AND DISCUSSION ...... 83 3.3.1 Vesicle fusion requires hydrophilic surface: ...... 84 3.3.2 AFM imaging and lipid phase morphology on gold: ...... 85 3.3.3 AFM force map of DEC221 on gold:...... 86 3.3.4 Comparison with micasupported bilayers:...... 91 3.3.5 The effect of the substrate stiffness on apparent modulus: ...... 94 3.3.6 Thermal history of lipids:...... 95 3.3.7 Surface charge density of substrates ...... 97

3.4 CONCLUSIONS ...... 100

3.5 ACKNOWLEDGEMENTS ...... 102

3.6 REFERENCES...... 102

CHAPTER 4: LIPID BILAYERS ON NANOAPERTURE ARRAYS...... 112

4.1 INTRODUCTION ...... 112

4.2 MATERIALS AND METHODS...... 117

4.3 RESULTS AND DISCUSSION...... 120 4.3.1 Unpatterned Gold...... 120 4.3.2 Nanohole arrays...... 124 4.3.3 Nanoslit array ...... 133 4.3.4 Annular Aperture Arrays ...... 136 4.3.5 Transmission spectra...... 139

4.4 CONCLUSIONS AND PERSPECTIVES...... 142

4.5 ACKNOWLEDGEMENTS ...... 145

4.6 REFERENCES...... 145

viii

CHAPTER 5: INCORPORATION OF MEMBRANE PROTEINS EXTRACTED FROM CELL CULTURES INTO ARTIFICIAL BILAYERS...... 150

5.1 INTRODUCTION ...... 150

5.2 MATERIALS AND METHODS...... 152

5.3 RESULTS AND DISCUSSION...... 155

5.4 CONCLUSIONS AND FUTURE PERSPECTIVES ...... 163

5.5 ACKNOWLEDGEMENTS: ...... 165

5.6 REFERENCES...... 165

CHAPTER 6: SUMMARY OF PART I AND FUTURE PERSPECTIVES...... 168

6.1 THE IMPORTANCE OF LIPID BILAYERS IN BIOSENSORS ...... 168

6.2 EXISTING METHODS OF LIPID SURFACE FUNCTIONALIZATION ...... 169

6.3 SUMMARY OF RESULTS...... 171 6.3.1 Phase segregation of lipids on gold...... 171 6.3.2 Model lipids on nanohole, nanoslit and annular aperture arrays...... 174 6.3.3 Membrane protein incorporation into model lipid bilayers...... 176

6.4 FUTURE WORK...... 177

6.5 FINAL REMARKS...... 179

6.6 REFERENCES...... 181

PART II. LIPID BILAYER FUNCTIONALIZATION OF PLASMONIC NANOPARTICLES...... 184

CHAPTER 7: MODEL LIPID MEMBRANE ENCAPSULATION OF SERS NANOPARTICLES...... 185

7.1 INTRODUCTION ...... 185

7.2 MATERIALS AND METHODS...... 188

7.3 CHARACTERIZATION OF LIPID‐ENCAPSULATED SERS PARTICLES...... 190

7.4 STABILITY OF LIPID ENCAPSULATED SERS PARTICLES...... 199

7.5 CONCLUSIONS ...... 203

7.6 ACKNOWLEDGEMENTS ...... 206

7.7 REFERENCES...... 206

CHAPTER 8: REVIEW OF NANOPARTICLE OPTICAL PROBES: BIOLOGICAL LABELLING AND TOXICITY 211

8.1 GENERAL INTRODUCTION ...... 211

8.2 BACKGROUND ...... 212 8.2.1 Surface functionalization ...... 213 8.2.2 Enhanced optical properties ...... 215

8.3 PLASMONIC NOBLE METAL NANOPARTICLES...... 216 8.3.1 Introduction: ...... 216 8.3.2 Localized Surface Plasmon Resonance:...... 217 8.3.3 Tuning Metal Nanoparticle Surface Plasmon Properties:...... 218 8.3.4 Surface Enhanced Raman Spectroscopy (SERS): ...... 222 8.3.5 In vitro toxicity of metal nanoparticles:...... 228

8.4 FLUORESCENT SEMICONDUCTOR QUANTUM DOTS ...... 231 8.4.1 Background...... 232 8.4.2 Surface functionalization ...... 239 8.4.3 In vitro assays ...... 245 8.4.4 In vivo assays...... 246 8.4.5 Toxicity...... 250

8.5 CONCLUSIONS AND FUTURE PERSPECTIVES:...... 258

8.6 ACKNOWLEDGEMENTS ...... 260

8.7 REFERENCES...... 260

CHAPTER 9: SUMMARY OF PART II AND FUTURE PERSPECTIVES ...... 275

9.1 LIPOSOMES AND NANOPARTICLE FUNCTIONALIZATION ...... 275

9.2 SUMMARY OF RESULTS...... 277

9.3 FUTURE WORK...... 278

9.4 FINAL REMARKS ...... 280

List of Tables

Table 7-1: Assignment of observed SERS bands of MGITC based on ref 28. “Band” column is the observed wave-number of a peak. (*) indicates that correlation of band position with ref 28 is > 5 cm-1, and that the assignment is the closest match. “Chemical group” and “Mode” columns indicate the chemical group and type of vibrational mode assigned by ref 28. Multiple assignments to a band are separated by semicolons. Parentheses indicate location of the chemical group within the molecule...... 191

Table 7-2: Assignment of observed SERS bands of Rho-PE based on ref 38 (Rhodamine B) and based on ref 39 (Rhodamine 6G). “Band” column is the observed wave-number of a peak. (*) indicates that correlation of band position with ref 39 is > 5 cm-1, and that the assignment is the closest match. (†) indicates correlation with ref 38 only. “Chemical group” and “Mode” columns indicate the chemical group and type of vibrational mode assigned by ref 38 and 39. Multiple assignments to a band are separated by semicolons. Parentheses indicate location of the chemical group within the molecule...... 197

Table 7-3: Assignment of observed SERS bands based on ref 43. “Band” column is the observed wave-number of a peak. (*) indicates that correlation of band position with ref 43 is > 5 cm-1, and that the assignment is the closest match. “Chemical group” and “Mode” columns indicate the chemical group and type of vibrational mode assigned by ref 43. Multiple assignments to a band are separated by semicolons. Parentheses indicate location of the chemical group within the molecule...... 199

Table 7-4: Summary of colloidal stability test: Lipid-coated particles, lipid vesicles, and stock particles are subjected to high salt and low pH conditions. Descriptions of the suspensions are summarized...... 202

List of Figures

Figure 1-1 Diagram of a supported lipid bilayer on a NAA surface and the capture of a CD4+ T- cell from solution via an immunological synapse...... 10

Figure 1-2: Illustration of a super-array of nanoaperture arrays. Optical transmission is collected by a CCD array (green) placed behind the super-array. Signals from individual NAAs are separated spatially on the image produced by the CCD. The linear optical path makes high throughout parallel sensing possible...... 11

Figure 2-1: Illustration of surface plasmon dispersion relative to light line. Solid line is the surface plasmon dispersion for a smooth metal/air interface. Dashed line (- -) is the light line in air. Dash-dot line (-.) is the light line in a medium with refractive index n1...... 18

Figure 2-2: Schematic illustration of di-bock co-polymer templating method...... 47

Figure 2-3: AFM scan of spin-casted PS-PMMA diblock copolymer thin film that spontaneously phase segregates into cylindrical domains oriented in perpendicular to the surface. Cross section through a defect shows film thickness...... 48

Figure 2-4: PS-PMMA diblock copolymer film after oxygen plasma treatment. Self-assembled pattern of polymer is as deep as the defect in the film, which indicated the PMMA matrix has been etched, leaving an array of PS posts...... 49

Figure 2-5: Schematic illustration of electron beam lithography of NAA. Sequence of steps: top row left  right, then bottom row left  right ...... 50

Figure 2-6 (A): AFM image of residual PMMA on nanohole arrays made by e-beam lithography. Thickness of layer from cross section: ~200 nm. (B): Transmission spectra of nano-hole array in solvents with different dielectric constants. Base lines for increasing dielectric constant (dielectric constant stated in parentheses) are shifted up for clairity. Stationary peak at 775 nm is likely due to PMMA layer. Peaks due to solvent are red-shifted relative to arrays without PMMA. Peaks shift due to change in dielectric constant despite thickness of PMMA layer...... 51

xii

Figure 2-7: AFM topography (A) and SEM of a cross section (B) of a nanoaperture array. Hole size is 200 nm x 200 nm, and array period is 500 nm. Fabrication and SEM performed by Todd Simpson, University of Western Ontario, London, Canada...... 52

Figure 2-8: (A) Transmission spectra through NAA (hole size – 200 nm square, period = 600 nm) under different dielectric environments (dielectric constants noted in brackets). (B): Plot of peak position with dielectric constant of solvent. The peak position red-shifts linearly with increasing dielectric constant. The slope indicates a sensitivity of 256 nm/dielectric constant unit...... 53

Figure 2-9 AFM topography image of DPPC bilayers on a glass cover slip made by vesicle fusion. Scan size is roughly 8 µm x 8 µm. The section analysis of defects in the bilayer indicates a thickness of 5.4nm. Literature values for DPPC bilayer thickness is 4.6 nm...... 54

Figure 2-10: (A) AFM topography image of bare commercial gold substrate. RMS roughness of image is 2.38 nm, while the RMS roughness in the area beneath the white box is 2.35 nm. (B): AFM image (5 µm x 5 µm) of the gold substrate after vesicle fusion treatment. Features may be lipid patches, but the section analysis (C) is inconclusive...... 55

Figure 2-11 Hydrogen flame annealed Au surfaces (A): AFM topography shows the surface reconstruction. Grain sizes are on the order of microns in diameter. (B): 880 nm x 880 nm AFM topography image of a single grain over the region indicated by the white box in Panel A. Regular triangular features are atomic step terraces characteristic of (111) crystal surface...... 56

Figure 2-12 (A): AFM topography image (scan size = 1 µm) of flame annealed gold after treatment with 11-mercaptoundecanoic acid using the method of Wang et al.92. Features have heights < 2 nm and lateral sizes of ~ 20 – 30 nm. Surface contact angle: ~ 30 degrees (B): AFM AC-mode images of vesicle fusion on surface similar to (A) (scan size = 1.5 µm). Patches have heights of ~ 5.5 nm relative to substrate, and diameters of ~30 nm...... 57

Figure 2-13: Fluid phase AFM AC-mode image of nearly defect free DPPC bilayer on COOH- alkanethiol modified Au. Alkane thiol monolayer produced by a modified method of ref92. Scan size: 850 nm x 850 nm. Surface contact angle after COOH SAM formation was < 15 degrees.. 58

xiii

Figure 2-14: Select force indentation curves (green) of DPPC bilayer on 11-mercaptoundecanoic acid modified ultraflat Au. Red curve is Sneddon model fit to indentation. Elastic modulus obtained as fitting parameter was about 3.9 MPa for both cases...... 59

Figure 3-1: Typical AFM force indentation curve of the ternary lipid bilayer on gold (Top) featuring a breakthrough event characteristic of these bilayers. Features of the indentation curve are labeled A through D. The four-panel diagram (not to scale) represents the interaction between the AFM tip and the lipid bilayer during the indentation process. Letters A-D refer to labeled features on the example indentation curve. The bilayer constituents (DOPC, egg sphingomyelin, and cholesterol) are drawn to represent the phase-segregation of the DOPC from the egg sphingomyelin into ld and lo phases respectively. The bilayer is supported by the mercaptoundecanoic acid (MUDA) modified gold substrate through an assumed hydration layer...... 77

Figure 3-2: AC-mode AFM height (A) and phase (B) images of MUDA-modified, flame- annealed gold. Lines intersecting in triangular patterns are the terraces between (111) crystal planes of the Au. (C-H): AFM imaging and force mapping of DEC221 (DOPC/egg-SM/Chol mixed in 2:2:1 molar ratio) bilayers on MUDA-modified, flame annealed Au at room temperature. AC-mode AFM height (C) and phase (D) images reveal phase segregation into liquid ordered (lo) and disordered (ld) phases. 64x48 force maps of the same area are processed to produce Young’s modulus and breakthrough force maps (E and F, respectively). Contrast between lo and ld phases are clearer in the breakthrough force map, and follow the features on the phase image (D). G and H are the histograms of the data mapped in E and F respectively. The histogram of breakthrough forces (H) has a largely bimodal distribution attributed to the lo phase

(peak at ~ 3nN), and the ld phase (peak at ~ 2.1nN)...... 88

Figure 3-3: Breakthrough force contrast (see Figure 3-2 D and F) used to separate data from the

Young’s modulus map (Figure 3-2 C and E) into contributions from the lo phase (A and C) and the ld phase (B and D). Modulus maps A) and B) clearly corresponds to features of the lo phase and the ld phase respectively, indicating the data sorting is reasonable. C and D are histograms of data seen in modulus maps A and B respectively. Log-normal fits to the histograms (black lines in C and D) are used to determine the peak value: The average modulus of the lo phase is 100 ± 2

MPa while that of the ld phase is 59.8 ± 0.9 MPa...... 90

xiv

Figure 3-4: DEC221 on mica via vesicle fusion at 50˚C. Contact mode AFM height map (A) clearly shows two lipid phases lo and ld (higher and lower respectively) as well as defects exposing bare mica. Breakthrough force map (B). White areas in (B) are excluded data points detected by software as non-breakthrough events. Features in (B) correspond well with features in (A). Breakthrough force histogram shows characteristic bimodal distribution. Lower force peak corresponds to ld phase while higher force peak belongs to the lo phase...... 92

Figure 3-5: As with the lipid on gold, breakthrough force data (Figure 3-4 C) used to sort modulus data into contributions from lo and ld phases. A and C are the modulus map and associated histogram for the lo phase. B and D are the modulus map and histogram for the ld phase. Average moduli acquired from log normal fits to histogram are 164 ± 2 MPa and 67 ± 1 40 MPa for the lo and ld phase. These results agree well with literature values ...... 93

Consistent with the analysis discussed in the text, lower apparent Poisson’s ratio of the ld phase on MUDA/gold than on mica could lead to the same modulus value of actual bilayer, as indicated by the dotted grey line...... 99

Figure 4-1 Schematic representation of AFM force indentation (top), and example force curves on nanoaperture arrays (botton). From left to right: ‘Breakthrough’ event indicative of lipid bilayers, ‘Hard Contact’ event were no breakthough is detected, and ‘Other’ event that is neither a Breakthrough nor Hard Contact...... 116

Figure 4-2 AFM height (A) and phase (B) of unpatterned area containing several gold grains, defects and phases. (C):Force classification map over the same area as A; Grey is ‘Breakthrough’, black is ‘Hard Contact’ and white is ‘Other’. (D): Map of second force value (ex: breakthough force for ‘breakthrough events’)...... 123

xv indentations over the gold result in breakthrough events (Grey) while hard-contact events (Black) correspond to defects detectable in (B). Indentation over holes produces “Other” events. Breakthrough force map (D) cannot resolve phase segregation. Likewise, breakthrough force histogram (E) shows a broad, single-mode distribution of forces. Two modes in indentation depths (F) represent indentation over holes and over the support...... 126

Figure 4-4: Modulus maps and associated histograms of nanohole array separated by depth of indentation. Shallow indentations (A,C) indicate indentation over supported lipid; fitted peak at 84 MPa. Deep indentations (B,D) indicate indentations over apertures; fitted peak at 12 MPa...... 128

Figure 4-5: Lipid phase segregation on nanohole arrays. AFM topography (A) and phase (B) images of nanohole array over selected area with one grain and fewer grain boundaries. Height difference between domains is resolvable in (A), while some boundaries of domains are visible in (B). Modulus map (C) and...... 130

Figure 4-6: Modulus contributions from lo and ld phases separated via breakthrough forces. Left side: Modulus map and histogram of force curves having breakthrough forces greater than 1.75 nN, corresponding to lo domain. Histogram peak at 64 MPa from fit. Right side: Modulus map and histogram on force curves with breakthrough forces ≤ 1.75 nN, corresponding to ld phase. Histogram peak at 50 MPa from fit...... 132

Figure 4-7 DEC221 bilayers on nanoslit arrays. Regular array of slits is discernable by AFM topography and phase images (A and B respectively). Lipid domains are obscured by large variations in topography from pattern and grain boundaries, but defects in the bilayer appear as white features in the phase image. Force classification map (C) shows most indentations over the gold result in breakthrough events (Grey) while hard-contact events (Black) correspond to defects. Indentation over slits and most grain boundaries produces “Other” events. Breakthrough force map (D) cannot resolve phase segregation. Likewise, breakthrough force histogram (E) shows a very broad, single-mode distribution of forces. Two modes in indentation depths (F) represent indentation over holes and over the support...... 134

Figure 4-8: Modulus maps and associated histograms of nanoslit array separated by depth of indentation. Shallow indentations (A,C) indicate indentation over supported lipid; fitted peak at

xvi

29 MPa. Deep indentations (B,D) indicate indentations over apertures; fitted peak at 7.6 MPa...... 135

Figure 4-9: DEC221 bilayers on annular aperture arrays. Regular square array of annuli is discernable by AFM topography and phase (A and B respectively). Lipid phases are likely obscured by large variations in topography, but hole depths suggest lipids are spanning many holes. Force classification map (C) shows most indentations over the gold result in breakthrough events (Grey) while hard-contact events (Black) correspond to defects. Indentation over grain boundaries produces “Other” events. Breakthrough force map (D) cannot resolve phase segregation. Likewise, breakthrough force histogram (E) shows a broad, single-mode distribution of forces. Two modes in indentation depths (F) represent indentation over holes and over the support...... 137

Figure 4-10: Modulus maps and associated histograms of annular aperture array separated by depth of indentation. Shallow indentations (A,C) indicate indentation over supported lipid; fitted peak at 20 MPa. Deep indentations (B,D) indicate indentations over apertures; fitted peak at 5 MPa...... 138

Figure 4-11: Visible (left side) and NIR (Right side) transmission spectra through the same nanohole array (A,B), nanoslit array (C,D), and annular aperture array (E,F) used for force mapping before (grey line) and after (black line) bilayer formation (confirmed by force mapping). For nanohole arrays (A,B), peak position is determined by fitting to Gaussian (markers). The Vis peak (A) red-shifts 4 nm, while the NIR peak (B) red-shifts 6 nm. Broad- band transmission spectra shown in inset...... 141

Figure 5-1: Normalized fluorescence excitation (grey) and emission (black) spectra of GFP- EGFR containing cell extracts ...... 155

Figure 5-2: AFM magnetic tapping mode topography of pure DEC221 bilayer (A), and DEC + Membrane proteins (B, C). A and B are 5 µm x 5 µm scans and panel C is a 1 µm x 1 µm scan.

Raft-like lo domains appear as light grey irregular, but rounded structures in a (dark grey) ld matrix in A and B. Proteins appear as small white spots in B and C...... 156

Figure 5-3: From left to right, consecutive AFM topography scans over a 10 µm x 10 µm area tracking the migration of a protein fiber in the lower left portion of the image...... 157 xvii

Figure 5-4: Comparison of force maps over pure DEC221 and DEC-MP bilayers. A, B and C are topography, classification, and breakthrough-force maps respectively of pure DEC221 bilayers deposited in PBS but imaged in pure water. Breakthrough events were not detected over the lo domains despite using a high force setpoint, likely due to the presence of residual ions from buffer. D, E and F are topography, classification, and breakthrough-force maps, respectively, of

DEC-MP bilayers in PBS. Many double breakthrough events are detected on ld phase due to tip contaimination...... 159

Figure 5-5: Young’s modulus maps and histograms of DEC-MP bilayers separated by classification. A and C are the double breakthrough events (ld phase), and B and D are the hard contact events (lo phase). Fits to the histograms in C (ld) and D (lo) are 29.4±0.3 MPa and 40.3±0.7 MPa respectively...... 161

Figure 5-6: Young’s modulus maps and histograms of DEC221 bilayers separated by classification. A and C are the breakthrough events (ld phase), and B and D are the hard contact events (mostly lo phase). Fits to the histograms in C (ld) and D (mostly lo) are 29.0±0.5 MPa and 31.5±0.3 MPa respectively...... 162

Figure 7-1: A) Schematic illustration of three methods of incorporating dye molecules to produce lipid-encapsulated gold SERS particles. The solid circle represents the gold nanoparticle, the rings represent lipid bilayer vesicles, and the hexagons represent small dye molecules. B) Chemical stuctures of the lipid mixture used to encapsulate the nanoparticles...... 188

Figure 7-2: Characterization of Malachite Green isothiocyanate (MGITC) lipid-encapsulated SERS nanoparticles produced using Method 1. A) The chemical structure of MGITC, B) TEM image of a bilayer-encapsulated nanoparticle. The ring around the particle is the lipid bilayer, which measures 4.8 nm averaged from 5 measurements of 5 isolated particles. C) DLS histograms of hydrodynamic radius for stock citrate-coated particles (hatched,grey) and MGITC- lipid-coated SERS nanoparticles (outline,black). Gaussian centroids report average Rh = 30.5 nm and 36.4 nm for stock particles and SERS particles respectively. D) LSPR absorption in UV- Vis spectrum of stock (grey) and MGITC-lipid-coated particles (black). LSPR absorption peaks at 534 nm for stock particles and 538 nm for MGITC-lipid-coated particles. E) SERS spectrum of MGITC-lipid-coated nanoparticles showing strong SERS spectrum recognizable as that of MGITC...... 192 xviii

Figure 7-3: Characterization of lipid-encapulated SERS nanoparticles with 1 mol% Rhodamine- Lissamine-DSPE (Rho-PE) produced using Method 2. A) The chemical structure of Rho-PE, B) TEM image of a bilayer-encapsulated nanoparticles. The ring around the particles is the lipid bilayer. Clustering is caused by drying of the sample on the TEM grid. The bilayer measures 5.8 nm averaged from 10 measurements of 5 clusters of particles. C) DLS histograms of hydrodynamic radius (bars) and fits to Gaussian distributions (curves) for stock citrate-coated particles (grey) and Rho-lipid-coated SERS nanoparticles (black). Gaussian centroids report average Rh = 30.5 nm and 36.8 nm for stock particles and SERS particles respectively. D) LSPR absorption in UV-Vis spectrum of stock (grey) and Rho-lipid-coated SERS particles (black). LSPR absorption peaks at 534 nm for stock particles and 536 nm for Rho-lipid-coated particles. E) Fluorescence spectra of Rho-lipid vesicle supernatant (dashed) and of Rho-lipid-coated SERS particles (solid). F) UV-Vis absorption of Rho-lipid-coated SERS particles prepared with Ca2+. Grey solid line is the abs. spectrum of stock Au particles, black line is the abs. spectrum of Rho- lipid-coated particles, and grey dashed line is the abs. spectrum of Rho-PE/DEC vesicles. G) SERS spectrum of Rho-lipid-coated nanoparticles prepared with Ca2+ showing strong SERS spectrum recognizable as that of Rhodamine...... 194

Figure 7-4: Tryptophan as a Raman active molecule in a lipid-coated SERS nanoparticle. A) Chemical structure of Tryptophan. B) UV-Vis absorption spectrum of bare gold (grey) and Trp- Lipid-coated SERS particle (black). Peak shifts form 535nm to 538.6nm. C) SERS spectrum of Trp-lipid-coated SERS nanoparticle. Significant bands are marked by a vertical line and the wave number...... 198

Figure 7-5: Stability of MGITC-lipid-coated particles and Rho-lipid-coated-particles. A) SERS spectrum of MGITC-lipid-coated particles collected on day of synthesis, 12 days, and 25 days after synthesis. B) SERS spectrum of Rho-lipid-coated-particles collected on day of synthesis, and 7 days after synthesis. For both cases, particles were stored in water at 4 °C between measurements...... 200

xix

P.K., Huang, X., El-Sayed, I.H. & El-Sayed, M.A. Noble Metals on the Nanoscale: Optical and Photothermal Properties and Some Applications in Imaging, Sensing, Biology, and Medicine. Accounts of Chemical Research 41, 1578-1586 (2008). Copyright © 2008, American Chemical Society...... 220

Figure 8-2: A) Tunability of surface plasmon resonant wavelength of 30 nm length Ag nanocube sacrificial templates after they have been treated with different volumes of HAuCl4 in a galvanic replacement reaction. B) and C) Location of resonant wavelength and components of extinction coefficient for Au nanocages of dimentions shown inset in each figure72. Reproduced from Skrabalak, S.E. et al. Gold Nanocages for Biomedical Applications. Adv. Mater. Weinheim 19, 3177-3184 (2007). Copyright © 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim. Reprinted with permission...... 221

Figure 8-3: SERS labelled cells show distinct spectrum, which is the same as the pure tag in solution. Control cells incubated with unlabelled nanoparticles, or NPs with nonspecific antibody labels do not show spectral features.38 Reproduced with permission from Qian, X. et al. In vivo tumor targeting and spectroscopic detection with surface-enhanced Raman nanoparticle tags. Nat Biotech 26, 83-90 (2008). Copyright © 2008, Nature Publishing Group. Reprinted with permission...... 227

Figure 8-4: Comparison of electronic energy levels of bulk semiconductors having a continua of valence and conduction states (left), and semiconductor QDs having discrete conduction and valence sates (right).122 Reproduced with permission from Brus, L. Electronic wave functions in semiconductor clusters: experiment and theory. The Journal of Physical Chemistry 90, 2555- 2560 (1986). Copyright © 1986, American Chemical Society. Reprinted with permission...... 234

Figure 8-5: Illustration of quantum size effect. The absorption peak appears at lower wavelengths for smaller particles.6 Reproduced with permission from Murray, C.B., Norris, D.J. & Bawendi, M.G. Synthesis and characterization of nearly monodisperse CdE (E = sulfur, selenium, tellurium) semiconductor nanocrystallites. Journal of the American Chemical Society 115, 8706- 8715 (1993). Copyright © 1993, American Chemical Society...... 235

Figure 8-6: Illustration of functionalized quantum dot used by Chan et al., featuring hydrophilic surface groups, and bioconjugated protein.14 Reproduced with permission from Chan, W. & Nie,

S. Quantum dot bioconjugates for ultrasensitive nonisotopic detection. SCIENCE 281, 2016- 2018 (1998). Copyright © 1998 The American Association for the Advancement of Science. Reprinted with permission...... 238

Figure 8-7: An adaptable method for attaching different functional groups to a quantum dot through dithiol groups. Carbon disulfide is mixed with amine terminated functional group to form a dithiocarbamate group. (2a-2i) are examples of various dithiocarbamate groups made with this synthesis. The dithiol group attaches to the particle, and the functional groups interface with the surroundings.28 Reproduced with permission from Dubois, F., Mahler, B., Dubertret, B., Doris, E. & Mioskowski, C. A versatile strategy for quantum dot ligand exchange. Journal of the American Chemical Society 129, 482-483 (2007). Copyright © 2007, American Chemical Society...... 242

Figure 8-8: Versatile amphiphilic copolymer coatings for QDs. Different amine-terminated functional groups can be attached to the polymer backbone via reaction between amine and maleic anhydride group. The resulting polymer wraps around the QD with the hydrophobic side chains intercalating with the hydrophobic coating, and the hydrophilic side chains interfacing with the solvent.30 Reproduced with permission from Yu, W.W. et al. Forming Biocompatible and Nonaggregated Nanocrystals in Water Using Amphiphilic Polymers. Journal of the American Chemical Society 129, 2871-2879 (2007). Copyright © 2009, American Chemical Society...... 244

Figure 8-9: Demonstration of labelling and tracking membrane proteins by two QDs with distinct fluorescence emissions and different targeting moieties.25 Reproduced with permission from Roullier, V. et al. High-Affinity Labeling and Tracking of Individual Histidine-Tagged Proteins in Live Cells Using Ni2+ Tris-nitrilotriacetic Acid Quantum Dot Conjugates. Nano Letters 9, 1228-1234 (2009). Copyright © 2009, American Chemical Society...... 246

Figure 8-10: Multiphoton fluorescence image of deep skeletal muscle vasculature highlighted by PEG-QDs suspended in blood.61 Reproduced with permission from Bateman, R.M., Hodgson, K.C., Kohli, K., Knight, D. & Walley, K.R. Endotoxemia increases the clearance of mPEGylated 5000-MW quantum dots as revealed by multiphoton microvascular imaging. J. Biomed. Opt. 12, 064005-8 (2007). Copyright © 2007 Society of Photo Optical Instrumentation Engineers...... 249

xxi

Figure 8-11: Comparison between cell death rates over a series of concentrations of CdSe QDs that were unfunctionalized (black bars), or had 750 Mw or 6000 Mw PEG coatings (dark grey and light grey bars respectively).137 Reproduced from Chang, E., Thekkek, N., Yu, W.W., Colvin, V.L. & Drezek, R. Evaluation of quantum dot cytotoxicity based on intracellular uptake. Small 2, 1412-1417 (2006). Copyright © 2006 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim. Reprinted with permission...... 253

Figure 8-12: Nucleation of amyloid fibrils of ß2m in the presence of various nanoparticles monitored by onset of thioflavin-T fluorescence. Blue: 70nm hydrophilic polymer nanosphere; Red: 70nm hydrophobic polymer nanosphere; Cyan: 200nm hydrophilic polymer nanosphere; Pink: 70nm hydrophobic polymer nanosphere; Orange: 16nm hydrophilic QD; Green: 6nm diameter multi-walled carbon nanotube; Yellow: 16nm cerium oxide nanoparticle; Black: ß2m without particles.139 Reproduced with permission from Linse, S. et al. Nucleation of protein fibrillation by nanoparticles. Proceedings of the National Academy of Sciences 104, 8691-8696 (2007). Copyright © 2007 by The National Academy of Sciences of the USA...... 255

xxii

List of Abbreviations

AFM – Atomic Force Microscopy

Chol. – Cholesterol

DLS – Dynamic Light Scattering

DOPC – Diolioylphosphatidylcholine

ESM – Egg Sphingomyelin

FIB – Focused Ion Beam Milling

IR – Infrared

lo – Liquid Ordered (lipid domain)

ld – Liquid Disordered (lipid domain)

LSPR – Localized Surface Plasmon Resonance

SEM – Scanning Electron Microscopy

SERS – Surface Enhanced Raman Scattering

SPP – Surface Plasmon Polariton

SPP-BW – Surface Plasmon Polariton Bloch Wave

SPR – Surface Plasmon Resonance

Vis. – Visible (i.e. the visible portion of the electromagnetic spectrum)

UV – Ultraviolet

xxiii

Part I. Lipid Bilayer Functionalization of Planar Nanoaperture Arrays

Part I of this two-part thesis is concerned with the use of lipid bilayers as functional modifications for planar biosensors, while Part II is concerned with the use of lipid bilayers on metal nanoparticle-based biosenors. In particular, Part I deals with the formation of supported lipid bilayers that mimic some important aspects of cell membranes on a promising class of optical surface-based sensors called nanoaperture arrays, with potential applications in drug screening, lab-on-chip device integration, and in fundamental biophysics. Chapter 1 provides a brief introduction of topics, and a description of the project in the context of potential applications that motivates the experimental portion of Part I. Chapter 2 provides some of the context and background information for the experimental portion of this thesis, including a brief literature review on topics ranging from surface plasmons to lipid biophysics, a description of the basic techniques used in the experimental portion of Part I, and some preliminary results that have influenced the experimental direction of this project. Chapters 3 to 5 detail the experimental work undertaken. A progression of findings is presented from the deposition and characterization of raft-forming lipid bilayers on a gold surface in Chapter 3, to the deposition of these lipids on nano-patterned gold surfaces that comprise the nanoaperture arrays in Chapter 4. Chapter 5 summarizes experiments on the incorporation of natural membrane proteins extracted from cells into supported lipid bilayers. A summary of findings for Part I and perspectives on future work are the topic of Chapter 6.

- 1 -

Chapter 1: Introduction 1.1 General Introduction

Optical tools, including the eye, have always been essential to the study of biology. However, at the length-scales relevant for proteins and DNA, limitations imposed by diffraction prevent the direct visual observation of processes that are fundamental to biology. Although, x-ray crystallography, and nuclear magnetic resonance, electron microscopy, and atomic force microscopy have become standard techniques for visualizing biology at the molecular level, the kind of structural information attained by these methods tell only part of the story. Thus various optical spectroscopies, fluorescence, electrochemical spectroscopy, and surface plasmon resonance, to name a few, are also widely employed, particularly when structural information proves difficult to obtain. For instance, fluorescence and electrochemical techniques1 are staple tools in studying the physiological functions of membrane proteins, for which structural information is sparse due to inherent physical limitations in current techniques.

Indeed, cellular and subcellular membranes, and especially membrane proteins, represent one of the deepest enigmas in modern biology: Considering that membrane proteins play important roles in processes such as ovum fertilization2, cardiac rhythm3, and the immune system4, as well as in diseases such as autoimmune disorders5, HIV/AIDS6, and cancer7, relatively little is known about membrane proteins compared to soluble proteins, in terms of function, and especially in terms of structure. Difficulties in obtaining structural information about membrane proteins is highlighted by the fact that there are over 15000 entries in the Protein Data Bank, and only 144 transmembrane protein structures have been determined, according to the membrane protein topology database8. This is in spite of the fact that membrane proteins comprise about 20-30% of most genomes9. This gulf in structural information can be attributed to the amphiphilic nature of membrane proteins, which necessitates the use of detergents in their purification. Detergents are obstacles both to crystalization and to the analysis of NMR information, which are the primary tools for structural characterization. Clearly, additional tools for the study of

- 2 -

membrane proteins can potentially have an impact both in fundamental and applied health science.

Though membrane proteins themselves are interesting targets for study, understanding their function often cannot be done by studying the proteins in isolation. For example, cells often require multiple proteins to act in concert to start the chain reaction of molecular events that comprise the cell’s response to an external stimulus. Thus, not only is there a need for tools to study the interactions between the membrane and external factors (the stimulus) but also, the interactions between multiple membrane proteins in a fluid membrane (the response).

Widely employed to study the kinetics of binding of soluble analytes to specific targets immobilized on their surfaces, surface plasmon resonance (SPR) and related techniques may represent an ideal method by which to study the interactions between membranes and the variety of targets that act on membranes such as viruses, amyloid peptides, signaling molecules, and drugs. Although planar SPR sensors utilizing the Kretschmann configuration may be considered a mature technology, with commercially available devices that offer unparalleled bulk sensitivity, there exist, some sizeable barriers to their application for the detection of physiologically relevant membrane interactions. One barrier is the need to simultaneously detect multiple targets (multiplexing), and the other is the functionalization of the surface with a relevant membrane.

The first barrier may be overcome by variants of surface plasmon-active materials such as Nanoaperture Arrays (NAAs) 10, which by virtue of their small size and simplified optical geometry, can potentially provide greater parallelization, and with that, a greater potential for device integration (such as lab-on-a-chip devices that promise to previde sample preparation or separation to in addition to detection). NAAs and other related

- 3 -

technologies will be addressed in greater detail in a later section. Some apparent advantages of NAA sensors include:

• High sensitivity for label free detection of biomolecules

• Simplified implementation for simultaneous mutiplexed sensing

• Miniaturization and integration: Their size (10 – 100 μm) makes them

suitable for high density integration, and thus bench‐top

implementations

• Modification of their surface chemistry is widely studied

Thus, NAAs represent a promising, platform technology for studying cellular interactions. But because the technology is burgeoning, there is a great deal of potential for study both fundamentally and technologically, regardless of how these predicted points of merit turn out in practice.

The second barrier, namely the functionalization of sensor surfaces with relevant model membrane systems that mimic or incorporate some features of natural membranes, is largely the focus of this thesis. Like many biological or chemical sensors, SPR and related technologies are able to detect changes in an environment within a very confined volume adjacent to the sensor surface. In order to detect analytes of interest, they must be brought into the volume of detection. Thus, in order to detect what is occurring in the vicinity of a cellular membrane, the sensor must be brought very close to the membrane. In the case of SPR, this distance is typically less than 1 micron, and decays exponentially away from the surface. Conversely, the membrane may be deposited directly on the sensor surface so that targets that interact with the membrane will enter the detection volume.

- 4 -

There remains an active interest in the preparation of model, and natural lipid membranes on planar surfaces by such techniques as vesicle fusion and variations on the Langmuir- film approach. By casting bilayers on surfaces, they become more suitable for characterization by surface analytical techniques such as Atomic force microscopy, time- of-flight secondary ion mass spectrometry, and neutron diffraction, to name a few. For these techniques, mica and glass supports are well suited. For SPR however, a noble metal surface is required. There are however, some significant challenges to producing and characterizing lipid layers on metal surfaces, and accordingly, this represents an important topic in this thesis. Many existing methods of casting lipid membranes on metal surfaces may limit the choice, and properties of lipid species used in such layers.

This is an important consideration in the context of the lipid raft hypothesis, that suggests that cellular membranes are not random distributions of lipids, proteins and polysaccharides, but instead, are capable of forming functional clusters of lipids and proteins called rafts. These rafts act to provide a means of organizing membrane components in the plane of the membrane, which plays a role in processes such as cell signaling and in mechanisms of diseases such as Alzheimer’s disease and the permeation of a cells by the HIV virus, among others. Various artificial lipid mixtures are known to reproduce some of the aspects of rafts, and are used as model lipid systems in the study of rafts. The use of such a lipid system in a biosensor would be a natural extension of the application of such model lipid systems to study their interactions with known targets such as viruses and drugs with greater relevance to natural cell membranes. A fair portion of this thesis is dedicated to the engineering challenges involved with the formation and characterization of model lipid systems on metal surfaces, and on various NAA structures.

- 5 -

In the broader context of biosensing, lipid bilayers represent a unique and versatile surface functionalization route. Traditionally surface functionalization serves 3 purposes: (1) to control the surface chemistry of the sensor and provide an interface with the analyte phase, (2) to impart selectivity towards specific analytes of interest, and (3) to prevent non-specific adsorption of molecules that may also be present but are not of interest in the detection assay. These functions are often provided by various chemical species that are covalently grafted to the sensor surface, which may be delivered by solution and deposited by a self-assembly process (such as through thiol-gold chemistry). Because of a variety of natural, synthetic, and head-group-functionalized lipids that are commercially available, mixtures of lipid species can potentially be used to provide these functionalities. They can also be delivered in solution and self-assembled by vesicle fusion. The unique aspect, in addition to mimicking cell surfaces, is the potential for these functional lipid components to remain mobile in two dimensions across the sensor surface. Additionally, these lipid layers can be more easily removed and new lipids recast, which improves the reusability of the sensor surface, though at the cost of robustness.

Ultimately, this thesis contributes to the goal of developing a sensitive analytical technique to probe the interactions between a model or natural lipid bilayer and various soluble targets. Possible applications of this technology may be found in fundamental biology, where studying the interactions between membranes and such analytes as proteins, peptides, viruses, or antibodies may be of interest. Applications may also be found in drug discovery and testing, since the membrane is a relevant therapeutic target. Also, membranes or membrane proteins may also be the targets of in vitro diagnostics, and may be adapted for use in other planar sensor systems.

- 6 -

1.2 Project description

The scope of this thesis project can be framed within the context of developing a plasmonic biosensor with a model lipid bilayer as a surface functionalization layer. This section describes some of the aspects of a hypothetical multiplexed, Nanoaperture-array- based biosensor with a lipid functional layer. The contributions of the experimental work that forms the bulk of this thesis, presented in detail in later chapters, is highlighted in this context.

Nanoaperture arrays are photonic structures consisting of an array of subwavelength apertures in an opaque metal film that exhibit structured optical transmission spectra. These apertures arrays may take the form of a 2D array of nanoholes, 1D array of nanoslits, an array of annuli, or a variety of other aperture geometries. In the case of nanohole arrays, the apertures themselves are subwavelength, and therefore do not allow propagating modes of light transmission, and although the metal film itself is thick enough to be opaque, they transmit light with unexpected efficiency. For example, in some cases a film that is 10% porous can transmit up to 20% of light at certain wavelengths that impinge on the array 11.

The optical transmission phenomenon is believed to be a result of surface plasmon polaritons (SPPs) that can be excited on the metal surface of the array. While individual holes support localized surface plasmon resonances (LSPR), arrays of these holes, when arranged so that LSPR resonances overlap and interact with one another over extended areas support surface plasmon polariton Bloch waves (SPP-BW). The transmission spectrum observed through a NAA depends largely on the wavelength of the SPP-BW modes allowed by the array. Specifically, peaks in the transmission spectrum represent the direct far-field observation of scattered SPPs. The wavelength of the transmission resonances depend on the geometry of the array, as well as material properties of the metal/dielectric interface. As a sensing element, the mechanism by which signals are

- 7 -

transduced is similar to other evanescent optical wave sensors such as Surface Plasmon Resonance (SPR) sensors, in that changes in the dielectric constant of the material interfacing with the metal cause changes in the wavelength of the SPP (or SPP-BW in the case of arrays). This is observed as a shift in the colour or wavelength position of resonant peaks in the transmission spectrum.

Because SPPs are evanescent fields, the analyte of interest must be brought within close proximity (roughly 100-500 nm 12) to the array surface in order to ensure maximum interaction between the surface plasmons and the analyte. However, the wavelength of the plasmon and the resulting transmission peak can be affected by any detectable change in the dielectric constant, which can be caused by any adsorbate without an inherent means of discrimination. Therefore, detection of specific analytes requires control of surface chemistry to promote the selective adsorption of analytes of interest. Strategies for selectively adsorbing biological analytes of interest to a surface usually involve very specific molecular interactions such as those between complementary strands of poly- nucleic acids, between antibodies and their antigens, or between biotin and avidin. Usually, one part of the pair is immobilized onto the surface, and the complementary molecule is either the analyte of interest, or acts as a label that is chemically attached to the analyte.

On metal surfaces, the immobilized surface functionalization group typically consists of a hetero-bifunctional molecule with a thiol group on one end and a reactive functional group on the other. The thiol binds to the gold surface and the other functional group is exposed at the interface where it may be conjugated to functional moieties such as single- stranded DNA, biotin, or antibodies, which provide specific adsorption sites for analytes of interest. In the case of lipid bilayer functionalization, the bilayer serves the role of the hetero-bifuctional layer. If pure lipids are used, the sensor can probe the interaction of various soluble species such as amyloid forming peptides, or drug molecules with the bilayer itself. Alternatively, commercial sources of lipids with modified headgroups are

- 8 -

available, which enables dyes, biotin, or reactive species to be anchored into the bilayer. A reactive species may be used to anchor peptides or antibodies to the bilayer, and be used to capture analytes of interest.

There are however, a number of practical challenges to depositing lipid bilayers on metal surfaces and further challenges still to the characterization and verification of bilayers on smooth and patterned metal surfaces. The experimental portion of this work is focused primarily on overcoming these challenges. Specifically, a new method of depositing bilayers on gold surfaces is presented which differs from existing methods insofar as it produces true bilayers by self-assembly without the requirement for specialized lipids with modified headgroups to promote the lipid interaction with the metal surface. This method therefore constitutes a generalized approach to depositing nearly any lipid species or mixture thereof onto gold surfaces. This is important in the context of including functionalized lipid species to act as specific receptor sites for analytes of interest. Furthermore, the versatility of this method enables the deposition of biologically relevant model lipid systems that mimic biologically important aspects of cellular membranes such as lipid rafts. This is demonstrated experimentally by depositing ternary lipid mixtures -known to phase segregate into raft-like domains- onto gold substrates and imaging and identifying the lipid phases. The method is also applied to producing phase segregated lipid bilayers on various nanoaperture arrays including nanohole arrays, nanoslit arrays, and annular aperture arrays.

The lipid layers are characterized by atomic force microscopy (AFM) using both imaging modalities and force spectroscopic methods. Specifically, force indentation maps, which involve indenting the AFM tip into the bilayer to probe its mechanical properties as a function of position, are used to verify, identify and visualize the coexisting lipid phases. The challenges and innovations required to perform this characterization are detailed in later chapters.

- 9 -

1.3 Future applications

As an alternative to the traditional use of antibodies or peptides as artificial receptor sites, natural receptors in the form of cellular membrane proteins can be inserted into the bilayer and used to capture analytes of interest. By using natural membrane-bound proteins, a natural cell membrane is most closely mimicked. Indeed, this strategy has been reported in literature to capture live cells on a surface, and could perhaps be a part of a live-cell assay. Figure 1-1 is a schematic illustration of how a lipid bilayer on a nano-hole array, hosting a transmembrane protein can mimic a cell surface and be used to capture a live cell. The proteins illustrated in the figure form an immunological synapse, which is known to be a fairly strong intercellular junction between a helper T-cell and a B-cell.

Figure 1-1 Diagram of a supported lipid bilayer on a NAA surface and the capture of a CD4+ T-cell from solution via an immunological synapse

Later chapters demonstrate how membrane extracts from live cells can be reconstituted with artificial lipids and cast onto mica surfaces. AFM images of these supported

- 10 -

bilayers suggest that many membrane-bound proteins are integrated into the supported bilayer.

Regardless of the choice of lipid system, binding of analytes to the lipids can be monitored by collecting Vis-NIR transmission spectra through the arrays. This involves illuminating one side of the sensor with white light, and collecting the far-field light transmitted through the array. Ideally, the use of a CCD could enable simultaneous collection of light form multiple arrays fabricated on the same substrate. Here, spectral information can be obtained by software analysis of RGB data collected from the CCD.

The advantage of using a CCD is that it can be integrated very closely with the sensor, and enable a high density of active sensing elements to be monitored simultaneously. Specifically, multiple NAAs can be fabricated on a gold film supported by a glass substrate, and the glass side of the sensor can be placed in contact with the CCD. Images of the CCD will consist of the light transmitted though all of the NAAs, which can be separated by isolating the pixels on the images corresponding to each array. RGB

- 11 -

information from each cluster of pixels can be used to monitor the colour of the transmission peak. Figure 1-2 illustrates this concept.

In this work however, a microscope is used to collect the light emerging from one array at a time and sent to a spectrometer for spectral analysis. Specifically, a lipid bilayer is deposited on a large centimeter-scale chip that contains several nanoaperture arrays with various geometries. The lipid layer is applied simultaneously to all arrays, and spectral shifts due to the deposition of the lipids is recorded one array at a time by translating the sample across the microscope stage, and into the area where light is captured.

Microfluidics can, in principle, be used to deliver different analytes to different NAA elements on the multiplexed array. The same microfluidics can be used to deliver different vesicles hosting different surface proteins to different NAAs, where they can self-assemble into functional bilayers specific to different molecules.

As a whole, the ideal device will consist of an array of NAAs milled into a gold film supported by a glass slide. A lipid bilayer deposited on the gold will host transmembrane proteins, or lipid-bound antibodies or peptides, which act as receptors for analytes of interest. Potentially, multiple NAAs can support lipid bilayers with different surface proteins or antibodies capable of targeting different analytes. Microfluidic channels will be used to deliver both the functionalized lipids, and the analyte. The substrate will be placed in close proximity to a CCD so that light transmitted through multiple NAAs can be collected simultaneously and separated on the image, and spectrally decomposed by software.

Possible applications of such a device may include:

- 12 -

• Parallel testing and screening drugs to characterize their interaction with

cellular membranes, which are targets for many drugs, and important

determinants of their delivery, and dosing.

• Medical diagnostics: Capturing and detecting cells, viruses, and antibodies.

• Characterizing the pathological mechanisms of diseases that act on

membranes, notably HIV and Alzheimer’s. (For example, by measuring the

binding properties of HIV envelope proteins, or Amyloid‐ß proteins to

various supported membranes)

• Fundamental research: measuring the water/membrane partition coefficient

of proteins, peptides, and small molecules.

1.4 Conclusions

Overall, the experimental work presented in this thesis represents important innovations in the deposition and characterization of lipids for biosensing applications. Although model lipid membranes, nanoaperture arrays, and cancer cell membrane extracts are used as examples to demonstrate the principles of the preparation and characterization methods employed, they also serve as powerful examples of the versatility and potential of these methods.

1.5 References

1. Cordero-Morales, J.F. et al. Molecular determinants of gating at the potassium- channel selectivity filter. Nat. Struct. Mol. Biol. 13, 311-318 (2006).

2. Sun-Wada, G. et al. A Proton Pump ATPase with Testis-specific E1-Subunit Isoform Required for Acrosome Acidification. J. Biol. Chem. 277, 18098-18105 (2002).

- 13 -

3. Sanguinetti, M.C. & Tristani-Firouzi, M. hERG potassium channels and cardiac arrhythmia. Nature 440, 463-9 (2006).

4. Grakoui, A. et al. The Immunological Synapse: A Molecular Machine Controlling T Cell Activation. Science 285, 221-227 (1999).

5. Todd, J. et al. A molecular basis for MHC class II--associated autoimmunity. Science 240, 1003-1009 (1988).

6. Siliciano, R.F. The role of CD4 in HIV envelope-mediated pathogenesis. Curr. Top. Microbiol. Immunol 205, 159-79 (1996).

7. Cordon-Cardo, C. et al. Expression of the multidrug resistance gene product (P- glycoprotein) in human normal and tumor tissues. J. Histochem. Cytochem 38, 1277- 87 (1990).

8. Jayasinghe, S., Hristova, K. & White, S.H. MPtopo: A database of membrane protein topology. Protein Sci 10, 455-458 (2001).

9. Wallin, E. & Heijne, G.V. Genome-wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms. Protein Science 7, 1029-1038 (1998).

10. Ebbesen, T.W., Lezec, H.J., Ghaemi, H.F., Thio, T. & Wolff, P. Extraordinary optical transmission through sub-wavelength hole arrays. Nature 391, 667-669 (1998).

11. Krishnan, A. et al. Evanescently coupled resonance in surface plasmon enhanced transmission. Optics Communications 200, 1-7 (2001).

12. Homola, J. Present and future of surface plasmon resonance biosensors. Analytical and Bioanalytical Chemistry 377, 528-539 (2003).

- 14 -

Chapter 2: Background 2.1 Introduction

This chapter aims to contextualize and motivate the methods and experimental findings of this thesis. First, a brief review of relevant background information form selected literature is presented. This is followed by a brief description of the general methods employed in later chapters. Then, early experimental results that have influenced the methods used in the experiments presented in the chapters that follow are reviewed. Finally, important lessons from the reviewed literature, and early experiments are summarized.

2.2 Brief Review of Literature

This section is a brief review the literature necessary to both contextualize the project, and justify the methods employed. Since the project represents the assimilation of topics in optics, biology, and chemistry, this section will be presented in the following sub- sections:

• 2.2.1 Surface plasmons and SPR sensing

• 2.2.2 Surface plasmons on nanophotonic structures

• 2.2.3 Exploiting cellular interactions

• 2.2.4 Phospholipid chemistry

• 2.2.5 Supported lipid bilayers and methods of preparation

- 15 -

2.2.1 Surface plasmons and SPR sensing

The interface between a metal and a dielectric is a boundary through which conduction electrons from the metal cannot freely propagate. Thus, conduction electrons become bound in one dimension at the interface, forming a two dimensional gas on the metal surface. This 2D electron gas can be driven to oscillate with an external oscillating electric field, typically at optical or infrared frequencies. A collective oscillation of surface electrons driven in resonance with an external optical field is generally called a

Surface Plasmon (SP)1.

Among surface plasmons, there are three main types: (1) localized surface plasmons

(LSPs)2, (2) surface plamon polaritons (SPPs), and (3) surface plasmon polariton Bloch waves (SPP-BWs)3, which differ both in the method by which they are excited, and their intrinsic physical properties. LSPs exist on metal nanoparticles, or isolated holes in a metal film, while SPPs exist on smooth planar metal surfaces, and SPP-BWs exist on periodic metal/dielectric interfaces.

SPPs are the most generic form of surface plasmon, in addition to being the best characterized and most widely employed in applications. Mathematically, SPPs can be described as a plane wave propagating along a direction within the plane of the interface.

Specifically, if the interface lies in the X-Y plane, a SPP can described by the equation for a general oscillating wave in one dimension:

- 16 -

ψ = A ⋅ exp(ikspp x −ωt + φ) Equation 1

€ where kSPP is the SPP wave vector, the magnitude of which is defined by the properties of the interface 4:

Equation 2

Here, εd is the dielectric constant of the insulator, and εm is the dielectric constant of the metal at the frequency of the external oscillating field. This expression incorporates the effect of the dielectric properties of the interface, and thus encompasses the mechanism by which SPPs can be used for biochemical sensing. This expression also describes how the dispersive properties of the metal influence SPP dispersion.

The dispersion diagram for a SPP provides insight into a very important property of SPPs with implications to the method by which they can be experimentally produced. Figure

2-1 depicts the dispersion diagram for a SPP as well as that for a plane wave in air. The fact that these two lines do not intersect, suggests that energy from a planewave cannot be directly coupled into a SPP mode. At any given frequency, a planewave in vacuum has less momentum (proportional to |k|) than a SPP. This disparity in momentum can be overcome either by changing the dispersion of SPPs or by changing the light line (

- 17 -

where n is the optical index of the medium). The latter can be accomplished if the medium in which the excitation field is propagating has a higher optical index than that in which the SPP is propagating1.

Practically, this can be accomplished using either the Kretschman configuration or the Otto configuration. In the case of the Kretschman configuration, light is projected through a glass prism (n = 1.45, for example) and incident on a metal film deposited on the base of the prism. The light traveling in the prism, with dispersion ω = c0k/n1, will have an evanescent component that exists on the metal/prism interface. If the film is thin

- 18 -

enough (typically < 100 nm 4), this evanescent component can tunnel through the metal film to the metal/air interface, where it would then have enough momentum to excite a SPP. In this case, because the light is incident on the metal at some angle, the relevant magnitude of k is that of the in-plane component. This means that by changing the angle of incidence, the magnitude of the in-plane component of k can change, and as the angle of incidence approaches normal, the magnitude of the in-plane component diminishes. Also, at some critical angle, SPPs can no longer be excited.

In practical terms, SPPs are observed by monitoring the light specularly reflected from the metal/glass interface. At certain angles of incidence, light is efficiently coupled into SPPs on the metal/air interface, and since light is absorbed, specular reflectance is minimal. This observation is critical to the operation of commercial surface plasmon resonance (SPR) biosensors. Here, the metal surface is functionalized to selectively bind the analyte of interest. When the analyte binds to the surface, it changes the dielectric constant (εd in Equation 2) of the SPP interface, which changes the angle of incidence at which SPPs are resonantly excited, that is, the angle at which specular reflectance is zero.

The advantages of SPR sensors are that they offer very high sensitivity, with current state of the art devices having sensitivities on the order of 3000-8000 nm per refractive index unit (RIU) 4, and are also suitable for monitoring binding in real-time, providing kinetic information. The main disadvantage of SPR is that it is difficult to monitor the binding of a large number of different analytes simultaneously, as doing so would require the alignment of multiple optical paths for multiple specularly reflected beams at multiple angles of incidence. Another limiting factor is spatial resolution of the device, as having a higher spatial resolution would enable more analytes to be distinguished on the same surface.

- 19 -

These difficulties are partially mitigated with SPR imaging, which have been operated with around 100 channels. In this case, the base of a prism is patterned with a number of different active layers each measuring 400x800 µm in size. Active layers are paired together to produce a differential signal. Active regions are differentiated spatially from inactive areas by polarization contrast. A CCD collects an image, which has each channel spatially separated. The refractive index resolution of the device is reportedly 5x10-6 RIU.4

The other types of surface plasmons, namely LSPs, and SPP-BWs can be explained in relation to the plasmonic structures that support them.

2.2.2 Surface plasmons on nanophotonic structures

LSPs can be considered as surface plasmons that are confined to one dimension. The best known physical examples of LSPs are those supported by metallic nanoparticles5. LSP excitation begins when incident light can penetrate the entire metal nanoparticle, and cause the conduction electrons to oscillate relative to the atoms that make the particle’s core, creating an oscillating dipole. When light of a specific frequency is resonant with an eigenmode of the dipole oscillation, a LSP resonance occurs. For example, when gold nanoparticles are made to be about 10’s of nanometers in diameter 6, they take on a red colour as a result of the absorption of green light by LSPs.

The optical properties of gold nanoparticles of various shapes and sizes have been well characterized7-10. The size and shape of the nanoparticle has a dramatic effect on the colour of the nanoparticle suspension. Nanorods of different lengths will also have different absorption spectra, and a wide range of size-tunable colours. Current research towards the application of LSPs on gold nanoparticles, nanorods, and nanoboxes includes their use as anticancer treatments11, and as chemical sensors6. Nanoparticles or rods with

- 20 -

resonant frequencies in the range of 700-1000 nm 6,11, the optical band-pass region of biological tissue, can be excited within the body using a light source outside the body, and are being explored as non-invasive probes and treatments.

Isolated subwavelength holes in metal films can also support LSPs. Like metal nanoparticles, nanoholes that support LSPs have also been well studied in terms of size and shape12-14. In this case, their fixed geometry is convenient both for direct imaging of surface plasmons and other near-field optical features near the holes using nearfield microscopy12,2,15 and for studies of the effect of incident polarization on the optical transmission of isolated holes16.

Understanding the near-field behaviour of holes either in isolation or in small interacting clusters can aid in understanding the behaviour of large arrays of holes such as NAAs. Our own observations suggest that when small clusters of holes are placed closer together than the decay length of a surface plasmon, the plasmons can overlap and interact, and this interaction can be modeled by considering the LSP from each hole as a point source for surface bound optical wave, and calculating their interference 2.

Periodic structures in metal films can range from one-dimensional corrugations on the surface, to two dimensional arrays of apertures of various shapes. In any case, the structures introduce spatial periodicity in the dielectric constant of materials, which introduces periodic boundary conditions for surface plasmons. This gives rise to surface plasmon polariton Bloch waves (SPP-BWs) which are analogous to electronic Bloch waves in semiconductor electronics, and likewise gives rise to dispersion and a band gap, much like a 2D photonic crystal would. The transition from a flat dispersion to a Bloch- wave dispersion was clearly demonstrated by measuring dispersion on a series of samples that ranged from an array of non-interacting metal islands to a periodic array of holes17.

- 21 -

The most common model used to describe the excitation of surface plasmons on periodic surfaces involves momentum matching between diffracted orders and surface plasmons4,1. Recall that the dispersion diagram from a surface plasmon on a smooth surface does not intersect the light line, implying that light at a given frequency, lacks sufficient momentum to excite a surface plasmon. Positive orders of diffraction have a greater momentum than the incident beam and is proportional to the wave vector given by

Equation 3

where a is the period of the grating, and 2mπ/a is the grating vector, which is proportional to the additional momentum that the mth diffracted order has over the in-plane component of the incident field. When |kx|m equals the magnitude of the wave vector of a surface plasmon given by Equation 2, then it is possible for incident light to couple into surface plasmon polariton modes. The surface plasmon dispersion depends on the dielectric constants of the interfacing materials with εm being the dielectric constant of the metal, while εd is the dielectric constant of the interfacing dielectric material.

For two-dimensional gratings, two orthogonal indexed components of the diffracted order must be considered, but the treatment is otherwise the same 18. Of course the angle of the in-plane component relative to the principle axes of the 2D grating is also a factor. The k vector of the diffracted beam is the vector sum of the incident k, and the two indexed k’s from the grating. These must have a magnitude at least as great as that of a surface plasmon in order for resonant coupling to occur, giving the condition for SPP excitation at normal incidence:

- 22 -

Equation 4

Here, i and j are the indices of the orthogonal components of the diffracted order. This condition can be rearranged and solved for wavelength, giving an expression to predict the surface plasmon wavelength:

Equation 5

This depends on the dielectric properties of the interfacing materials (ε), the periodicity of the array (a), and the mode indices i and j.

Though this model is widely accepted, it fails to accurately predict the wavelength of surface plasmons observed in practice, and in fact, the observed surface plasmons have a wavelength that is often longer than those predicted by this model. Nevertheless, this model serves well as a rough prediction until such time as a more accurate, yet simple model becomes available. This model treats the holes as points in space, and as such, analytical models based on this do not exist for more complicated hole geometries. In fact, modeling of plasmonic structures is often done by numerically solving Maxwell’s equations for optical fields using the finite difference time domain (FDTD) method3,19,20.

Experimentally, the existence of surface plasmons is justified by the observation that, under certain conditions, some nanoaperture arrays that are only 10% porous by area can

- 23 -

exhibit up to 20% optical transmission for certain wavelengths21. In a classical optics interpretation, this would suggest that some light that impinges on the metal will also make it through the film. Moreover, for holes with diameters below the diffraction limit, the expected optical transmission efficiency is less than 0.1% for isolated holes18,22. The generally accepted explanation for this behaviour suggests that incident light excites SPP- BWs on the array and since SPPs are evanescent waves, they can tunnel efficiently through the holes in the array where they scatter off of the features on the back side into propagating fields that can be observed in the far field. The observation of a peak in the transmission spectrum whose wavelength changes linearly with array period supports the Bloch-wave argument. Additionally, the observation that the peak position depends on the dielectric constant of the environment also corroborates the role of surface plasmons in the mechanism of transmission.

Published systematic experimental studies on the optical properties of these arrays are also valuable to determine design parameters of these arrays. Studies with various hole shapes23,24, hole sizes25, and array periods26 can be used to optimize both the optical throughput of the array as well as the position of the main transmission peak. The effect of the dielectric constant on the optical properties has likewise been published21,27. A summary of relevant observations from these studies are:

• The increase in transmission efficiency with increasing hole size begins to plateau at hole sizes of about 200 nm

• Arrays of round and square holes behave in qualitatively similar ways • Periods of about 600 nm place the peak transmission near 680 nm in air • Immersion in increasing dielectric media causes a linear red shift in the peak position

• Typical sensitivities are 200-300 nm/dielectric constant

- 24 -

Several published studies have proposed the use of similar structures as transduction elements in bio/chemical sensors. Another promising structure is the Annular Aperture Array (AAA), which is an array of ring-shaped holes. In the visible wavelengths, AAAs behave much like NAAs. In fact, preliminary results from our collaborators suggest that the dielectric dependence of the peak position is nearly identical to NAAs. Unlike NAAs however, AAAs have guided modes in the near infrared (NIR) afforded by their geometry. In the literature these modes are also attributed to cylindrical surface plasmons (CSPs). These NIR peaks exhibit up to 50% transmission. The dielectric dependence of the NIR modes has not been characterized.

To the knowledge of this author, at the time of this writing, no one has demonstrated the use of NAA structures to detect cellular interactions, or published work that combines NAAs with artificial lipid bilayers for sensing. There is however, one group that has combined lipid bilayers with Au or Ag films patterned with a random distribution of holes that support LSP resonances28. They used biotinylated lipids in a lipid bilayer matrix to conjugate neutravidin from solution, and detect binding by observing LSPR shifts. In their experiment, the metal film was coated with a thin layer of SiOx to enhance vesicle fusion. In this case, the authors believed that the lipid followed the topography of their surface.

A recent study demonstrated lipid bilayers deposited by vesicle fusion can span nanoscale holes in Au-coated porous alumina29. In this case the Au has been functionalized with a negatively charged monolayer, and lipid vesicles, which contained positively charged headgroups spontaneously fused to form a bilayer on the surface. Also, because the size of the vesicles was larger than the holes, they were unable to enter the holes before rupturing, which produced lipid bilayers that spanned the holes. The plasmonic properties of the holes were not investigated. In the literature, pore-spanning lipid bilayers, in the form of black lipid membranes (BLMs) were the focus of early studies of artificial lipid bilayers30. Various means of preparing BLMs have been reported

- 25 -

including painting of the lipids by brush, and transfer from Langmuir films30-32. In this way, lipids were made to span holes with mm2 dimensions in Teflon or loops of hair. Proteins have also been incorporated in these systems. Additionally, calcium-induced phase segregation has been observed in bilayers spanning sub-micron holes in porous carbon grids using electron microscopy33. The lipids were transferred to the porous carbon grids by the Langmuir-Blodgett technique. These early systems were studied primarily by microscopy and electrochemical techniques. Additionally, membranes from live cells have been transferred directly to porous silicon nitride (SiN) by stamping the SiN membrane against a live cell. The porous SiN was part of a semiconductor field- effect transistor used to probe the electrophysiology of the cell membrane34. Neither optical spectroscopy nor membrane protein mediated binding to the surface was reported.

2.2.3 Exploiting cellular interactions

The idea of using supported lipid bilayers to mimic cell surfaces on sensors is has existed nearly as long as supported bilayers themseves. In 1984, Harden McConnell’s group demonstrated that antigen presentation via class-II MHC proteins on planar artificial bilayers were capable of activating helper T-cells captured from solution35. In 1986, McConnell’s group combined this idea with an evanescent wave device (TIRF) to observe fluorescence resonant energy transfer (FRET) from labelled MHC-II molecules incorporated into the supported membrane, and labelled TCRs on live T-cell hybridomas that were captured from solution36.

More recently, Jay Groves and coworkers have used artificial antigen presenting surfaces consisting of MHC molecules and intercellular adhesion molecule-1 (ICAM-1), which binds to leukocyte function-associated antigen-1 (LFA-1) on the T-cell surface, to capture T-cells37. This study demonstrated that multiple TCR/MHC complexes for the core of the immunological synapse, while ICAM-1/LFA-1 complexes surround them. Using patterned barriers on the surface to restrict lateral diffusion of the surface proteins, they

- 26 -

also showed that the immunological synapse can tolerate these spatial restrictions and still form a tight interface.

Other methods of capturing T-cells that do not involve MHC proteins are also possible. For example, antibodies to CD4, or to other surface markers on T-cells, can be covalently attached to the sensor surface. In this case, the functional molecules have no lateral mobility. If, however, otherwise soluble ligands are anchored to a fluid bilayer they will remain mobile across the surface38. This is helpful for situations where the cell contacts the surface through multiple molecular interactions that need to reorganize on the surface to allow for tight binding. For example, antibodies to the T-Cell receptor, CD4, and LFA-1 can be used simultaneously to mimic the immunological synapse. This method is particularly attractive because MHC proteins are not readily available from commercial sources, and their purification from recombinant sources could be difficult.

Lipid bilayers have also been suggested in a variety of sensor constructs that involve the detection of soluble analytes. Because lipid bilayers are important to so many processes in biology, sensors utilizing lipid bilayers are desirable as tools to understand these interactions. For example, sensors with lipid bilayer functionalzations are used in many stages of drug discovery and testing39, as well as in basic research to understand the mechanisms of disease40,41, or functions of membrane proteins42-44. Furthermore, in the context of biosensor functionalization methods, lipid bilayers have several advantages. First they preserve the lateral mobility of the surface receptors that are determinants of the quality of analyte binding, especially in multivalent interactions, which have configurational and length-scale dependences37,45. Secondly, previous reports suggest that in some cases, cells, proteins (such as human fibrinogen, human serum albumin, and human immunoglobulin), and platelets do not adhere readily to model lipid bilayers46-49, thus the lipid bilayer itself acts to resist nonspecific adsorption. Non-specific adsorption is one of the greatest challenges to surface-based biosensors.

- 27 -

2.2.4 Phospholipid chemistry

The properties of natural and model lipid bilayers can be understood in terms of the physical and chemical properties of the component molecules. This section reviews the structure and properties of the main components of model lipid mixtures, and the consequences of their molecular properties on the properties of structures such as vesicles, and bilayers that form from assemblies of lipid molecules. The relationship between the properties of model systems, and the properties of natural cell membranes is also summarized.

2.2.4.1 Phospholipid structure and properties

Typical phospholipids are amphiphilic molecules that consist of a polar phosphate- containing head-group and two fatty acid chains joined by a glycerol backbone. In length, the fatty acid groups are typically much larger than the headgroup and glycerol combined, and is often referred to as the tail-group. The tail-group may consist of 14 to 24 carbon atoms joined in a chain by single or double bonds. The length and degree of saturation strongly influence the physical properties and phase behaviour of the lipids. Additionally, the two tail-groups may differ in length and saturation. Variations in headgroups can also occur, but are typically comprised of a phosphate group and a small organic molecule such as choline. Depending on the nature of this small organic molecule headgroups can vary in net charge: positive or negative, or can be zwitterionic. Examples are given below. The glycerol portion of the molecule can vary as well, such as in sphingomyelin, which replaces the diglyceride with a sphingosine group.

By convention, a four-letter naming system is commonly used in the literature where the first two letters indicate the two fatty acid chains, while the last two letters indicate the head-group chemistry. For example, POPC stands for 1-Palmitoyl-2-oleoyl-sn-glycero- 3-phosphocholine: this indicates that the two fatty acid chains that form the tail group

- 28 -

differ in chemistry and are described by their common names Palmitic acid and Oleic acid which are a 16-carbon saturated fatty acid, and a 18-carbon mono-unsaturated fatty acid respectively. The PC stands for phosphatidylcholine, which is zwitterionic. For another example, DSPE stands for 1,2-Distearoyl-sn-glycero-3-phosphoserine, which indicates the two fatty acid chains are identical (symmetric) and consist of Stearic acid (an 18-carbon saturated fatty acid), and the head-group consists of phosphtidylserine, which has a net-negative charge.

Like other amphiphiles, phospholipids form molecular assemblies such as vesicles when present at concentrations higher than the critical micelle concentration (CMC). The assembly of surfactant aggregates sequesters the hydrocarbon tails of the surfactant, resulting in a favourable decrease in the interfacial energy between water and the hydrocarbon tail, and in some cases a possible increase in electrostatic repulsion between neighboring head-groups as well. However, the assembly of aggregates is also accompanied by the unfavourable decrease in entropy for each molecule in the aggregate. A balance between the energies associated with these factors determines the critical micelle concentration. The type of aggregate that is formed, for example a micelle or bilayer vesicle, is also determined by these factors in addition to geometric constraints, which limit the size and curvature of the assemblies. Detergents with single acyl chains and large headgroups readily form micelles, which favour large head-groups and limited hydrocarbon volume. In the case of phospholipids with two acyl chains, the effective volume of the hydrocarbons doubles without appreciable differences in the area occupied by headgroups. This geometry severely restricts the allowed curvature of a potential structure, and thus, planar bilayers are favoured over micelles. Bilayers of finite size have high interfacial energy at their edges, thus larger bilayer structures are favoured. However, larger structures have an entropic penalty. Bilayer vesicles are intermediate: eliminating the interfacial energy between water and lipid, but also having a finite size (and thus, a finite entropic penalty).

- 29 -

In bilayer structures, the thickness of the bilayer is determined by the length of hydrocarbon chain. Typical biological membranes have thicknesses of 4 to 5 nm, and integral membrane proteins tend to have congruent structures, with hydrophobic portions of the protein being about the same thickness of the bilayer. Thus, lipids for model systems are chosen with appropriate length to mimic this thickness.

2.2.4.2 Phase behaviour of lipids

Within the bilayer, interactions between adjacent phospholipid molecules determine their physical properties. These properties can be described in terms of the phase-behaviour of the lipids. Saturated lipids, (and some monounsaturated lipids with the double bond near the terminus of the acyl chain, for example: sphingomyelin) tend to form a stiff gel-phase at room temperature because the linear geometry of the alkane chains allows for close- packing of the lipids. In this configuration, the strong van der Waals interactions between adjacent alkane chains dominate their interaction. At higher temperatures phase 50 transition from gel to 2D liquid occurs (for example Tm ≈ 39°C for sphingomyelin ). Phospholipids with unsaturated acyl chains can be kinked, which sterically disrupts the packing of adjacent lipid molecules. Consequently, bilayers of these lipids tend to form a 2D liquid at room temperature. A liquid-to-gel transistion typically occurs at low temperatures (for example Tm = -22°C for DOPC51).

Cholesterol in the bilayer also has a profound effect on the phase behaviour of lipid bilayers. Evidence suggests that cholesterol associates preferentially with sphingomyelin because the long, straight acyl chains provide a surface with which the cholesterol can interact52-54. Additionally, the cholesterol is believed to hydrogen bond with the sphingosine group in SM53. In model systems SM and cholesterol from a liquid-ordered

(lo) phase at room temperature. The lo phase, compared to the gel phase formed by pure SM, is more fluid, yet retains a large degree of order between lipid molecules when compared to the liquid phase of DOPC at room temperature, which is often referred to as

- 30 -

the liquid-disordered (ld) phase. In mixtures of DOPC, SM and Chol, these two liquid phases have been observed to coexist. In fact, cholesterol has been identified as a critical component in the separation of liquid phases. In mixtures of SM and DOPC, phase segregation does not spontaneously occur in the absence of cholesterol, or in the presence of modified sterol derivatives53,55.

2.2.4.3 Lipid rafts and model membranes

Natural cell membranes can vary greatly in composition depending on cell or organelle function and environmental conditions56, but consist primarily of zwitterionic monounsaturated PC and PE phospholipids, saturated, zwitterionic sphingomyelin (SM), and cholesterol (chol). Phospholipids with various charged head-groups such as PS, and phosphotidyl inositol (PI) are also present in smaller relative quantities in most membranes. The choice of lipids used in model systems reflects the composition of natural cell membranes. In particular, ternary mixtures of DOPC, SM, and chol are widely used in literature, where DOPC features a symmetric monounsaturated fatty acid tail group, and a zwitterionic PC head-group. These lipid mixtures are known to phase segregate into co-existing lipid phases that mimic lipid rafts.

The lipid raft hypothesis has gained wide acceptance in the current view of cellular membranes57,58,52. Recent experimental findings suggest that a simple model of cell membranes that describes them as random mixtures of lipids, cholesterol, and proteins is insufficient in accounting for some aspects of membrane function, and that some degree of lateral organization within the plane of the bilayer is necessary59,60. The lipid raft provides a construct by which proteins and lipids can organize and/or segregate themselves in the plane of the bilayer.

- 31 -

Early experiments with detergent resistant membranes found that in model bilayers, lipid phases containing tightly-packed saturated lipids (such as gel, or liquid-ordered phases) were more resistant to being dissolved in detergent solution than lipid phases composed of unsaturated lipids (such as the liquid-disordered phase), and that natural membranes also exhibit this behaviour, suggesting the coexistence of phases - and thus rafts - in natural cell membranes61. More recently, it has been suggested that the presence of the detergent can induce the formation of phases in cellular membranes, and thus some controversy still exists62,63,53. Though phase segregation of membranes in live cells has not been directly observed, the existence of coexisting lipid phases has been inferred from a variety of indirect evidence, which includes evidence gathered from the behaviour of model lipid bilayers.

Advancements in study of supported model lipid bilayers and high resolution imaging of lipid bilayers by atomic force microscopy (AFM) have added evidence that model lipid mixtures that contain components found in natural membranes spontaneously separate into coexisting phases64,65. Additionally, there is experimental evidence that the pathologies of diseases such as Alzheimer’s and HIV/AIDS depend on the presence of lipid rafts66,67. These findings suggest that in order to better mimic the function of cell membranes, supported lipid bilayer surface functionalizations would ideally contain a mixture of lipids that mimic the function of rafts.

The phase behaviour of model lipid mixtures of DOPC, SM and chol have been well studied68-71, and the formation of raft-like domains in a fluid matrix (i.e. segregation of the mixture into two phases) has been directly observed using fluorescence microscopy using probes that preferentially sequester into one of the two phases72, and by AFM, where the two phases can be directly observed by their difference in height64,65, and identified by differences in their mechanical properties73.

- 32 -

2.2.5 Supported lipid bilayers and methods of preparation

In order to incorporate transmembrane proteins or membrane surface-bound proteins onto the surface of the sensor, it is necessary to construct a lipid bilayer across the surface of the sensor. There are several well-studied methods to make supported bilayer lipid membranes (sBLMs), including the Langmuir-Blodgett technique, and the vesicle fusion technique.

2.2.5.1 Vesicle fusion

The vesicle fusion method involves incubating a suspension of bilayer vesicles with a hydrophilic surface. The vesicles spontaneously adsorb to the surface, where their tendency to wet the surface creates a surface tension on the vesicle wall. When the surface tension reaches a critical level, the vesicle will rupture and spread on the surface, thereby forming a lipid bilayer on the surface 74. With sufficient time, the whole surface can be coated by a continuous and uniform bilayer. Experimental details about the technique will be addressed in the next section.

Proteins can be incorporated into the bilayer while in vesicular form, by introducing detergent-solublized transmembrane proteins into the vesicle suspension. Repetitive freezing and thawing of the solution allows some of the protein to become incorporated into the vesicular bilayer. The orientation of the protein in the bilayer is random, meaning that once the vesicles are fused to the surface, some proteins will be facing the substrate while others will be facing solution75,76.

It has been suggested that the expression of membrane proteins with large extramembraneous domains on supported bilayers leads to the denaturing of the protein

- 33 -

due to its interaction with the substrate77. However, it has been demonstrated that transmembrane proteins can be hosted within pore-spanning lipid bilayers because the aqueous space within the pore can accommodate the extramembraneous domains42. Literature suggests that the lipid layers deposited by vesicle fusion can either span the hole or coat the interior of the hole.28,29. It is likely that if the vesicles are made to be much larger than the holes, then they will bridge over the holes upon rupturing whereas smaller vesicles can diffuse into the hole and rupture at the bottom of the hole. It is conceivable however, that annular apertures are more easily spanned, as the physical size of the gap is smaller. The results presented in the experimental section provide evidence that pore-spanning lipid layers have been achieved over various apertures.

Although there are many advantages to the vesicle fusion method that have contributed to its popularity, there are some shortcomings of the technique when applied to nanoaperture arrays. Although the vesicle fusion technique is reliable on glass and mica surfaces, it is difficult to produce on metal surfaces. There are however several reported techniques that involve modifying the metal surface with more hydrophilic groups78,79, or 28 with a thin layer of sputtered SiO2 . In this thesis, a relatively simple method to depositing lipid bilayers on gold surfaces is demonstrated, which will be summarized in both the methods and the results sections. Briefly, gold substrates were modified with

SH-(CH2)11-COOH self assembled monolayers to improve hydrophilicity. Vesicles then fused spontaneously to the surface to from bilayers.

2.2.5.2 Langmuir-Blodgett/Langmuir-Schaefer method

The Langmuir-Blodgett (LB) method is a means of transferring a floating monolayer from water surfaces to a solid substrate by vertically dipping the substrate into, or pulling a substrate out of a floating monolayer with the substrate being perpendicular to the water’s surface. The substrate can be dipped in and pulled out repeatedly to make multilayer films 80.

- 34 -

The process requires a specially made trough that consists of a shallow teflon-coated pan with a deep well that is large enough to accommodate a vertically dipped substrate. The trough is filled to the brim with water. The floating monolayer must be an amphiphile such as a fatty acid, or a phospholipid, and is cast on the water’s surface by dissolving it in a volatile organic solvent, and dropping the solvent onto the water’s surface. The solvent spreads and evaporates, leaving the amphiphile behind. Movable barriers that skim the water surface can be used to reduce the available surface area for the film, and thereby laterally compress the amphiphiles. A plate vertically suspended in the interface attached to a transducer, is used to monitor the surface pressure, which is an indication of the amount of surface compression. A feedback loop is used to keep the pressure constant during film transfer. Drawing a hydrophilic surface up though the monolayer transfers a portion of the monolayer onto the substrate with the hydrophilic headgroup facing the substrate.

The Langmuir-Schaefer (LS) method involves dipping a substrate into a Langmuir film with the substrate parallel to the film surface. To make a bilayer, a substrate with a monolayer deposited by the LB method (with the hydrophobic tail groups sticking out) is horizontally dipped onto the surface of a Langmuir film 81. This deposits the outer leaflet of the bilayer where the hydrophobic tail groups of both leaflets face each other, and the hydrophilic headgroups face the substrate and the environment.

The advantage of this method is that it may be possible to construct a bilayer on a gold surface without the need to chemically modify it first. Also, it is plausible that a monolayer transferred by the LB method can span the holes of the NAA. During transfer, water will fill the holes, and support the monolayer. Subsequent transfer of the outer leaflet via the LS method should preserve this geometry.

- 35 -

Protein incorporation into the membrane can occur either after transfer of a bilayer, or before the transfer of the first monolayer to a substrate. In some cases, an unfolded solubilized transmembrane protein can fold and insert into the bilayer 82. In other cases, peptide fragments can be covalently linked to lipids and spread with the monolayer 83, and subsequently transferred to a substrate. Alternatively, protein-containing bilayer vesicles can be spread on the air/water interface, and then transferred to a substrate 80. This process is essentially vesicle fusion at an air/water interface, and in some cases would be easier to apply directly to the substrate.

A possible advantage of Langmuir trough methods is that chemical modification of the gold surface is not required. In fact, there are a several reports that demonstrate the deposition of lipid bilayers on unmodified gold surfaces using a Langmuir trough84,85.

A comparison of the merits of vesicle fusion and trough-based techniques would identify that the advantage of the trough methods are the control over the surface pressure and therefore the quality of the lipid bilayer. On the other hand, the advantages of the vesicle fusion method are that, when combined with microfuidics, can be used to functionalize different parts of the substrate with bilayers loaded with different surface markers or proteins, which is particularly useful for large scale parallel sensing.

2.2.6 Summary of Background

The preceding was a brief review of the principles of surface plasmon resonance and their role in the extraordinary optical transmission observed through nanoaperture arrays. The efficiency of light transmission is wavelength dependent, and thus the transmission spectra typically containing a variety of peaks and valleys, which correspond to these

- 36 -

efficiencies. Enhanced transmission occurs at wavelengths corresponding to the wavelength of the SPP-BW, which depends on the geometry of the features in the array and the dielectric properties at the metal surface. The former dependence allows the optical properties to be engineered, while the latter dependence forms the basis of the sensing mechanism.

Also reviewed were examples from literature where lipid bilayer surface modifications were used to capture cells and soluble molecules to surfaces, usually for sensing applications. These applications exemplify the breadth of interactions that are mediated by lipid bilayers, and are of interest to fundamental and applied biology. The experimental portion of this thesis endeavours to make an impact on the use of lipids in optical biosensors, especially model lipid systems that more closely mimic cellular membranes than current solutions, specifically insofar as enabling the formation of lipid rafts on these surfaces.

Various existing techniques for depositing lipid bilayers on various substrates were also reviewed including reports of lipid bilayers on porous substrates. These techniques are based on either the spontaneous self-assembly of bilayers from vesicles or on Langmuir trough deposition. The experimental portion of this thesis focuses on vesicle fusion, and the lessons learned form literature have lead to a method of depositing lipids on gold without the requirement for specially modified lipids or tethers, which is detailed in Chapter 3.

These topics and others are reviewed in more detail and in the context of the experimental work presented in the introductory sections of the chapters that follow. The remainder of this chapter meanwhile briefly summarizes the methods employed in the experiments presented, with more details given in their respective chapters. Additionally, some of the

- 37 -

early experimental results, which have influenced the evolution of the experimental methods employed, are also reviewed.

2.3 General Methods

2.3.1 Focused Ion Beam Milling

Several methods of fabricating nanoaperture arrays were explored, including top down techniques such as electron beam lithography, and focused ion beam milling, and a mixed bottom up/top down technique that employed self assembled diblock co-polymer structures that are selectively etched for use as templates for metal deposition. Among these methods, focused ion beam milling (FIB), once it became conveniently accessible, has been the most reliable method. Consequently, this section will briefly describe the method of FIB. In the literature, there are numerous sources of information on e-beam lithography86, and a body of primary literature on using diblock copolymers as templates87. These techniques and the processes involved are discussed in more detail in section 2.4.2.

For substrates patterned by focused ion beam milling were supplied by collaborators at the University of Pittsburgh. A typical fabrication process involved first depositing a film of gold on a quartz slide using an electron-beam assisted thermal evaporator. Gold, chromium and titanium sources and a clean quartz slide were placed in the evaporator under high vacuum. Heat, and an electron beam aimed at the metal targets were used to vapourize the metals, which then deposited on the quartz substrate. Mechanical shutters placed between each target and the quartz substrate were used to control the thickness of the metal films. Typically the film consisted of 3-5 nm of either titanium or chromium as an adhesion layer directly on the quartz substrate. Gold was then deposited on top to a final thickness of 200 nm.

- 38 -

Arrays of apertures were then milled into the metal films using focused ion beam milling (Seiko SMI 3050SE, Seiko instruments inc, Chiba, Japan) where the metal-coated substrate was exposed to a beam of ionized gallium, focused to a fine point, and controlled by electromagnetic fields in much the same way that electron beams are manipulated in electron microscopy. The beam of gallium ions focused on the metal surface mechanically sputters atoms from the surface. A computer script was written to control the beam parameters and the pattern that the beam cut out of the surface. Typically, the smallest feature sizes were about 20nm on gold surfaces. Due to limitations in computing power, hole array sizes were limited to 100 µm x 100 µm. Such arrays typically contained about 40,000 apertures. With AAAs, array larger array sizes are attainable, improving optical throughput and consequently signal to noise.

2.3.2 Spectroscopy

To verify the far-field optical properties of the arrays, and to determine their response to different dielectric environments, an Ocean Optics HR2000+ fiber optic coupled UV/Vis/NIR spectrometer (Ocean Optics, Duneedon, FL) was used. It was modified with a 50 micron entrance slit, which allowed ten-fold more optical throughput over the stock configuration but reduced spectral resolution from 0.66 nm to 1.85 nm. Source and collection optics were aligned parallel to an optical table so the sample can be mounted vertically in a glass colorimeter cell with a 5 mm path length. Light from a bright broadband light source, (either a tungsten light bulb, or a tungsten-halogen lamp) was focused onto the sample through a condenser lens salvaged from a commercial Leica microscope. A home-built horizontal microscope with a 40X achromatic microscope objective was used to collect light emerging from one array at a time. Light collected by the objective was focused into a multimode optical fiber through a plano-convex lens and sent to the spectrometer.

- 39 -

Alignment of the sample was achieved by mounting the colorimeter cell on a XYZ stage, and first visually checking the alignment and focus with a periscope attachment and a 10X ocular lens. After removing the periscope, alignment was fine tuned by observing the raw counts collected by the spectrometer and adjusting the in-plane position of the sample to maximize signal. Reference spectra were collected through a square hole cut into the same Au film to the same overall dimensions of the nanoaperture array. The dielectric environment was changed by filling the colorimeter cell with various solvents such as methanol, acetone, isopropyl alcohol, toluene, and hexadecane. Software provided by the spectrometer manufacturer was used to collect, and average spectra. Typically over 50 spectra were averaged.

In later experiments presented in Chapter 4, the transmission spectra were collected using the same Ocean Optics spectrometer for visible spectra, as well as an Ocean Optics near infrared spectrometer (NIR512) connected via optical fiber to the front port of a Nikon inverted microscope (Nikon Eclipse TE2000, Nikon Canada Inc., Mississauga, Canada).

AFM fluid cells, both custom made, and commercially available (Asylum Research,

Santa Barbara, CA, USA) were used to mount the arrays, and contain a water environment.

2.3.3 Vesicle fusion

In the vesicle fusion method, substrates were incubated for at least 30 minutes in a solution of unilamellar bilayer vesicles in biological buffer or MilliQ water at temperatures above the melting temperature (Tm) of all lipid components. During incubation, the vesicles adsorbed to the substrate surface, and spontaneously ruptured to form a bilayer. On the hydrophilic surface, spherical vesicles spread on the surface and

- 40 -

took on an oblong shape, which created surface tension on the vesicle walls. This led eventually to their rupture. After rupturing, the vesicles unfolded onto the surface to form a bilayer. As more vesicles ruptured more of the surface became covered in the bilayer until most of the surface was covered by a continuous bilayer. Following incubation, the substrates were rinsed very carefully under a light stream of buffer or water to remove unbound lipid.

Two was of preparing unilamellar bilayer vesicles were employed. In both cases unilamellar vesicles were refined from aqueous solutions of multilamellar vesicles. The multilamellar vesicles were made by creating a solution of phospholipids (1,2- Dipalmitoyl-sn-Glycero-3-Phosphocholine, Avanti Polar Lipids Inc., Alabaster, AL) in chloroform, which were rotary-evaporated under vacuum over night. This formed a multilayer cake on the walls of a round bottom flask (RBF). Biological buffer (phosphate buffered saline) or water was added, and the flask was sonicated in a bath sonicator. The miltilayers that lifted off the walls of the RBF formed multilamellar vesicles. The aqueous solution of multilamellar vesicles was refined into unilamellar vesicles by repeated extrusion through a Teflon filter with 100 nm diameter pores using a specially made device (Avanti Mini Extruder, Avanti Polar Lipids Inc., Alabaster, AL) in which the lipid suspension was passed from one glass syringe to another through the polycarbonate filter (Whatman, Piscataway, NJ, USA). After 12-14 passes through the filter, the solution became mostly clear with only a slight cloudiness detected. Light scattering studies reported by the manufacturer indicated that as the number of passes increases, the size distribution narrowed and became centered around the pore size. Typically only an odd number of passes was employed to ensure the lipid suspension ended up in the opposite syringe from which it started. This ensured that any dust or large contaminants accidentally introduced into the lipid suspension before extrusion, had been filtered out of the final vesicle suspension. This process was performed around 55 °C, which was above the melting temperature of the lipid. The extrusion method was preferred over the alternative, namely tip sonication, because the latter can leave behind metal particles from the tip. Furthermore, the extrusion method was inexpensive, and

- 41 -

reproducible while affording control of vesicle size through the use of polycarbonate filters with various sized pores. Thus, larger vesicles required for spanning larger surface features, were made by extrusion through filters with larger pores.

In addition to hydrophilicity, another factor that affeced the ability of vesicles to fuse on surfaces was the choice of buffer. According to88 the choice of buffer depended on the surface charge of the substrate. On negatively charged surfaces such as COOH-SAMs (at neutral pH) and mica, buffers with large molecular cations tended to gather near the surface and sterically interfere with adsorption. Small metal cations on the other hand did not interfere sterically and vesicles adsorbed to the surface. The condition was reversed for positive surface charges.

2.3.4 Atomic Force Microscopy of Lipid Bilayers

Formation of the lipid bilayer were verified by Atomic Force Microscopy (AFM). Conventionally, lipid bilayers on mica or glass coverslips were imaged under aqueous buffer or pure water in either contact, or intermittent contact mode. Defects in the bilayer were used to determine the thickness of the bilayer. Typical bilayer thicknesses were between 3 and 7 nm, depending on the length of the lipid chain. Many groups have studied various supported lipid systems using AFM 89-91.

As illustrated in the next section, lipids on rough surfaces such as thermally deposited gold, were much more difficult to image than on flat surfaces such as mica. The thermally deposited gold had a mean roughness close to 2 nm, which was about half the thickness of a bilayer. This made resolving the height of the bilayer difficult against the topographic noise. Two methods were employed to circumvent this. The first method employed flame annealed gold as the substrate. Over single grains, the mean roughness of flame-annealed gold was less than 1nm. This reduced the topographic background,

- 42 -

which allowed the ~1nm height difference between lipid domains to be observed directly. The procedure for flame annealing is described in the next subsection. The second method employed force indentation over the bilayer to differentiate between a bare gold surface and one coated with a soft bilayer, which was performed in tandem with imaging to help verify the nature of topographic features that were difficult to discern from the height image alone.

Chapters 3-5 descibe how force indentation mapping was employed, the details of which are presented therein. Briefly, the AFM tip was indented into the surface at regular intervals within a fixed area, and force vs indentation was recorded as a function of x,y coordinate. Force information, which included the characteristic breakthrough force, and Young’s modulus of the layer, was derived from each force curve by offline computer analysis. The calculated force data was then mapped to (x,y) coordinates to form a map of mechanical properties. This technique proved powerful not only to verify the presence of lipids, but also to visualize and identify coexisting lipid phases.

2.3.5 Ultraflat gold via hydrogen flame annealing

In the literature, several methods have been reported to produce ultra-flat gold. Hydrogen flame annealing was chosen for its simplicity. Commercially acquired gold substrates (Arrandee 11 mm x11 mm gold on glass, Werther, Germany) were heated by moving a hydrogen flame emitted from a quartz pipette across the surface of each substrate, until a bright incandescence was achieved. At these temperatures gold surfaces were able to reconstruct, and crystals, which were roughly 100 nm in diameter before annealing grew to about 1 to 2 µm in diameter, with deep grain boundaries separating adjacent grains. The surface of each grain however was a (111) face of the crystal with a distinct pattern of atomic step edges and a mean roughness of about 0.2 nm. This was determined to be suitable for examining lipid bilayers via AFM. These structures were used for milling

- 43 -

nanoaperture arrays, however their thickness affects both the quality of the milled features and the quantity of transmitted light. Before annealing, the backside of each substrate was cleaned of adhesive (used by the manufacturer to secure the substrates for shipping) and used otherwise as is for annealing. The flame annealed substrates were used immediately. In Chapters 3 to 5 all of the gold surfaces used were flame annealed to simplify AFM measurements of the bilayer.

2.3.6 Self assembled monolayers.

Self assembled monolayers (SAM) of carboxylic acid terminated alkanethiols were used to improve the hydrophilicity of the gold surfaces. The –SH group at the end of the chain forms a covalent bond to the gold surface, and at high surface coverage, the COOH groups collectively created a polar surface. Continuity of the surface coverage greatly affected vesicle fusion. The quality of the SAM was measured by contact angle (KSV Instruments LTD, CAM101, Linthicum Heights, MD, USA). Typical static contact angles for MilliQ water on the unmodified, but cleaned commercial substrates were about 30-50 degrees. After flame annealing and SAM formation, contact angles became low enough (< 10 degrees) that contact angle instrument was unable to measure it. Intermediate contact angles of about 20 to 30 degrees indicated incomplete SAM coverage and exhibit patchy bilayer deposition determined by AFM.

Substrates were immersed in an ethanolic solution of 11-mercaptoundecanoic acid (MUDA) (0.1 wt%) immediately after flame annealing, and left to incubate overnight. Following incubation, the substrates were rinsed thoroughly under a stream of ethanol, dried under a stream of N2 and used immediately for vesicle fusion. Using this method, both SAM and lipid bilayers were patchy, with patches roughly about 5 – 10 nm in diameter as determined by AFM imaging of surfaces treated with SAM and surfaces that were treated by vesicle fusion as well.

- 44 -

An improved method of SAM deposition for COOH alkanethiols based on a method reported by Wang et al.92 was used. It was determined that hydrogen bonding between the COOH groups of molecules bound to the surface and those free in solution caused the formation of a partial second layer with the hydrophobic alkane group facing solution. Deposition of the SAM under acidic conditions disrupted hydrogen bonding and thus prevented the partial formation of a second layer. Additionally, the acidic conditions also protonated the acid group, which prevented charge-charge repulsion that could have inhibited the formation of dense COOH monolayers. Following incubation, substrates were rinsed with ethanol made basic by the addition of 10 % (v/v) NaOH, as well as by absolute ethanol. It was found empirically that the use of acetic acid at 5% by volume to acidify the SAM solution produced comparable contact angles to Wang et al.92 where 2% by volume CF3COOH was used. The AFM images of lipid bilayers formed on the latter surface (presented in the next section) revealed very few defects.

All solvents were filtered through PTFE (for organics) or nitrocellulose (for aqueous) filters with 0.2 µm pores.

2.4 Early results and challenges

2.4.1 Introduction

This chapter reviews some early results that have influenced the direction and experimental approach employed in this project. These experiments include several methods of producing nano-hole arrays including top-down and bottom-up approaches, optical characterization of nanohole arrays, and initial studies of lipid bilayers on gold substrates. These initial lipid studies have motivated the use of flame annealed gold,

- 45 -

force indentation as a means of characterization, and lipid phase segregation to help visualize lipid deposition.

2.4.2 Fabrication methods

Several methods for fabricating nanoaperture arrays have been explored, included oft- published focused ion beam milling, electron beam lithography, and a bottom-up diblock copolymer template method. Ultimately, FIB, when it became accessible, became the fabrication method of choice, since it provides direct patterning of the substrate without the use of masks, and generally produces reliable and reproducible patterns. The other approaches to fabrication however, are also worth discussion.

The diblock copolymer template method had the potential to offer larger scale patterning than do top down methods such as FIB and E-beam lithography. The approach involved the use of polystyrene-polymethylmethacrylate (PS-PMMA) dibolock copolymers, which under certain conditions, have been shown to phase segregate and self-assemble into hexagonal arrays of cylinders where the cylinders are composed primarily of one block, while the surrounding matrix is composed of the other. Under conditions where the PS was the minority component by volume, the cylinders were comprised of PS, and the surrounding matrix was comprised primarily of PMMA.

The process of exploiting this self-assembly process to create a template for the deposition of a gold film required several steps that are illustrated schematically in Figure 2-2. A solution of the diblock copolymer was first spin-coated onto a substrate, and allowed to anneal in an acetone vapour environment. The thickness of the layerwas known to be an important determinant of whether the cylinders were oriented vertically. Next, the PMMA matrix was preferentially removed by oxygen plasma treatment, though exposure to UV light followed by rinsing with acetic acid was also employed (not

- 46 -

shown). Both methods selectively removed the matrix component, and left an array of posts. Though it was not accomplished, the deposition of gold onto the patterned substrate was expected to fill the voids between the posts. The PS posts would then be removed by soaking in an organic solvent, leaving behind a gold film with a hexagonal pattern of perforations.

Figure 2-2: Schematic illustration of di-bock co-polymer templating method.

In practice, this process has several limitations. The first is that the patterns and shapes of apertures available in the final film were limited compared to the diversity of structures that are accessible by FIB; a hexagonal array of cylindrical holes was effectively the only accessible structure that was potentially relevant for the sensor applications that were the interest of this project. The second is that, although changing the size and spacing of the cylinders in the pattern was possible by changing the molecular weights of the two polymer blocks, the diameter of the cylinders is inherently tied to the period of the array since only a restricted relative volume fraction of the blocks would result in the formation of cylinders. The third limitation was that the thickness of the film is closely tied to relative size and period of the cylinders. Films of insufficient thickness resulted in the formation of worm-like structures. This required that spin- coating speeds and conditions be re-optimized for different block sizes, so as to achieve the optimal thickness of the film that results in upright cylinders.

- 47 -

In the example below, a PS-PMMA diblock copolymer was spin-cast on a silicon substrate and annealed over night in an acetone vapour environment. Figure 2-3 is an AFM image and cross-section through a defect. The cylindrical domains were about 80 nm in diameter, while the period was about 150 nm. The cross section through the defect revealed the film thickness was about 30 nm, while the difference in height between the domains was about 5nm.

The sample is then treated with an oxygen/nitrogen plasma for 90 s. An AFM scan of a different location on the same substrate after plasma etching (Figure 2-4) revealed that the height of the cylinders approached the thickness of the film as determined by comparing the cross section through a defect and through the features that remain. This indicated that the PMMA was selectively removed by the plasma treatment. Still, the thickness of the film in this case was limited to about 25 nm. A metal film deposited through this mask would also have a maximum thickness of about 25 nm in order to maintain the pattern and allow solvent access to the PS cylinders for their removal. At 25 nm, a gold film would have exhibited significant bulk optical transmission, which would

- 48 -

have contributed strongly to the optical transmission spectrum when compared to the surface-plasmon mediated transmission. The bulk optical transmission band was not of interest in the context of sensing since it was not expected to be sensitive to changes in dielectric constant.

25nm

Another method of fabrication explored was electron beam lithography. Figure 2-5 illustrates the steps involved in fabricating a nanoaperture array. First an quartz slide is coated with a 200 nm thick gold film using thermal or electron-beam assisted vapour deposition, and then a layer of PMMA electron-beam resist is spin coated onto the surface. The electron beam is used to write a pattern onto the surface, and when developed, results in an array of nano-holes in the PMMA. The patterned PMMA acts as a mask through which the pattern is transferred to the gold. The whole substrate is exposed to a bulk-ion beam milling process, which etches the whole sample. The gold is etched where it is exposed through the pattern in the mask. The mask is also etched, albeit at a slower rate. Once the pattern is transferred, the PMMA layer can be removed, leaving the patterned gold substrate.

- 49 -

Figure 2-5: Schematic illustration of electron beam lithography of NAA. Sequence of steps: top row left  right, then bottom row left  right

The E-beam approach yielded relatively large arrays measuring 500 µm x 500 µm. However, some residual E-beam resist remained on the surface even after aggressive measures such as oxygen plasma etching and Piranha treatment were employed. The residual resist layer was observed under AFM to be nearly 200 nm thick, and unsurprisingly, had a profound effect on the transmission spectra of the arrays. Nevertheless, despite this layer, immersing the array into solvents of various dielectric constants still provided a linear shift of SPP bands. The results are shown in Figure 2-6

Even in air, the spectra of the arrays with the residual resist appeared significantly red- shifted from reports in literature and from measurements on bare arrays made by FIB. This implied that material deposited onto the surface of the array but not within the holes, was still able to affect the wavelength of the SPP-BW, even though the most significant field enhancement occurs within the holes. Addition of solvents with various dielectric

- 50 -

constants changed the dielectric material within the holes and also above the remaining resist, and is also able to affect the SPP-BW wavelength. This indicated that changes in the dielectric properties of materials at the surface and within the holes both contribute to the observed shift in the spectrum.

A B

Toluene (2.24)

Hexadecane (2.04)

Isopropanol (1.89)

Acetone (1.84)

Methanol (1.76)

Air (1.00)

2.4.3 Nanoaperture arrays and optical properties

Optical spectra through an array were obtained with the array immersed in a series of different dielectric solvents. The array consisted of square holes 200nm x 200nm in size arranged in a square array with a period of 600 nm, and a total size of 100 µm x 100 µm. AFM and SEM images of typical NAAs are shown in Figure 2-7 A and B respectively.

- 51 -

A B

In Figure 2-8A, the zero-order transmission spectra in various solvents were up-shifted for clarity. Because spectra were referenced against a single square hole of the same total area as the array, the optical transmission percent was measured relative to the total area of the array. The transmission peak centered around 680 nm in air, red shifted in higher dielectric environments. This family of peaks was attributed to the (i,j) = (1,0) Bloch- wave mode of the surface. Other spectral features such as peaks around 520 nm can be attributed to higher order modes (such as (1,1)) of the SPP-BWs, as higher order modes appear at lower wavelengths according to Equation 5. A transmission peak also appeared at 500 nm that persists in different solvents. This was likely due to the volume transmission of light through the Au film. This was visible with the naked eye as a weak transmission through the substrate when held up to a bright light source.

- 52 -

Figure 2-8B plots the (1,0) peak position against the dielectric constant of the solvent. The redshift is linear, and a sensitivity of 256 nm/dielectric constant unit (630 nm/RIU) can be calculated from the slope. This result is comparable to sensitivities around 400 nm/RIU published by other groups using similar NAAs. This, according to Brolo et al.27, is comparable to grating based SPR (300-630 nm/RIU), and slightly better than gold colloids (66.5 nm/RIU), and gold nanoshells (328.5 nm/RIU). This however is much less than SPR via the Kretschmann configuration (3000-8000 nm/RIU). However, the response of the NAA sensors is due to a smaller total number of analyte molecules due to the limited surface area of the array.

2.4.4 Artificial Bilayers on gold

Vesicle fusion experiments on glass and mica substrates have been widely studied, and as such serves as a model for adapting the techniques to gold surfaces. Lipid bilayers were

- 53 -

deposited on glass coverslips by the vesicle fusion method described earlier. Figure 2-9 is the AFM height image of the deposited DPPC bilayer. The accompanying trace through defects in the membrane shows the step height of the bilayer was about 5.4 nm, which is agreeable with the literature value of 4.6 nm.

Initial attempts to adapt this method to gold substrates were largely inconclusive because the RMS roughness of the commercially acquired gold substrates was about 2.5 nm (Figure 2-10A), which was about half the thickness of a bilayer. AFM images (Figure 2-10B) revealed topographic features that may have been patches of lipid. However, in the cross sectional analysis (Figure 2-10C) the thickness the patch was difficult to measure compared to cross sections of the lipids on glass due to the background roughness of the gold.

- 54 -

For purposes of testing the vesicle fusion method, ultraflat gold substrates were created by hydrogen flame annealing. Figure 2-11A shows 10 µm x 10 µm AFM scans of the flamed surfaces that depict the reconstructed morphology consisting of irregularly shaped flat grains of micron sizes, separated by deep grain boundaries. A 880 nm x 880 nm scan of an individual grain’s surface reveals a pattern of equilateral triangles that result from atomic terraces of a gold (111) surface. The RMS roughness of this selected area is 0.15 nm (Figure 2-11b).

- 55 -

These substrates were subsequently modified with COOH-alkanethiol chemistry as described in the previous section, to increase the hydrophilicity of the surface. This proved to be critical to the formation of lipid bilayers on the gold surfaces. Conventional methods of depositing SAMs resulted in poor SAM coverage, and contact angles of about 30 degrees (unmodified gold had contact angles of about 50 degrees). Following the improved procedure of Wang et al.92, contact angles of < 30 degrees were achieved. This however, resulted in patchy SAM formation as depicted by Figure 2-12A and consequently patchy bilayer formation, as shown in Figure 2-12B. The patchy features in this image had heights of about 5 nm, which is the expected thickness of a bilayer of this lipid.

- 56 -

Figure 2-12 (A): AFM topography image (scan size = 1 µm) of flame annealed gold after treatment with 11- mercaptoundecanoic acid using the method of Wang et al.92. Features have heights < 2 nm and lateral sizes of ~ 20 – 30 nm. Surface contact angle: ~ 30 degrees (B): AFM AC-mode images of vesicle fusion on surface similar to (A) (scan size = 1.5 µm). Patches have heights of ~ 5.5 nm relative to substrate, and diameters of ~30 nm.

These findings suggested that a complete SAM would be required to form a complete bilayer on gold. By using acetic acid (5%), and only fresh ethanol in the SAM solution produced contact angles below 15 degress. Vesicle fusion on these surfaces resulted in nearly defect-free bilayer coverage, and thus the thickness of the bilayer could not be measured from the topographic data. On visual inspection, the lipid-covered grain depicted in Figure 2-13 appeared clearly different than a bare grain, with the crystal terraces being mostly obscured.

To verify the presence of an overlayer, the AFM tip was used to indent the surface at various locations. Some indentation curves are shown in Figure 2-14. Whereas hard surfaces like gold tend to show a sharp change in the force at the point of contact, these surfaces showed a gradual increase in the force as the tip made contact, which indicated a soft surface. Though not strictly quantitative, these curves were fit to the Sneddon model for indentation assuming a parabolic tip shape. The fits indicated depths of indentaion of

- 57 -

less than 10 nm, and in most cases between 4 to 6 nm, which supports the assertion that the overlayer was in fact a bilayer.

Figure 2-13: Fluid phase AFM AC-mode image of nearly defect free DPPC bilayer on COOH-alkanethiol modified Au. Alkane thiol monolayer produced by a modified method of ref92. Scan size: 850 nm x 850 nm. Surface contact angle after COOH SAM formation was < 15 degrees.

- 58 -

The Sneddon fit also produces an elastic modulus of indentation as a fitting parameter. In this case modulus values around 3.9 MPa were produced. Modulus values in literature range in order from MPa to GPa. This is likely because the mechanical properties of the bilayer can depend on many factors such as the chain length, temperature, headgroup chemistry, the ionic strength of the buffer solution, and also the presence and concentration of divalent cations. Thus, literature values for the elastic modulus are as varied as the conditions under which they are measured. Furthermore, the Sneddon model assumes that indentation is occurring into an infinite half-space of a homogeneous material. A model93 that addresses this issue by modeling the substrate as a thin layer on an elastic substrate identifies a tendency for force-indentation measurements to be nonlinear (in log-log scale), and as such linear fits tend to overestimate the modulus by a factor of 2 to 4.

2.5 Conclusions

The brief review of literature presented in this chapter provides the general background information and context by which to frame the experiments presented in later chapters.

- 59 -

Reviewed were: (1) Surface plasmon resonance and the role that surface plasmons play in planar biosensing contructs; (2) Lipid bilayers and their current use and potential as functional interface for planar biosensors; (3) Current techniques for depositing lipid bilayers on metal surfaces. These topics are reviewed in more detail in the introductory sections of the chapters that follow to provide better context for the specific experimental contribution presented.

To summarize, NAAs are advantageous for biosensors because the simple linear geometry the optical path makes high levels of simultaneous multiplexing possible. The porous nature of the NAAs makes them particularly suited for lipid-functionalized optical biosensors: In the case of pore-spanning lipid layers, the apertures act as reservoirs that accommodate the large intra- and extracellular domains of some proteins.

Model lipid bilayers are suitable as surface functionalizations in biosensing for 3 main reasons: (1) They have been shown to resist the nonspecific binding of some cells, platelets and proteins such as human IgG and human serum albumin.46-49 (2) They provide the necessary environment natively host membrane proteins but are also adaptable to host lipid-anchored receptors. (3) The lateral mobility of functional molecules in the supported bilayer is similar to physiological cell membranes, allowing for the two dimensional reorientation of surface receptors to optimize geometry for multivalent analyte attachment.

The results of some initial experiments conducted in pursuit of lipid-functionalized sensors were also presented. These results have played a significant role in guiding the methodologies and experiments presented in the chapters that follow, in terms of fabrication, lipid deposition, and characterization.

- 60 -

Focused ion beam milling was chosen as the fabrication method for arrays used in the experiments described in subsequent chapters. In addition to being conveniently available through collaborators, it provided more reliable results while providing flexibility in the shape, and arrangement of apertures: Elements that were difficult to control in the bottom up block co-polymer templating method. This was of particular interest with regards to fabricating a variety of nanoaperture arrays with well-defined geometry, such as the nanohole arrays, nanoslit arrays, and annular aperture arrays employed in Chapter 4. Additionally the method provided the flexibility to mill arrays into flame annealed gold, which enabled the characterization of lipid layers cast on these arrays. E-beam lithography was comparable to FIB in these resepects, but was not as readily available. Furthermore, the initial batch of arrays was found to have a persistent layer of residual e-beam resist, that proved extremely difficult to remove. Being a dielectric surface coating, it also affected the transmission spectra of the arrays.

Initial lipid deposition experiments on gold led to two main contributions. The first was the use of flame annealed gold to better visualize the lipids by AFM: The roughness of thermally deposited gold was too high to clearly observe defects, and would have prevented the identification of coexisting lipid phases. The second was the use of carboxylic acid SAMs and the optimization of that process to allow for a more hydrophilic surface. It was found in later experiments that the SAM modified gold resisted fouling, and therefore retained its hydrophilicity much longer than Piranha or UV cleaned gold. This robustness against fouling enabled the substrates to be cleaned and bilayers redeposited on the same gold substrate multiple times. Because the substrate retained the hydrophilicity required for vesicle fusion, it was not necessary to clean the samples aggressively. This was especially important for patterned substrates, whose features may be affected by treatment with Piranha solution. Gentle treatment of the patterned substrates allowed them to be reused. This treatment also greatly simplified sample preparation and reduced the turn-around time on repeated experiments. Furthermore, when fouling became an issue, redeposition of the SAM restored the hydrophilic properties.

- 61 -

The initial characterization experiments also influenced the main experiments in a number an important ways. It was found that the deposition of a homogeneous defect-free bilayer was difficult to detect by AFM imaging alone. This was addressed by employing force indentation to mechanically characterize the bilayer. Force indentation and the extension of that method to force mapping became a staple tool in the experiments descibed in the chapters that follow, and has been indispensible for verifying, identifying, and characterizing the lipid layer. This was true especially with regards to resolving the coexisting lipid phases. The use of phase-segregated lipids has been identified as important to mimicking some of the aspects of lipid rafts, which are known to play a role in many biological processes.

The remaining chapters detail several experiments that represent important contributions to the current state of the art. Chapter 3 details the deposition of phase-segregated model lipid bilayers on gold. The method of surface modification does not, in principle, restrict the type of lipids that can be deposited, making it possible to deposit model membranes on gold. It also lays the foundation for characterizing these lipids and identifying the lipid phases on gold by force indentation maps. Chapter 4 details experiments that demonstrate the deposition of model lipid membranes on nanohole arrays, nanoslit arrays, and annular aperture arrays, and applies the techniques of Chapter 3 to characterizate and identify the lipid bilayers and phases by imaging and force mapping. Chapter 5 demonstrates the use of a commercial kit to extract natural cell membranes and proteins from cultured cancer cells, as well as their incorporation into model lipids and deposition onto mica surfaces for AFM characterization.

- 62 -

2.6 References

1. Raether, H. Surface plasmons on smooth and rough surfaces and on gratings. 111, (Springer-Verlag: 1988).

2. Stebounova, L. et al. Field localization in very small aperture lasers studied by apertureless near-field microscopy. APPLIED OPTICS 45, 6192-6197 (2006).

3. Chang, S., Gray, S. & Schatz, G. Surface plasmon generation and light transmission by isolated nanoholes and arrays of nanoholes in thin metal films. Opt. Express 13, 3150-3165 (2005).

4. (Ed) Homola, J. Surface plasmon resonance based sensors. 04, (Springer-Verlag: 2006).

5. Link, S. & El-Sayed, M.A. Shape and size dependence of radiative, non-radiative and photothermal properties of gold nanocrystals. International Reviews in Physical Chemistry 19, 409 (2000).

6. Shalaev, V.M., Kawata, S. & NetLibrary, I. Nanophotonics with Surface Plasmons. (Elsevier: 2007).

7. Link, S. & El-Sayed, M.A. Shape and size dependence of radiative, non-radiative and photothermal properties of gold nanocrystals. International Reviews in Physical Chemistry 19, 409 (2000).

8. Oldenburg, S.J., Averitt, R.D., Westcott, S.L. & Halas, N.J. Nanoengineering of optical resonances. Chemical Physics Letters 288, 243-247 (1998).

9. Hao, E. & Schatz, G.C. Electromagnetic fields around silver nanoparticles and dimers. J. Chem. Phys. 120, 357-366 (2004).

10. Chen, J. et al. Gold Nanocages: Bioconjugation and Their Potential Use as Optical Imaging Contrast Agents. Nano Letters 5, 473-477 (2005).

- 63 -

11. Chen, J. et al. Gold nanocages: Engineering their structure for biomedical applications. ADVANCED MATERIALS 17, 2255-2261 (2005).

12. Chen, F. et al. Imaging of optical field confinement in ridge waveguides fabricated on very-small-aperture laser. Appl. Phys. Lett. 83, 3245-3247 (2003).

13. Xu, J., Xu, T., Wang, J. & Tian, Q. Design tips of nanoapertures with strong field enhancement and proposal of novel L-shaped aperture. Opt. Eng. 44, 018001-9 (2005).

14. Lesuffleur, A., Kumar, L.K.S. & Gordon, R. Enhanced second harmonic generation from nanoscale double-hole arrays in a gold film. Appl. Phys. Lett. 88, 261104-3 (2006).

15. Sonnichsen, C. et al. Launching surface plasmons into nanoholes in metal films. Appl. Phys. Lett. 76, 140-142 (2000).

16. Degiron, A., Lezec, H., Yamamoto, N. & Ebbesen, T. Optical transmission properties of a single subwavelength aperture in a real metal. Optics Communications 239, 61- 66 (2004).

17. Murray, W.A., Astilean, S. & Barnes, W.L. Transition from localized surface plasmon resonance to extended surface plasmon-polariton as metallic nanoparticles merge to form a periodic hole array. Phys. Rev. B 69, 165407 (2004).

18. Ebbesen, T.W., Lezec, H.J., Ghaemi, H.F., Thio, T. & Wolff, P. Extraordinary optical transmission through sub-wavelength hole arrays. Nature 391, 667-669 (1998).

19. Taflove, A. Review of the formulation and applications of the finite-difference time- domain method for numerical modeling of electromagnetic wave interactions with arbitrary structures. Wave Motion 10, 547-582 (1988).

20. Ruan, Z. & Qiu, M. Enhanced Transmission through Periodic Arrays of Subwavelength Holes: The Role of Localized Waveguide Resonances. Phys. Rev. Lett. 96, 233901 (2006).

- 64 -

21. Krishnan, A. et al. Evanescently coupled resonance in surface plasmon enhanced transmission. Optics Communications 200, 1-7 (2001).

22. Bethe, H.A. Theory of Diffraction by Small Holes. Phys. Rev. 66, 163 (1944).

23. Koerkamp, K.J.K., Enoch, S., Segerink, F.B., van Hulst, N.F. & Kuipers, L. Strong Influence of Hole Shape on Extraordinary Transmission through Periodic Arrays of Subwavelength Holes. Phys. Rev. Lett. 92, 183901 (2004).

24. Degiron, A. & Ebbesen, T.W. The role of localized surface plasmon modes in the enhanced transmission of periodic subwavelength apertures. J. Opt. A: Pure Appl. Opt 7, S90-S96 (2005).

25. van der Molen, K.L., Segerink, F.B., van Hulst, N.F. & Kuipers, L. Influence of hole size on the extraordinary transmission through subwavelength hole arrays. Appl. Phys. Lett. 85, 4316-4318 (2004).

26. Barnes, W.L., Dereux, A. & Ebbesen, T.W. Surface plasmon subwavelength optics. Nature 424, 824-830 (2003).

27. Brolo, A., Gordon, R., Leathem, B. & Kavanagh, K. Surface Plasmon Sensor Based on the Enhanced Light Transmission through Arrays of Nanoholes in Gold Films. Langmuir 20, 4813-4815 (2004).

28. Jonsson, M.P., Jonsson, P., Dahlin, A.B. & Hook, F. Supported lipid bilayer formation and lipid-membrane-mediated biorecognition reactions studied with a new nanoplasmonic sensor template. Nano Lett. 7, 3462-3468 (2007).

29. Steltenkamp, S. et al. Mechanical Properties of Pore-Spanning Lipid Bilayers Probed by Atomic Force Microscopy. Biophys. J. 91, 217-226 (2006).

30. Ti Tien, H. & Louise Diana, A. Bimolecular lipid membranes: A review and a summary of some recent studies. Chemistry and Physics of Lipids 2, 55-101 (1968).

31. Montal, M. & Mueller, P. Formation of Bimolecular Membranes from Lipid

- 65 -

Monolayers and a Study of Their Electrical Properties. Proceedings of the National Academy of Sciences of the United States of America 69, 3561 -3566 (1972).

32. Mueller, P., Rudin, D.O., Ti Tien, H. & Wescott, W.C. Reconstitution of Cell Membrane Structure in vitro and its Transformation into an Excitable System. Nature 194, 979-980 (1962).

33. Lieser, G. Electron microscopic investigations on free-standing mixed lipid Langmuir-Blodgett-Kuhn monolayers: phase separation and aging process. BBA - Biomembranes 1192, 14-20 (1994).

34. Danelon, C., Perez, J.B., Santschi, C., Brugger, J. & Vogel, H. Cell membranes suspended across nanoaperture arrays. Langmuir 22, 22-25 (2006).

35. Watts, T.H., Brian, A.A., Kappler, J.W., Marrack, P. & McConnell, H.M. Antigen presentation by supported planar membranes containing affinity-purified I-Ad. Proceedings of the National Academy of Sciences of the United States of America 81, (1984).

36. Watts, T.H., Gaub, H.E. & McConnell, H.M. T-cell-mediated association of peptide antigen and major histocompatibility complex protein detected by energy transfer in an evanescent wave-field. Nature 320, 179-181 (1986).

37. Mossman, K.D., Campi, G., Groves, J.T. & Dustin, M.L. Altered TCR Signaling from Geometrically Repatterned Immunological Synapses. Science 310, 1191-1193 (2005).

38. Nam, J., Nair, P.M., Neve, R.M., Gray, J.W. & Groves, J.T. A Fluid Membrane- Based Soluble Ligand-Display System for Live-Cell Assays. Chembiochem 7, 436- 440 (2006).

39. Cooper, M.A. Optical biosensors in drug discovery. Nat Rev Drug Discov 1, 515-528 (2002).

40. Kremer J.J. & Murphy R.M.[1] Kinetics of adsorption of beta-amyloid peptide

- 66 -

Abeta(1-40) to lipid bilayers. Journal of Biochemical and Biophysical Methods 57, 159-169 (2003).

41. Hoffman, T.L., Canziani, G., Jia, L., Rucker, J. & Doms, R.W. A biosensor assay for studying ligand-membrane receptor interactions: Binding of antibodies and HIV-1 Env to chemokine receptors. Proceedings of the National Academy of Sciences of the United States of America 97, 11215 -11220 (2000).

42. Reimhult, E. & Kumar, K. Membrane biosensor platforms using nano- and microporous supports. Trends Biotechnol 26, 82-89 (2008).

43. Nikolelis, D.P. & Krull, U.J. Establishment and control of artificial ion-conductive zones for lipid membrane biosensor development. Analytica Chimica Acta 257, 239- 245 (1992).

44. Nikolelis, D.P., Hianik, T. & Krull, U.J. Biosensors based on thin lipid films and liposomes. Electroanalysis 11, 7-15 (1999).

45. Yu, J. et al. Polyvalent interactions of HIV-gp120 protein and nanostructures of carbohydrate ligands. NanoBioTechnology 1, 201-210 (2005).

46. Kam, L. & Boxer, S.G. Cell adhesion to protein-micropatterned-supported lipid bilayer membranes. J. Biomed. Mater. Res 55, 487-495 (2001).

47. Andersson, A., Glasmästar, K., Sutherland, D., Lidberg, U. & Kasemo, B. Cell adhesion on supported lipid bilayers. Journal of Biomedical Materials Research 64A, 622-629 (2003).

48. Glasmästar, K., Larsson, C., Höök, F. & Kasemo, B. Protein Adsorption on Supported Phospholipid Bilayers. Journal of Colloid and Interface Science 246, 40- 47 (2002).

49. Kono, K., Ito, Y., Kimura, S. & Imanishi, Y. Platelet adhesion on to polyamide microcapsules coated with lipid bilayer membrane. Biomaterials 10, 455-461 (1989).

- 67 -

50. Ruiz-Argüello, M.B., Veiga, M.P., Arrondo, J.L.R., Goñi, F.M. & Alonso, A. Sphingomyelinase cleavage of sphingomyelin in pure and mixed lipid membranes. Influence of the physical state of the sphingolipid. Chem. Phys. Lipids 114, 11-20 (2002).

51. Cronan, J.E. & Gelmann, E.P. Physical properties of membrane lipids: biological relevance and regulation. Bacteriol. Rev. 39, 232-256 (1975).

52. Brown, D.A. & London, E. Functions of Lipid Rafts In Biological Membranes. Annu. Rev. Cell. Dev. Biol. 14, 111-136 (1998).

53. Brown, D.A. Lipid rafts, detergent-resistant membranes, and raft targeting signals. Physiology (Bethesda) 21, 430-439 (2006).

54. Chachaty, C., Rainteau, D., Tessier, C., Quinn, P.J. & Wolf, C. Building Up of the Liquid-Ordered Phase Formed by Sphingomyelin and Cholesterol. Biophys. J. 88, 4032-4044 (2005).

55. Xu, X. & London, E. The Effect of Sterol Structure on Membrane Lipid Domains Reveals How Cholesterol Can Induce Lipid Domain Formation†. Biochemistry 39, 843-849 (2000).

56. Spector, A. & Yorek, M. Membrane lipid composition and cellular function. J. Lipid Res. 26, 1015-1035 (1985).

57. Simons, K. & Vaz, W.L. Model Systems, Lipid Rafts, and Cell Membranes. Annu. Rev. Biophys. Biomol. Struct. 33, 269-295 (2004).

58. Brown, D. & London, E. Structure and Origin of Ordered Lipid Domains in Biological Membranes. J. Membr. Biol. 164, 103-114 (1998).

59. Cheng, P.C. et al. Floating the raft hypothesis: the roles of lipid rafts in B cell antigen receptor function. Seminars in Immunology 13, 107-114 (2001).

60. Simons, K. & Toomre, D. Lipid rafts and signal transduction. Nat Rev Mol Cell Biol

- 68 -

1, 31-39 (2000).

61. London, E. & Brown, D. Insolubility of lipids in Triton X-100: physical origin and relationship to sphingolipid/cholesterol membrane domains (rafts). BBA- Biomembranes 1508, 182-195 (2000).

62. Hancock, J.F. Lipid rafts: contentious only from simplistic standpoints. Nat Rev Mol Cell Biol 7, 456-462 (2006).

63. Lichtenberg, D., Goñi, F.M. & Heerklotz, H. Detergent-resistant membranes should not be identified with membrane rafts. Trends in Biochemical Sciences 30, 430-436 (2005).

64. Johnston, L.J. Nanoscale Imaging of Domains in Supported Lipid Membranes. Langmuir 23, 5886-5895 (2007).

65. Rinia, H.A. & de Kruijff, B. Imaging domains in model membranes with atomic force microscopy. FEBS Lett. 504, 194-199 (2001).

66. Devanathan, S., Salamon, Z., Lindblom, G., Gröbner, G. & Tollin, G. Effects of sphingomyelin, cholesterol and zinc ions on the binding, insertion and aggregation of the amyloid ABeta(1-40) peptide in solid-supported lipid bilayers. FEBS Journal 273, 1389-1402 (2006).

67. Campbell, S.M., Crowe, S.M. & Mak, J. Lipid rafts and HIV-1: from viral entry to assembly of progeny virions. J. Clin. Virol. 22, 217-227 (2001).

68. Veatch, S. & Keller, S. Miscibility Phase Diagrams of Giant Vesicles Containing Sphingomyelin. Phys. Rev. Lett. 94, (2005).

69. Almeida, R.F.D., Aleks, Fedorov, R. & Prieto, M. Sphingomyelin/Phosphatidylcholine/Cholesterol Phase Diagram: Boundaries and Composition of Lipid Rafts. Biophysical Journal doi:10.1016/S0006-3495(03)74664- 5

- 69 -

70. Quinn, P.J. & Wolf, C. The liquid-ordered phase in membranes. Biochimica et Biophysica Acta (BBA) - Biomembranes doi:10.1016/j.bbamem.2008.08.005

71. Tessier, C., Staneva, G., Trugnan, G., Wolf, C. & Nuss, P. Biointerfaces : Liquid– liquid immiscibility under non-equilibrium conditions in a model membrane: An X- ray synchrotron study. Colloids and Surfaces B doi:10.1016/j.colsurfb.2009.07.033

72. Veatch, S.L. & Keller, S.L. Seeing spots: Complex phase behavior in simple membranes. Biochimica et Biophysica Acta (BBA) - Molecular Cell Research 1746, 172-185 (2005).

73. Sullan, R.M.A., Li, J.K. & Zou, S. Direct Correlation of Structures and Nanomechanical Properties of Multicomponent Lipid Bilayers. Langmuir 25, 7471- 7477 (2009).

74. Johnson, J.M., Ha, T., Chu, S. & Boxer, S.G. Early Steps of Supported Bilayer Formation Probed by Single Vesicle Fluorescence Assays. Biophys. J. 83, 3371-3379 (2002).

75. Geertsma, E.R., Nik Mahmood, N.A.B., Schuurman-Wolters, G.K. & Poolman, B. Membrane reconstitution of ABC transporters and assays of translocator function. Nat Protoc 3, 256-266 (2008).

76. Slade, A., Luh, J., Ho, S. & Yip, C.M. Single molecule imaging of supported planar lipid bilayer--reconstituted human insulin receptors by in situ scanning probe microscopy. J. Struct. Biol. 137, 283-291 (2002).

77. Raguse, B. et al. Tethered Lipid Bilayer Membranes: Formation and Ionic Reservoir Characterization. Langmuir 14, 648-659 (1998).

78. Lahiri, J., Kalal, P., Frutos, A., Jonas, S. & Schaeffler, R. Method for Fabricating Supported Bilayer Lipid Membranes on Gold. Langmuir 16, 7805-7810 (2000).

79. Stelzle, M., Weissmueller, G. & Sackmann, E. On the application of supported bilayers as receptive layers for biosensors with electrical detection. The Journal of

- 70 -

Physical Chemistry 97, 2974-2981 (1993).

80. Roberts, G.G. Langmuir-Blodgett Films. (Plenum Presss: New York, 1990).

81. Ulman, A. An Introduction to Ultrathin Organic Films: From Langmuir-Blodgett to Self-Assembly. (Boston, 1991).

82. Surrey, T. & Jähnig, F. Refolding and oriented insertion of a membrane protein into a lipid bilayer. Proceedings of the National Academy of Sciences of the United States of America 89, (1992).

83. Thompson, N.L., Brian, A.A. & McConnell, H.M. Covalent linkage of a synthetic peptide to a fluorescent phospholipid and its incorporation into supported phospholipid monolayers. Biochimica et Biophysica Acta (BBA) - Biomembranes 772, 10-19 (1984).

84. Chen, M., Li, M., Brosseau, C.L. & Lipkowski, J. AFM Studies of the Effect of Temperature and Electric Field on the Structure of a DMPC−Cholesterol Bilayer Supported on a Au(111) Electrode Surface. Langmuir 25, 1028-1037 (2009).

85. Li, M. et al. AFM Studies of Solid-Supported Lipid Bilayers Formed at a Au(111) Electrode Surface Using Vesicle Fusion and a Combination of Langmuir−Blodgett and Langmuir−Schaefer Techniques. Langmuir 24, 10313-10323 (2008).

86. Zailer I., Frost J.E.F., Chabasseur-Molyneux V., Ford C.J.B. & Pepper M. Crosslinked PMMA as a high-resolution negative resist for electron beam lithography and applications for physics of low-dimensional structures. Semiconductor Science and Technology 11, 1235-1238 (1996).

87. Shin, K. et al. A Simple Route to Metal Nanodots and Nanoporous Metal Films. Nano Lett. 2, 933-936 (2002).

88. Cha, T., Guo, A. & Zhu, X. Formation of Supported Phospholipid Bilayers on Molecular Surfaces: Role of Surface Charge Density and Electrostatic Interaction. Biophys. J. 90, 1270-1274 (2006).

- 71 -

89. Yuan, C., Furlong, J., Burgos, P. & Johnston, L.J. The Size of Lipid Rafts: An Atomic Force Microscopy Study of Ganglioside GM1 Domains in Sphingomyelin/DOPC/Cholesterol Membranes. Biophys. J. 82, 2526-2535 (2002).

90. Ross, M., Steinem, C., Galla, H.J. & Janshoff, A. Visualization of Chemical and Physical Properties of Calcium-Induced Domains in DPPC/DPPS Langmuir-Blodgett Layers. Biochim. Biophys. Acta 352, 10 (1974).

91. Xie, A.F., Yamada, R., Gewirth, A.A. & Granick, S. Materials Science of the Gel to Fluid Phase Transition in a Supported Phospholipid Bilayer. Physical Review Letters 89, 246103 (2002).

92. Wang, H., Chen, S., Li, L. & Jiang, S. Improved method for the preparation of carboxylic acid and amine terminated self-assembled monolayers of alkanethiolates. Langmuir 21, 2633-6 (2005).

93. Akhremitchev, B.B. & Walker, G.C. Finite sample thickness effects on elasticity determination using atomic force microscopy. Langmuir 15, 5630-5634 (1999).

- 72 -

Chapter 3: Phase-segregation of model lipid bilayers on Au by AFM imaging and force mapping.

Reproduced in part from Ip, S.; Li, J.K.; Walker, G. C. “Phase-segregation of untethered zwitterionic model lipid bilayers observed on mercaptoundecanoic acid-modified Au surfaces by AFM imaging and force mapping”, Langmuir, 2010, DOI: 10.1021/la100605t. Copyright © 2010 American Chemical Society.

J.K. Li wrote the software code foundation used in the analysis of force map data, which included procedures for file input/output, automated identification of basic force-features, automated Sneddon-model fitting, basic histogram generation. G.C. Walker provided supervisory guidance, contributed to background research, and contributed to the interpretation of the results. S. Ip performed all experiments, modified the analysis code with alternative procedures for automated force-feature identification, alternative fits of the data to different models, and for histogram analysis and data separation, performed data analysis, performed background research, and contributed to the interpretation of results.

Planar supported lipid bilayers (SLBs) have been widely studied as model cell membranes because they are accessible to a variety of surface-analytic techniques that have contributed to identifying membrane lateral organization and heterogeneity as fundamentally important to a range of cellular functions and diseases. This chapter describles the formation of phase segregated dioleoylphosphatidylcholine (DOPC)/sphingomyelin/cholesterol bilayers on mercaptoundecanoic acid-modified (111) gold by spontaneous fusion of unilamellar vesicles, without the use of charged, or chemically modified headgroups. The liquid-ordered (lo) and liquid-disordered (ld)

- 73 -

domains were observed by atomic force microscopy height and phase imaging. Furthermore, the mechanical properties of the bilayer were characterized by force- indentation maps. Fits of force indentation to Sneddon mechanics yielded average apparent Young’s moduli of the lo and ld phases of 100 ± 2 MPa and 59.8 ± 0.9 MPa respectively. The results were compared to the same lipid membrane system formed on mica with reasonable agreement, though modulus values on mica appeared higher. Semi- quantitative comparisons suggest that the mechanical properties of the lo phase were dominated by intermolecular van der Waals forces, while those of the fluid ld phase, with relatively weak van der Waals forces were influenced appreciably by differences in surface charge density between the two substrates, which manifested as a difference in apparent Poisson’s ratios.

3.1 Introduction

Planar supported lipid bilayers (SLB) have been widely studied as model cell membranes because they are accessible to a variety of surface-analytic techniques, such as scanned- probe microscopy1-6, fluorescence microscopy7-9, electrical impedance analysis10,11, time- of-flight secondary ion mass spectrometry12, and neutron scattering13 that allow detailed characterization of membrane morphology. Such characterization has been widely employed to study model lipid systems that phases-segregate into coexisting liquid- ordered and liquid-disordered phases. The coexistence of different lipid phases is of fundamental importance to the raft hypothesis for cellular membranes, which over the last 15 years, has evolved as a model for the lateral organization and heterogeneity of cell membranes, which has been found to play a role in processes such as cell signaling14-18, and also mechanisms of diseases such as Alzheimer’s19,20 and HIV/AIDS21,22.

Furthermore, studies on model systems show that inhomogeneous receptor distribution on surfaces at length scales comparable to rafts, profoundly affects the morphology of cells bound to these surfaces23-26. Thus, the role of phase segregation, and more specifically its

- 74 -

effect on receptor distribution and membrane interactions with cells, drugs, and viruses may be better understood by further characterization using diverse methods.

Specifically, surface plasmon resonance (SPR) and related plasmonic techniques are suitable for studying the kinetics of interactions between soluble ligands and membrane- bound proteins or the membrane itself27-30. Additionally, lipid bilayers have been identified as a possible platform for surface functionalization in biosensor applications29,31-33. The use and relevance of SPR and related biosensing platforms to this problem would be improved by the ability to spontaneously form bilayers of model, raft- forming lipid mixtures capable of inhomogeneous receptor display, such as those containing dioleoylphosphatidyl choline (DOPC), sphingomyelin and cholesterol, on noble-metal surfaces required for SPR. Instead, most studies utilize lipid membrane systems with charged, or chemically modified headgroups, unlike most found in nature, due to limitations in producing lipid bilayers on metal surfaces11. Furthermore, phase segregation and lateral heterogeneity, has not been adequately demonstrated on metals. Phase segregated lipids on metals may also be relevant to other characterizations methods that require conductive surfaces, such as scanning tunneling microscopy, which has already been employed to study lipids34-36.

On glass or mica surfaces, preparation of raft-forming model membranes is routine, and widely reported1,2,37-41. The vesicle fusion method, for example is straightforward: SLBs are self-assembled when unilamellar vesicles adsorb to a hydrophilic surface and spontaneously rupture, fusing together to form a laterally mobile bilayer 42-46,41. On metallic surfaces suitable for SPR or similar analytic techniques, bilayer formation (by vesicle fusion or otherwise) often requires a specific headgroup charge or chemical modification11 to promote a favorable interaction with the metal surface. Methods of spontaneous SLB formation on metals are diverse: (1) Lipid monolayers can be deposited on top of a hydrophobic alkanethiol monolayer preformed on the metal surface to produce a hybrid bilayer27,47. This precludes the incorporation of transmembrane proteins

- 75 -

into the layer that require hydration of both intracellular and extracellular domains. (2) The inner leaflet of the bilayer is anchored to the gold surface through proteins48. The anchoring at high concentrations reduces the lateral mobility of the SLB49. (3) SLBs containing lipids with cationic or anionic headgroups are formed spontaneously on gold surfaces with anionic or cationic surface modifications, respectively10,50. Strong electrostatic interactions, for example between cationic lipids and negatively charged mica, have been shown to greatly reduce lateral mobility and result in abnormal phase boundaries51. (4) Vesicles can spontaneously fuse on the end of a gold wire freshly cut in a suspension of vesicles52. This system however, is not suitable for SPR sensing. (5) Long chemisorbed tether molecules have been employed to produce bilayers with a larger aqueous compartment between the inner leaflet and the substrate. In principle, this method is suitable for the incorporation of transmembrane proteins with considerable cytosolic or extracellular domains53,54 but to the knowledge of this author, phase segregation has not been demonstrated on such a system. However, phase segregation has been demonstrated on similar systems using glass and quartz substrates55,56, and considering the degree to which the lipids are decoupled from the substrate, such a strategy could potentially produce phase-segregated bilayers on gold as well. The present study demonstrates the phase segregation of untethered ternary lipid bilayers on gold, which are more directly analogous to the widely studied ternary lipid bilayers on mica.

- 76 -

Figure 3-1: Typical AFM force indentation curve of the ternary lipid bilayer on gold (Top) featuring a breakthrough event characteristic of these bilayers. Features of the indentation curve are labeled A through D. The four-panel diagram (not to scale) represents the interaction between the AFM tip and the lipid bilayer during the indentation process. Letters A-D refer to labeled features on the example indentation curve. The bilayer constituents (DOPC, egg sphingomyelin, and cholesterol) are drawn to represent the phasesegregation of the DOPC from the egg sphingomyelin into ld and lo phases respectively. The bilayer is supported by the mercaptoundecanoic acid (MUDA) modified gold substrate through an assumed hydration layer.

- 77 -

Specifically, the formation of biologically relevant ternary lipid bilayers on mercaptoundecanoic acid-modified (111) gold by spontaneous fusion of unilamellar vesicles, without the use of lipids bearing charged, or chemically modified headgroups is reported. Furthermore, the phase segregation of the lipid into sphingomyelin-enriched liquid-ordered domains (lo) and DOPC-enriched liquid-disordered domains (ld) was observed by atomic force microscopy. The bilayer was also characterized by force- indentation maps, and compared the identical lipid membrane system formed on mica. Good agreement was found in terms of the magnitudes of the apparent Young’s moduli of the two phases. While there have been other studies that have demonstrated the deposition of neutral phospholipid bilayers on Au(111) surfaces57,58, and phospholipid/cholesterol mixtures58,59 the lipid systems employed, as in many cases, are not known to phase segregate, and are not, a priori, suitable for studying the functional consequences of lipid domain formation. Furthermore, the imaging and mechanical characterization of lipid phases on gold supports has not, to the knowledge of this author, been examined. To achieve this, two limitations were overcome: (1) Bare-gold surfaces are prone to contamination, which quickly reduces the hydrophilicity beyond a range where vesicle fusion can occur, and (2) direct observation of lipid phase segregation by scanned-probe microscopy is difficult on thermally deposited gold because the roughness is on the length scale of the lipid bilayer thickness. To address the first the gold is modified with mercaptoundecanoic acid (MUDA), which makes the surface negatively charged much like freshly cleaved mica, and remains hydrophilic much longer than bare gold. To address the second, the gold was flame annealed prior to MUDA deposition, to produce large (several square microns) single-crystal faces with RMS roughnesses that are an order of magnitude lower than as-deposited gold. These methods can be extended to prepare bilayers on generic gold surfaces.

- 78 -

3.2 Materials and Methods

Gold Surfaces: Gold-coated glass substrates (“11x11”) were purchased from Arrandee (Werther, Germany). The backside of each substrate was cleaned using lab tissue and acetone to remove adhesive applied by the manufacturer for transport. Hydrogen flame emitted from a quartz pipette was used to heat the substrate until incandescence, producing gold grains several square-microns in area with (111) surfaces exposed. RMS roughness of each grain was ~0.15 nm as evaluated by atomic force microscopy (data not shown).

Hydrophilic self-assembled monolayer: 11-mercaptoundecanoic acid (MUDA) monolayers were chemisorbed on the flame annealed gold surfaces at very high surface densities using a modified version of Wang et al60. Briefly, the substrate was immersed in a 1 mM solution of MUDA (95%, Sigma-Aldrich Canada Ltd, Oakville, ON, Canada), and 5% (V/V) glacial acetic acid in ethanol immediately after flame annealing and left overnight. Following incubation, substrates were rinsed with ~100 mL of solution containing 10% (V/V) amonium hydroxide and 90% (V/V) ethanol, then with ~300 mL ethanol, then copious amounts of 18 MΩ-cm H2O (Barnstead EASYpure II, Barstead/Thermo Fisher Scientific, Napean, ON, Canada). The substrate was then dried under a stream of N2 and epoxied to a clean 30mm-diameter round glass slide and left to cure in a laminar flow hood for 12 hrs. All ethanol was absolute, filtered through poly(tetrafluoroethylene) filters with pore size = 0.2µm (Whatman Inc, Piscataway, NJ, USA) using a vacuum filtration flask (VWR, Mississauga, ON, Canada). The MUDA modified Au substrates maintained their hydrophilicity longer than unmodified gold. Furthermore, after 12 hours of exposure to air in the laminar flow hood (while the epoxy cured), the substrate would no longer hold a continuous sheet of water, but a thorough rinse with the ammonium hydroxide/ethanol solution, followed by a rinse with ethanol, then with water restored the substrate’s ability to hold a thin sheet of water at its surface.

- 79 -

Mica substrates: Grade V1 muscovite mica discs (Ted Pella Inc, Product No. 50, Redding, CA, USA) were epoxied to 30mm-diameter round glass discs and cured over night in a lamellar flow hood. Immediately before lipid deposition, the sample was mounted in the fluid cell, and hydrated with 50°C 18MΩ-cm H2O. The mica was then cleaved under water using sharp tweezers, and the vesicle and salt solutions were applied immediately.

DOPC/egg sphingomyelin/cholesterol unilamellar vesicles: 1,2-dioleoyl-sn-glycero-3- phosphocholine, egg sphingomyelin, and ovine cholesterol (Avanti Polar Lipids, Alabaster, AL, USA) were mixed at 2:2:1 molar ratio at 1mg/mL in chloroform (reagent grade, Sigma-Aldrich; filtered with 0.2µm PTFE filters from VWR). Although fluorescence was not studied in this work, all alliquots contained fluorescent NBD- labeled phosphoethanolamine (1-palmitoyl-2-{12-[(7-nitro-2-1,3-benzoxadiazol-4- yl)amino]lauroyl}-sn-glycero-3-phosphoethanolamine) (Avanti Polar Lipids) at 0.2 mol%. 1 mL aliquots were dried under N2 stream overnight, sealed, and stored at -20˚C until use. Dried lipid cake was rehydrated with phosphate buffered saline to 1mg/mL at 60 °C, and vortexed to form multilamellar vesicles. The multilamellar vesicle suspension was refined to unilamellar vesicles by extrusion (Mini Extruder, Avanti Polar Lipids) at 60 °C using 0.1 µm polycarbonate filters (Whatman, Piscataway, NJ, USA) following the manufacturer’s directions. Each suspension was passed through the filter a minimum of 15 times, and always an odd number of times.

Vesicle fusion: In an AFM fluid cell, the mounted substrate was incubated in a solution containing 100 µg of lipid in unilamellar-vesicle form and CaCl2 at a final concentration of 10mM for 30 min. For the gold substrate, incubation occurred at room temperature, whereas for the mica substrate, incubation occurred at ~50°C. Afterwards, the substrate was rinsed with ~200 mL of 18MΩ-cm H2O without allowing the substrate to dry or be exposed to air.

- 80 -

Atomic force microscopy (AFM) imaging: Tapping mode AFM was performed in 18MΩ- cm H2O using an Asylum Research MFP3D (Asylum Research, Santa Barbara, CA, USA) equipped with magnetic tip oscillation (iDrive). An oscillating AC current is passed from one base of a triangular, metalized cantilever to the other creating an oscillating magnetic field, which causes the cantilever to oscillate in the static magnetic field supplied by a permanent magnet built into the cantilever holder. This produces amplitude and phase signals that are less noisy compared to traditional water-immersed AC-mode AFM, where cantilever oscillation is coupled strongly to resonances of the bulk fluid. It also necessitates the use of specialized gold-coated probes (nominal spring constant: 0.09 N/m ). The round glass slide to which the sample was epoxied, was mounted in a closed fluid cell (Asylum Research) according to the manufacturer’s instructions.

Before immersion in the fluid cell, the tip was equilibrated in air for 10 minutes, and cantilever spring constant calibrated by thermal method after indentation into freshly cleaved mica. Following immersion of the tip into the fluid cell, the system was allowed to equilibrate in situ with the sample and water for 1 hour prior to imaging.

Force mapping and automated analysis: Force map data was collected using instrument’s included software. Indentations into the sample are made at even intervals across a fixed area, which allows information from each force curve to be mapped back to an (x,y) coordinate to produce images from the force data. Each indentation is triggered using a relative force value to ensure the same maximum indentation force is reached for each point. Force curves were analyzed offline using custom software written in IgorPRO 6.02 (Wavemetrics. Portland, OR, USA). A typical force indentation curve is shown in Figure 3-1 with an illustration of the interaction between the AFM tip and the bilayer. The contact point, and rupture events are found by determining the peaks in the second

- 81 -

derivative of smoothed force vs. Z curves. These data are used to automatically identify the presence of a breakthrough event, and to record the rupture force. Each curve was also fit by the Sneddon model to obtain a Young’s modulus value.

Elastic indentation model fitting: Force vs. indentation data between the contact point and half-way to the breakthough point were fit by the Sneddon model, with the elastic modulus as the fitting parameter, and modified to account for differences between the physics of the model and the experiment. The Sneddon model assumes that the sample forms an infinite half-space below the tip, and thus differs from the experiment in which a thin soft material sits above a stiff substrate. Modifications to the model were made using two methods; the first according to Akhremitchev et al., and the second according to Dimitriadis et al. Neither correction method proved satisfactory, though indentation data collected on gold- and mica-supported bilayers can be compared when the same model is used to analyze each data set.

The Sneddon model relates the applied force (F) to the depth of indentation (δ) through experimental parameters such as probe tip geometry, and through material properties of the sample including the Young’s modulus (E), and the Poisson’s ratio (ν). For a parabaloid tip shape at low indentation depths, the model is mathematically identical to the Hertz model for a spherical indenter:

Equation 6

E was calculated as a fitting parameter assuming ν = 0.33, and R= 25 nm.

- 82 -

Corrections to the model were made according to Akhremitchev et al.61, where the correction factor to the fitted modulus can be obtained graphically from a normalized force-correction vs. indentation curve. The correction factor depends on the ratio of Lamé coefficients between the thin material and the underlying substrate, as well as the normalized maximum indentation given by Equations 7 and 8 respectively:

Equation 7

Equation 8

In Equation 7 the subscripts 1 and 2 refer to the bilayer and the substrate respectively. In

Equation 8, parabaloid tip geometry is assumed and δmax is the maximum indentation of the fit, R is the radius of curvature, and h is the thickness of the bilayer, which is assumed to be 5 nm. The application of this correction is discussed in the following section.

3.3 Results and discussion

Unlike many prior methods of producing lipid bilayers on gold surfaces, this method does not require head-group modified phospholipids. Instead, widely studied DEC221 bilayers, which are known to phase segregate into raft-like domains1,37-40,62, were self- assembled on MUDA-modified flame-annealed gold. The same bilayers were cast onto mica for comparison on the basis of morphology and mechanical properties. These results were compared not only to each other, but also to the results of others who have cast these lipids on mica1,40. The primary intent however, is not to quantify the true modulus of the bilayer as others have attempted, since doing so would require physically decoupling the bilayer from the solid support63, but rather to verify the coexistence of two lipid phases on gold, and to characterize the bilayer in supported form. Briefly, it was confirmed that a bilayer had formed on the gold substrate, and although there were

- 83 -

differences in morphology of the domains, the modulus values of the domains were similar to those measured on mica.

3.3.1 Vesicle fusion requires hydrophilic surface:

The ability to form a DEC221 bilayer on the MUDA modified gold, and the greater difficulty to do so on unmodified gold can be explained by differences in the hydrophilicity of the two substrates. The fusion of vesicles at a surface to form a supported bilayer can occur only on surfaces that are sufficiently hydrophilic. Otherwise, vesicles that settle on the surface tend to remain vesicular. On the other hand, when vesicles adsorb to strongly hydrophilic surfaces, they spread to maximize contact with the surface while causing increasing surface tension on the vesicle wall, and eventually rupture to form the supported bilayer. Adsorbed vesicles containing lipids with zwitterionic headgroups, rupture readily on freshly cleaved mica, which is negatively charged, provided the buffer environment does not contain large molecular counter ions that shield the surface charge45. The ability of adsorbed vesicles to spread on a surface can be estimated by the static contact angle made by a droplet of water on the surface. On freshly cleaved mica, the static contact angle of a drop of water is typically < 5 degrees. The unmodified gold substrates can have static contact angles < 10 degrees immediately after cleaning with oxygen plasma, UV-ozone, or Piranha solution but these angles are short-lived due to rapid contamination. The MUDA modified gold, either flame-annealed or not, showed static water contact angles < 10 degrees (data not shown) and remains hydrophilic longer (perhaps by resisting fouling), and can be restored easily by rinsing with the ammonium hydroxide/ethanol solution, followed by rinsing with ethanol, then water. In fact, the same MUDA-modified substrate can be reused several times by thoroughly rinsing in this fashion. And like mica, a high-density self-assembled monolayer of MUDA has a net negative surface charge at neutral pH.

- 84 -

3.3.2 AFM imaging and lipid phase morphology on gold:

Because oriented mica substrates are atomically smooth, they are ideal supports for AFM studies of the phase behavior of lipid bilayers, which typically exhibit topographic contrast between phases of ~1 nm, depending on temperature, and choice of lipids. Thermally deposited gold films on glass substrates, though useful for SPR and related techniques, were not found to be suitable as substrates for AFM studies of lipid bilayers because the root-mean-square (RMS) roughness of a polycrystalline gold film can be as high as 2.5 nm, which consequently dominates topographic images. The large (111) crystal faces of gold formed by flame annealing can have RMS roughnesses < 1 nm when measured over single grains. Though better than as-deposited gold films, flame annealed films can still contain topographic features that arise from the step-edges between (111) terraces on the crystal, which can range in height from 0.5 to 2 nm. Typical AFM height and phase images of MUDA-modified flame annealed gold are shown in Figure 2A and B respectively. The lines that formed triangular shapes in the images are the crystal terraces of the (111) gold surface typical of flame-annealed gold surfaces64. Although the topographic features of (111) Au complicated the interpretation of AFM height images relative to mica, such a treatment represents a great improvement over thermally deposited gold. Furthermore, the effect of topography in phase images and force maps was less pronounced, and consequently, these results are the focus of this discussion. The topographic images, and the effect of the substrate on phase segregation are also discussed.

An example of a DEC221 bilayer spontaneously formed on MUDA-modified, flame- annealed gold can be seen by AFM height and phase contrast in Figure 3-2C and D respectively. Note that the region scanned in Figure 3-2C and D is not the same region shown in Figure 3-2A and B. The phase segregation of the bilayer into lo and ld domains was discernable both by height and phase contrast, though the step edges between (111) Au terraces made the distinction in the topography difficult to perceive.

- 85 -

What complicated the interpretation of the topography was that the boundaries of lipid phase domains tended to align with the edges of the terraces in a many cases. It was hypothesized that the presence of topographic features on the scale of 1 to 2 nm (similar to the height mismatch expected between the lo and ld phases) can pin the migration of the phase boundaries. This consequently also affects the shape of the domains compared to how they appear on mica, which does not exhibit terracing. The coincidence of lipid phase boundaries with terrace edges had either an additive or negating effect on the perceived height difference. Line-traces of the height image showed height differences between the lo and ld phases which ranged from 0.9-1.2 nm, which is comparable to accepted height differences on mica of about 0.8 nm, if the height differences between terraces are considered. Thus, it was concluded that alternate mechanisms for contrast between the expected lipid phases was more reliable.

The morphology was more easily discernable in the phase image than in the height image. The lo phase appeared as the light grey features in the phase image, and the ld phase appeared as the dark grey regions. The coexistence of two lipid phases was verified by the breakthrough force map, which is discussed in the next section. Briefly, the presence of a breakthrough force event over most of the image area indicates good coverage by the bilayer. The map of the breakthrough force magnitudes reproduced the features in the phase image.

3.3.3 AFM force map of DEC221 on gold:

Force indentation curves of the bilayers on gold were collected over a square grid in a selected area, and information such as the breakthrough force and Young’s modulus were mapped to an (x,y) coordinate. Force indentation data was collected in a 64 x 48 grid over the same area as the AFM height image (Figure 3-2C), where Figure 3-2E and F are maps of the Young’s modulus and breakthrough force values extracted from the indentation data.

- 86 -

The breakthrough event is characteristic of an indentation curve collected over a bilayer. It occurs when a load is applied to the bilayer that is great enough to temporarily rupture the bilayer. At the point of rupture, the restoring force supplied by the bilayer is briefly lost, which is manifests in the force curve as a sudden, drop in measured force (Figure 3-1). It was found that this discontinuity coincided with a peak in the second derivative of the force vs. Z curve, which was detected by the algorithm, and the breakthrough force was taken as the force value just before the rupture. In Figure 3-2F, each point represents the magnitude of the force at the point where the tip ruptured the bilayer, or if no rupture event was detected, the maximum force was recorded but not displayed on the map.

Contrast between the lo and ld phases is clearly seen in Figure 3-2F, where features agree well with the AFM phase image (Figure 3-2D). Here, the force required to break through the lo phase was higher than the force required to break through the ld phase. This is expected because in the sphingomyelin-rich lo phase, the saturated acyl chains of the sphingomyelin allow closer packing of lipids than in the unsaturated DOPC-rich ld phase65,66, therefore leading to stronger interactions between the chains and headgroups of neighboring molecules. Another manifestation of this packing effect is the higher Tm of saturated lipids compared to unsaturated lipids67,68.

Each point on the modulus map (Figure 3-2E) represents a single force indentation curve where the Young’s modulus was determined by a fit of the force vs. separation curve between the contact point, and halfway to the breakthrough point using the Sneddon model. If no breakthrough point was detected for a given curve, the model was fit to the first half of the entire indentation. Contrast between the two phases exists in the Young’s modulus map as well, but is less apparent than in the breakthrough force map.

- 87 -

Figure 3-2: AC-mode AFM height (A) and phase (B) images of MUDA-modified, flame-annealed gold. Lines intersecting in triangular patterns are the terraces between (111) crystal planes of the Au. (C-H): AFM imaging and force mapping of DEC221 (DOPC/egg-SM/Chol mixed in 2:2:1 molar ratio) bilayers on MUDA- modified, flame annealed Au at room temperature. AC-mode AFM height (C) and phase (D) images reveal phase segregation into liquid ordered (lo) and disordered (ld) phases. 64x48 force maps of the same area are processed to produce Young’s modulus and breakthrough force maps (E and F, respectively). Contrast between lo and ld phases are clearer in the breakthrough force map, and follow the features on the phase image (D). G and H are the histograms of the data mapped in E and F respectively. The histogram of breakthrough forces (H) has a largely bimodal distribution attributed to the lo phase (peak at ~ 3nN), and the ld phase (peak at ~ 2.1nN).

- 88 -

Contributions to the Young’s modulus by each phase were separated by considering their contributions to the breakthrough map. This was accomplished by examining the histograms of the Young’s modulus and breakthrough forces shown in Figure 3-2 G and H respectively. As was the case with the modulus map, the contributions to modulus histogram made by the lo and ld phases were not clearly distinguishable. On the other hand, their contributions to the breakthough force histogram were clear. Here the force curves that were not identified as “breakthrough” events by the classification algorithm have been excluded. The resulting bimodal distribution was then used as a guide to threshold the data. The cleft between the two peaks in the histogram was centered near 2.43 nN. Therefore all the breakthrough force curves were separated into two categories: Those with breakthrough forces ≥ 2.43 nN, and < 2.43 nN. Figure 3-3 shows the result of the separation. The features present in the separated Young’s modulus maps (Figure 3C and D) show good agreement with the features attributed to their respective phases seen in Figure 3-2D. This indicates that the method of separation, however unsophisticated, was effective at isolating the two phases. The two separated Young’s modulus histograms, therefore, mostly represent the two phases respectively. Log-normal fits to 69-71 each distribution were used to estimate the Young’s modulus of the lo phase and the ld phase which were ~100 ± 2 MPa and ~59.8 ± 0.9 MPa, respectively.

- 89 -

Figure 3-3: Breakthrough force contrast (see Figure 3-2 D and F) used to separate data from the Young’s modulus map (Figure 3-2 C and E) into contributions from the lo phase (A and C) and the ld phase (B and D). Modulus maps A) and B) clearly corresponds to features of the lo phase and the ld phase respectively, indicating the data sorting is reasonable. C and D are histograms of data seen in modulus maps A and B respectively. Log-normal fits to the histograms (black lines in C and D) are used to determine the peak value: The average modulus of the lo phase is 100 ± 2 MPa while that of the ld phase is 59.8 ± 0.9 MPa.

These apparent moduli represent contributions to the modulus from the bilayer itself, in addition to contributions from the underlying gold substrate and perhaps even from the thin aqueous layer between them. This is a consequence of assumptions made in using the Sneddon model to fit the modulus. Specifically, the model assumes the tip is indenting into an infinite half-space of an elastic material, which is not representative of this supported lipid system. Indenting into a thin, soft film on a hard substrate can cause an over-estimation of the modulus value because the stiffness of the substrate is felt through the soft film. Consequently, the apparent modulus value measured here is not the true modulus of the bilayer, and comparisons to other mechanical studies that attempt to physically decouple the bilayer from the substrate63 are difficult to draw. As mentioned earlier, the intent here is to use the modulus values to help identify the lipid phases.

- 90 -

3.3.4 Comparison with mica-supported bilayers:

To help confirm the identity of the lo and ld phases on gold, the experiment was repeated on mica substrates for comparison both by imaging and by force-maping. Figure 3-4A shows a typical height-contrast contact-mode AFM image, of DEC221 on mica, where the domains can be clearly identified by height. Defects in the bilayer gave rise to a third phase representing the bare mica substrate. In the breakthrough force map (Figure 3-4B), the bare mica patches appeared as white because they were removed from this map by the classification subroutine (i.e., they were not identified as breakthrough events). Like the data collected on gold-supported samples, the contributions to the modulus map and histogram were separated by identifying a threshold value between the two peaks attributed to the two lipid phases in the breakthrough force histogram (Figure 3-4C). The two separated modulus maps (Figure 3-5 A and B) showed features that correspond well to the two lipid phases in the height image, indicating that the separation was reasonable. Fitting the separated modulus histograms to log-normal functions yielded mean apparent moduli of 164 ± 2 MPa and 67 ± 1 MPa for the lo and ld phases, respectively.

- 91 -

Figure 3-4: DEC221 on mica via vesicle fusion at 50˚C. Contact mode AFM height map (A) clearly shows two lipid phases lo and ld (higher and lower respectively) as well as defects exposing bare mica. Breakthrough force map (B). White areas in (B) are excluded data points detected by software as non- breakthrough events. Features in (B) correspond well with features in (A). Breakthrough force histogram shows characteristic bimodal distribution. Lower force peak corresponds to ld phase while higher force peak belongs to the lo phase.

- 92 -

Figure 3-5: As with the lipid on gold, breakthrough force data (Figure 3-4 C) used to sort modulus data into contributions from lo and ld phases. A and C are the modulus map and associated histogram for the lo phase. B and D are the modulus map and histogram for the ld phase. Average moduli acquired from log ± normal fits to histogram are 164 2 MPa and 67 ± 1 MPa for the lo and ld phase. These results agree well with literature values40.

Qualitatively, the lipids behaved similarly on the two substrates. In both cases, phase segregation into the SM-enriched lo and the DOPC-enriched ld phases indicates that, like the mica surface, the gold substrate did not prevent spontaneous phase segregation. And as expected, denser packing of saturated SM molecules compared to unsaturated DOPC led to higher apparent moduli and breakthrough forces on the lo than the ld phase on both substrates.

- 93 -

Overall, the apparent moduli for the lipid phases on both substrates were consistent with published results from a similar lipid system on mica40. The mechanical properties of the lipids however, have been shown to depend strongly on lipid composition, temperature, thermal history, Ca2+ concentration, probe tip geometry, and the supporting substrate3,40,46. Although a quantitative comparison of mechanical properties was not the initial purpose of this work, such an analysis provided interesting insight into the relative contributions to the mechanical properties of the lipids from the van der Waals interactions between adjacent lipid chains, and from the electrostatic interactions between the headgroups and the substrate. The average apparent moduli for both phases were higher on mica than on gold. To account for this difference, factors that play a role in the apparent modulus of the bilayer were considered: the stiffness of the substrate, the thermal history of the lipids, and the surface charge density of the substrate.

3.3.5 The effect of the substrate stiffness on apparent modulus:

That the apparent modulus is influenced by the supporting substrate can be expected intuitively. The presence of the support impedes out-of-plane bending of the bilayer in response to an applied load, thus the mechanical response can be understood in terms of the axial compression of the bilayer along with a lateral spreading of adjacent lipid molecules. Comparisons between the apparent moduli of lipid systems on different solid supports can be made if the relative contributions to the apparent modulus from the different substrates can be accounted for. The method of Akhremitchev et al. 61 was used to estimate the factor by which the modulus of a thin, soft layer over a hard substrate was overestimated by calculating the ratio of Lamé coefficients between the layer and the substrate, and transforming the force-indentation data to normalized coordinates as described therein. The ratio of Lamé coefficients (Equation 7) was approximated as the ratio of moduli between the lipid layer and either the gold or mica substrate, considering that the other factors in the equation are small compared to this ratio. The average maximum indentation over which the modulus fits were calculated was ~1 nm, which from equation 8 produced a maximum normalized indentation of ~2. It was estimated that these values represent an overestimation by a factor of > 2.3. This factor was similar for

- 94 -

bilayers on mica because the ratio of Lamé coefficients for mica-supported lipids was similar to that of gold-supported lipids, which is expected given the mismatch in moduli between the substrate and overlayer are quite similar. The modulus of mica was estimated between 50 and 180 GPa, which like gold, is 3 orders of magnitude greater than the modulus of the bilayer measured here. Therefore, relative comparisons between apparent moduli on gold and mica are reasonable; that is, neither substrate will influence the deposited lipid much more than the other.

Other models exist to account for substrate interactions. For example, the model presented by Dimitriadis et al.72 accounts for thin gel-phase samples that are either bonded or not bonded to a rigid support. Mathematically, the model modifies the Sneddon equation with a quartic equation in √δ where the coefficients differ depending on whether or not the film is bonded to the solid support. These coefficients depend nonlinearly on Poisson’s ratio. This model however, is not suitable for determining whether or not a layer is bonded to the substrate, but instead is more appropriate when the state of film/layer bonding can be determined.

3.3.6 Thermal history of lipids:

The phase behavior of lipid layers is complex, even for pure components. For example, x- ray scattering of multilayers of pure sphingomyelin analogues as well as natural sphingomyelin mixtures such as ESM can exhibit coexisting phases under certain conditions73-75. Two gel-phases of ESM have been identified: interdigitated, and non- interdigitated, where the former consists of a sub-population of molecules within the ESM mixture that have asymmetric lipid chains that can interdigitate at the midplane of the bilayer73. The addition of cholesterol at >30 mol% tends to inhibit temperature- dependent phase segregation, and is interpreted as preventing the formation of the interdigitated phase73. On the other hand, simultaneous x-ray diffraction and differential scanning calorimetry (DSC) of a synthetic, racemic sphingomyelin analogue suggests 3

- 95 -

phases are possible, depending on thermal history75. But with its symmetric acyl chains, the SM analogue is not expected to form an interdigitated phase, which suggests alternate possibilities for the structure of the phases. Furthermore, a stereochemically pure synthetic SM can also exhibit three distinct gel phases -one is stable, and two are metastable- depending on thermal history, as revealed by DSC74. Since the lipids on the two substrates in the present study have undergone different thermal treatment, it was not expected that the same mechanical properties be observed between the lo phases of the two samples, since they may be in different phases. Although a more systematic study of the thermal-history-dependent behavior of these supported bilayers is warranted, some insight into the relative effect of thermal history and surface interaction can nevertheless be had as a consequence of this treatment.

The lipids on the mica were incubated at a temperature of 50˚C, which is higher than the melting temperature of all lipid components, whereas on the gold substrate the lipids were incubated at room temperature, which is above the melting temperature of the DOPC (-20˚C76), but below the melting temperature of ESM, (41˚C77). Comparison of the apparent modulus values between the two phases on the two substrates revealed a greater difference between the moduli of the lo phases than between the ld phases. This result is not surprising since in both cases, the annealing temperature is above the melting temperature of the DOPC, which is believed to be the main component of the ld phase, whereas only the mica sample was heated above the Tm of the ESM, which is enriched in the lo phase. Thus the thermal history was a greater factor in the properties of the lo phase than the ld phase. Perhaps the higher temperature incubation has enabled the annealing of a metastable lo phase into a stable phase. Other explanations for the difference between the moduli of the lo phases cannot be excluded. For example, the difference in the shape of the lo domains between the two samples suggests a difference in line-tension at the phase boundaries78, which may affect the apparent modulus by affecting the packing of the tail groups or by changing the ability of the lipids to deform laterally.

- 96 -

Interestingly, the breakthrough force values of the lo phase were ~3 nN for both samples, which suggests the breakthrough forces were not affected by the thermal history. The breakthrough forces recorded over the ld phase, however, were higher on the gold- supported bilayer than on mica. Since the experiments occurred above the Tm of the DOPC for both samples, the thermal history was not expected to be a factor in the properties of the ld phase. This suggests that the breakthrough forces of the ld phase was affected more by the properties of the substrate.

3.3.7 Surface charge density of substrates

Electrostatic interactions between lipid headgroups and charged surfaces have been shown to have a profound effect on the formation of supported bilayers from vesicles. Furthermore, changes in effective surface charge affected by the presence of divalent ions can greatly impact bilayer formation and stability. Specifically, Ca2+ has been show to promote the adsorption and rupture of lipid vesicles containing negative headgroups onto negatively charged surfaces by electrostatically bridging headgroup charges to charged sites on the surface. The same principle applies to zwitterionic headgroups, where the Ca2+ ion induces a dipole in the headgroup. In contrast, the efficiency of bilayer formation of cationic lipids on a net-negative surface was not affected.43

The strong electrostatic coupling between the surface charges and the lipid headgroups suggests that the surface charge density must influence the stability of the bilayer. It follows that this interaction influences the ability of the bilayer to deform laterally under an applied load. Specifically, during AFM indentation, a load is applied perpendicular to the bilayer and substrate, and the bilayer is deformed both in the direction of the applied load as well as perpendicular to the applied load. The latter occurs by lateral reorganization of the lipid molecules. Strong coupling of the lipid molecules to the surface through electrostatic interactions can impede lateral relaxation, and thus resist lateral deformation. The ability of a material to deform perpendicular to an applied load is

- 97 -

captured in Poisson’s ratio, ν, which can take on values between -1.0 and 0.5. Incompressible materials with ν = 0.5 (such as rubber) expand perpendicular to an applied compressive load, while compressible materials with ν = 0 (such as cork) do not strain in directions perpendicular to an applied load. Strong coupling between supported lipids and the underlying substrate would serve as a barrier to lateral deformation, and thus reduce the apparent Poisson’s ratio. Therefore, surface charge density is expected to be a factor in determining the apparent modulus

From estimates found in literature, the surface charge density of a COOH terminated monolayer on Au can have surface charge densities of up to 7.7x10-10 mol/cm2 based on a perfect MUDA structure on (111) gold and complete ionization79, compared to a maximum of 3.3x10-10 mol/cm2 based on the mica lattice structure80. It follows that the apparent Poisson’s ratio of the lipid on the MUDA/gold surface should appear lower than on the mica surface. Figure 3-6 illustrates the effect of apparent Poisson’s ratio on the fitted modulus. The modulus value is recalculated from the data for Poisson’s ratios between 0 and 0.5. Poisson’s ratios less than 0 were not considered because it is not apparent how the lipid bilayer could contract laterally under a compressive force. If, for the moment, one assumes that the true modulus of the bilayer on both substrates is the same, then the difference in the apparent modulus may result from differences in the

Poisson’s ratio on the two substrates. For the ld phase (two lower lines), the modulus could be the same on both mica and gold if the apparent Poisson’s ratios were ~0.5 and ~0.38 respectively. Qualitatively, this is consistent with the conclusion that the MUDA/gold substrate, with its greater surface charge density, exhibits stronger coupling to the bilayer, leading to reduced lateral mobility and consequently lower apparent Poisson’s ratios than on mica.

- 98 -

Figure 3-6: The effect of apparent Poisson’s ratio on the fitted Young’s modulus values. Solid and open markers are recalculated modulus values of the two phases on mica and gold respectively. Black dashed and solid lines (guides to the eye) are the lo and ld phases respectively. Consistent with the analysis discussed in the text, lower apparent Poisson’s ratio of the ld phase on MUDA/gold than on mica could lead to the same modulus value of actual bilayer, as indicated by the dotted grey line.

On the other hand, the differences between the moduli of the lo phase on the two substrates (upper two lines) appear too great to be explained by differences in apparent Poisson’s ratio. This is not surprising because this implies that the intermolecular forces between the tightly packed sphingomyelin dominate over the surface effects. This is consistent with the expectation that thermal annealing affects the lo phase much more than the ld phase.

This view was also supported by the breakthrough force data. The breakthough event, where during indentation the tip reaches a critical force at which it penetrates the bilayer and makes contact with the underlying substrate, occurs via a breakage of intermolecular cohesive forces between adjacent lipids in the bilayer. The breakthrough force histograms on both mica and gold indicate that the breakthrough force over the lo phase was similar on both substrates, around 3 nN, which indicated a negligible effect of changing the substrate. Meanwhile, for the ld phase, the peak breakthrough forces were ~1.5 nN and

- 99 -

2.1 nN on mica and gold, respectively. This supports the expectation that the more mobile ld phase, with its weaker intermolecular forces, is more strongly influenced by electrostatic coupling of headgroups to the surface. Furthermore, the fact that the breakthrough force of the ld phase was greater on gold than on mica is consistent with the greater surface charge density of the MUDA/gold substrate.

These considerations taken together suggest that the mechanical properties of the lo phase were dominated by the thermal history and annealing due to the relatively large van der Waals forces present between the tightly packed alkyl chains. Meanwhile the mechanical properties of the ld phase, where the relative contribution from van der Waals forces was low, were dominated by the electrostatic coupling to the surface.

3.4 Conclusions

Previous methods of producing supported bilayers on gold have employed a variety of surface and lipid modifications in order to promote the adsorption and rupture of vesicles at the gold surface. Arguably, these modifications deviate from natural lipid bilayers. Although no supported lipid bilayer is identical to a cell membrane, ternary lipid mixtures of phosphatidylcholine lipids, sphingomyelin and cholesterol are routinely used as model membranes supported by mica substrates because like cellular membranes, they segregate into lo and ld phases. The deposition of model, phase-segregated lipid bilayers (specifically a 2:2:1 mixture of DOPC, egg-sphingomyelin, and cholesterol) on gold substrates using the standard vesicle fusion method has been demonstrated without the use of charged or modified lipid headgroups. Also, the phase segregation of the bilayers on gold was directly observed using both AFM imaging and force indentation mapping. To facilitate AFM evaluation, the gold surfaces were smoothed by hydrogen flame annealing. Additionally, a high density mercaptoundecanoic acid SAM was grafted on the surface to improve the hydrophilicity, and make the surface negatively charged to mimic mica.

- 100 -

The quality of the SLB on gold was evaluated in terms of the morphology and mechanical properties and compared to the same SLB on mica. Qualitatively, phase segregation was confirmed on both substrates from the AFM images. The lo and ld phases were indentified on gold by the height contrast as well as tapping mode phase contrast. Qualitatively, the mechanical properties, probed by force indentation mapping, were similar on both substrates: Statistically higher breakthrough forces were recorded over the lo phase than the ld phase on both substrates. Although the fitted moduli of the two phases differed between the gold- and mica-supported bilayers, they were of comparable magnitude. On both substrates however, the lo phase had a higher modulus than the ld phase.

Further qualitative comparisons of the two samples revealed that the mechanical behavior of the lo and ld phases had different origins in terms of the relative contributions of van der Waals and electrostatic forces. The dominant electrostatic forces originate from 2+ coupling of the lipid headgroups to surface charges through divalent cations such as Ca . Van der Waals forces originate from intermolecular interactions between adjacent alkyl chains of the lipids. The properties of the lo phase were dominated by strong van der Waals interactions attributed to tight packing of the lipids, and consequently, were affected more strongly by thermal history than by a change in surface charge density. Consequently, changing the surface charge density by changing the substrate did not affect the breakthrough force (~ 3 nN on both surfaces), which depends strongly on intermolecular interactions. The modulus of the lo phase was higher on the surface where thermal annealing of the bilayer took place (164 MPa vs 100 MPa). On the other hand, the mechanical properties of the ld phase were less sensitive to thermal history, and more sensitive to changes in surface charge density, which indicates that the electrostatic coupling of the lipids to the surface dominated over the relatively weak van der Waals interactions. In this case, differences in both the modulus values and the breakthrough force values have their origin in the surface charge densities. The MUDA-modified gold

- 101 -

substrate, with a higher surface charge density supported an ld phase with higher breakthrough forces (~2.1 nN vs ~1.5 nN) and lower apparent Poisson’s ratios (ν = 0.38 vs ν = 0.5). The lower apparent Poisson’s ratio resulted from the decreased lateral mobility of lipids that inhibits lateral strain, resulting in lower calculated modulus values (62 MPa vs 80 MPa). Though this is speculative, since it could be that the Young’s modulus of these two disordered phases were simply different.

The methods employed herein can be extended in future experiments to better understand the role of surface charge density on the morphology and mechanical properties of SLBs, for example, by using mixed –COOH and –OH terminated monolayers to control surface charge density. Furthermore, the ability to form model membranes on metallic surfaces will improve the biological relevance of methods such as SPR in characterizing lipid bilayers. The demonstrated real-time sensitivity of SPR could have an impact on understanding the interaction dynamics between soluble analytes such as drugs and membranes and membrane-bound proteins.

3.5 Acknowledgements

The author thanks James K. Li for the force map analysis code which formed the basis for automated force map analysis.

3.6 References

1. Johnston, L.J. Nanoscale Imaging of Domains in Supported Lipid Membranes. Langmuir 23, 5886-5895 (2007).

2. Connell, S.D. & Smith, D.A. The atomic force microscope as a tool for studying phase separation in lipid membranes (Review). Mol. Membr. Biol. 23, 17-28 (2006).

3. Dufrêne, Y.F. & Lee, G.U. Advances in the characterization of supported lipid films

- 102 -

with the atomic force microscope. BBA-Biomembranes 1509, 14-41 (2000).

4. Kruger, S., Kruger, D. & Janshoff, A. Scanning Force Microscopy Based Rapid Force Curve Acquisition on Supported Lipid Bilayers: Experiments and Simulations Using Pulsed Force Mode. ChemPhysChem 5, 989-997 (2004).

5. Slade, A., Luh, J., Ho, S. & Yip, C.M. Single molecule imaging of supported planar lipid bilayer--reconstituted human insulin receptors by in situ scanning probe microscopy. J. Struct. Biol. 137, 283-291 (2002).

6. Hollars, C.W. & Dunn, R.C. Submicron Structure in l-[alpha]- Dipalmitoylphosphatidylcholine Monolayers and Bilayers Probed with Confocal, Atomic Force, and Near-Field Microscopy. Biophys. J. 75, 342-353 (1998).

7. Korlach, J., Schwille, P., Webb, W. & Feigenson, G. Characterization of lipid bilayer phases by confocal microscopy and fluorescence correlation spectroscopy. Proc. Natl. Acad. Sci. U.S.A. 96, 8461-8466 (1999).

8. Mattjus, P. & Slotte, J.P. Does cholesterol discriminate between sphingomyelin and phosphatidylcholine in mixed monolayers containing both phospholipids? Chem. Phys. Lipids 81, 69-80 (1996).

9. Chiantia, S., Ries, J. & Schwille, P. Fluorescence correlation spectroscopy in membrane structure elucidation. BBA-Biomembranes 1788, 225-233 (2009).

10. Steinem, C., Janshoff, A., Galla, H. & Sieber, M. Impedance analysis of ion transport through gramicidin channels incorporated in solid supported lipid bilayers. Bioelectrochem. Bioenerg. 42, 213-220 (1997).

11. Steinem, C., Janshoff, A., Ulrich, W.P., Sieber, M. & Galla, H.J. Impedance analysis of supported lipid bilayer membranes: a scrutiny of different preparation techniques. BBA-Biomembranes 1279, 169-180 (1996).

12. Kraft, M.L., Weber, P.K., Longo, M.L., Hutcheon, I.D. & Boxer, S.G. Phase Separation of Lipid Membranes Analyzed with High-Resolution Secondary Ion Mass

- 103 -

Spectrometry. Science 313, 1948-1951 (2006).

13. Johnson, S. et al. Structure of an adsorbed dimyristoylphosphatidylcholine bilayer measured with specular reflection of neutrons. Biophys. J. 59, 289-294 (1991).

14. London, E. & Brown, D. Insolubility of lipids in Triton X-100: physical origin and relationship to sphingolipid/cholesterol membrane domains (rafts). BBA- Biomembranes 1508, 182-195 (2000).

15. Brown, D. & London, E. Structure and Origin of Ordered Lipid Domains in Biological Membranes. J. Membr. Biol. 164, 103-114 (1998).

16. Brown, D.A. & London, E. Functions of Lipid Rafts In Biological Membranes. Annu. Rev. Cell. Dev. Biol. 14, 111-136 (1998).

17. Simons, K. & Vaz, W.L. MODEL SYSTEMS, LIPID RAFTS, AND CELL MEMBRANES1. Annu. Rev. Biophys. Biomol. Struct. 33, 269-295 (2004).

18. Stillwell, W., Shaikh, S.R., Zerouga, M., Siddiqui, R. & Wassall, S.R. Docosahexaenoic acid affects cell signaling by altering lipid rafts. Reprod. Nutr. Dev. 45, 21 pages (2005).

19. Wong, P. et al. Amyloid-beta Membrane Binding and Permeabilization are Distinct Processes Influenced Separately by Membrane Charge and Fluidity. J. Mol. Biol. 386, 81-96 (2009).

20. D'Errico, G. et al. Interaction between Alzheimer's A beta(25-35) peptide and phospholipid bilayers: The role of cholesterol. BBA-Biomembranes 1778, 2710-2716 (2008).

21. Liao, Z., Graham, D.R. & Hildreth, J.E.K. Lipid rafts and HIV pathogenesis: virion- associated cholesterol is required for fusion and infection of susceptible cells. AIDS Res. Hum. Retroviruses 19, 675-687 (2003).

22. Campbell, S.M., Crowe, S.M. & Mak, J. Lipid rafts and HIV-1: from viral entry to

- 104 -

assembly of progeny virions. J. Clin. Virol. 22, 217-227 (2001).

23. Werbin, J.L., Heinz, W.F., Romer, L.H. & Hoh, J.H. Micropatterns of an Extracellular Matrix Protein with Defined Information Content. Langmuir 23, 10883- 10886 (2007).

24. Santore, M.M. & Kozlova, N. Micrometer Scale Adhesion on Nanometer-Scale Patchy Surfaces: Adhesion Rates, Adhesion Thresholds, and Curvature-Based Selectivity. Langmuir 23, 4782-4791 (2007).

25. Mossman, K.D., Campi, G., Groves, J.T. & Dustin, M.L. Altered TCR Signaling from Geometrically Repatterned Immunological Synapses. Science 310, 1191-1193 (2005).

26. Yu, J. et al. Polyvalent interactions of HIV-gp120 protein and nanostructures of carbohydrate ligands. NanoBioTechnology 1, 201-210 (2005).

27. Plant, A.L., Brighamburke, M., Petrella, E.C. & Oshannessy, D.J. Phospholipid/Alkanethiol Bilayers for Cell-Surface Receptor Studies by Surface Plasmon Resonance. Anal. Biochem. 226, 342-348 (1995).

28. Tawa, K. & Morigaki, K. Substrate-supported phospholipid membranes studied by surface plasmon resonance and surface plasmon fluorescence spectroscopy. Biophys. J 89, 2750-2758 (2005).

29. Jonsson, M.P., Jonsson, P., Dahlin, A.B. & Hook, F. Supported lipid bilayer formation and lipid-membrane-mediated biorecognition reactions studied with a new nanoplasmonic sensor template. Nano Lett. 7, 3462-3468 (2007).

30. Salamon, Z., Macleod, H.A. & Tollin, G. Surface plasmon resonance spectroscopy as a tool for investigating the biochemical and biophysical properties of membrane protein systems. II: Applications to biological systems. BBA-Rev Biomembranes 1331, 131-152 (1997).

31. Watts, T.H., Gaub, H.E. & McConnell, H.M. T-cell-mediated association of peptide

- 105 -

antigen and major histocompatibility complex protein detected by energy transfer in an evanescent wave-field. Nature 320, 179-181 (1986).

32. Castellana, E.T. & Cremer, P.S. Solid supported lipid bilayers: From biophysical studies to sensor design. Surf. Sci. Rep. 61, 429-444 (2006).

33. Avery, C.W., Som, A., Xu, Y., Tew, G.N. & Chen, Z. Dependence of Antimicrobial Selectivity and Potency on Oligomer Structure Investigated Using Substrate Supported Lipid Bilayers and Sum Frequency Generation Vibrational Spectroscopy. Anal. Chem. 81, 8365-8372 (2009).

34. Sek, S., Laredo, T., Dutcher, J.R. & Lipkowski, J. Molecular Resolution Imaging of an Antibiotic Peptide in a Lipid Matrix. J. Am. Chem. Soc. 131, 6439-6444 (2009).

35. Smith, D.P. et al. Images of a lipid bilayer at molecular resolution by scanning tunneling microscopy. Proc. Natl. Acad. Sci. U.S.A 84, 969-972 (1987).

36. Gregory, B.W., Dluhy, R.A. & Bottomley, L.A. Structural characterization and nanometer-scale domain formation in a model phospholipid bilayer as determined by infrared spectroscopy and scanning tunneling microscopy. J. Phys. Chem. 98, 1010- 1021 (1994).

37. Carrer, D., Kummer, E., Chwastek, G., Chiantia, S. & Schwille, P. Asymmetry determines the effects of natural ceramides on model membranes. Soft Matter 5, 3279-3286 (2009).

38. Dietrich, C. et al. Lipid Rafts Reconstituted in Model Membranes. Biophys. J. 80, 1417-1428 (2001).

39. Rinia, H.A. & de Kruijff, B. Imaging domains in model membranes with atomic force microscopy. FEBS Lett. 504, 194-199 (2001).

40. Sullan, R.M.A., Li, J.K. & Zou, S. Direct Correlation of Structures and Nanomechanical Properties of Multicomponent Lipid Bilayers. Langmuir 25, 7471- 7477 (2009).

- 106 -

41. Brian, A.A. & McConnell, H.M. Allogeneic stimulation of cytotoxic T cells by supported planar membranes. Proc. Natl. Acad. Sci. U.S.A. 81, 6159-6163 (1984).

42. Johnson, J.M., Ha, T., Chu, S. & Boxer, S.G. Early Steps of Supported Bilayer Formation Probed by Single Vesicle Fluorescence Assays. Biophys. J. 83, 3371-3379 (2002).

43. Richter, R.P. & Brisson, A.R. Following the Formation of Supported Lipid Bilayers on Mica: A Study Combining AFM, QCM-D, and Ellipsometry. Biophys. J. 88, 3422-3433 (2005).

44. Reviakine, I. & Brisson, A. Formation of Supported Phospholipid Bilayers from Unilamellar Vesicles Investigated by Atomic Force Microscopy. Langmuir 16, 1806- 1815 (2000).

45. Cha, T., Guo, A. & Zhu, X. Formation of Supported Phospholipid Bilayers on Molecular Surfaces: Role of Surface Charge Density and Electrostatic Interaction. Biophys. J. 90, 1270-1274 (2006).

46. Leonenko, Z.V., Carnini, A. & Cramb, D.T. Supported planar bilayer formation by vesicle fusion: the interaction of phospholipid vesicles with surfaces and the effect of gramicidin on bilayer properties using atomic force microscopy. BBA-Biomembranes 1509, 131-147 (2000).

47. Lingler, S., Rubinstein, I., Knoll, W. & Offenhausser, A. Fusion of Small Unilamellar Lipid Vesicles to Alkanethiol and Thiolipid Self-Assembled Monolayers on Gold. Langmuir 13, 7085-7091 (1997).

48. Rädler, U., Mack, J., Persike, N., Jung, G. & Tampé, R. Design of supported membranes tethered via metal-affinity ligand-receptor pairs. Biophys. J. 79, 3144-52 (3144).

49. Naumann, C.A. et al. The Polymer-Supported Phospholipid Bilayer: Tethering as a New Approach to Substrate−Membrane Stabilization. Biomacromolecules 3, 27-35 (2002).

- 107 -

50. Steltenkamp, S. et al. Mechanical Properties of Pore-Spanning Lipid Bilayers Probed by Atomic Force Microscopy. Biophys. J. 91, 217-226 (2006).

51. McKiernan, A.E., Ratto, T.V. & Longo, M.L. Domain Growth, Shapes, and Topology in Cationic Lipid Bilayers on Mica by Fluorescence and Atomic Force Microscopy. Biophys. J. 79, 2605-2615 (2000).

52. Ti Tien, H. Self‐assembled lipid bilayers for biosensors and molecular electronic devices. Adv. Mater. 2, 316-318 (1990).

53. Sinner, E. & Knoll, W. Functional tethered membranes. Curr. Opin. Chem. Biol. 5, 705-711 (2001).

54. Naumann, R. et al. Tethered Lipid Bilayers on Ultraflat Gold Surfaces. Langmuir 19, 5435-5443 (2003).

55. Garg, S., Rühe, J., Lüdtke, K., Jordan, R. & Naumann, C.A. Domain Registration in Raft-Mimicking Lipid Mixtures Studied Using Polymer-Tethered Lipid Bilayers. Biophys. J. 92, 1263-1270 (2007).

56. Kiessling, V., Crane, J.M. & Tamm, L.K. Transbilayer Effects of Raft-Like Lipid Domains in Asymmetric Planar Bilayers Measured by Single Molecule Tracking. Biophys. J. 91, 3313-3326 (2006).

57. Li, M. et al. AFM Studies of Solid-Supported Lipid Bilayers Formed at a Au(111) Electrode Surface Using Vesicle Fusion and a Combination of Langmuir−Blodgett and Langmuir−Schaefer Techniques. Langmuir 24, 10313-10323 (2008).

58. Chen, M., Li, M., Brosseau, C.L. & Lipkowski, J. AFM Studies of the Effect of Temperature and Electric Field on the Structure of a DMPC−Cholesterol Bilayer Supported on a Au(111) Electrode Surface. Langmuir 25, 1028-1037 (2009).

59. Zhang, L., Booth, C.A. & Stroeve, P. Phosphatidylserine/Cholesterol Bilayers Supported on a Polycation/Alkylthiol Layer Pair. J Colloid Interface Sci 228, 82-89 (2000).

- 108 -

60. Wang, H., Chen, S., Li, L. & Jiang, S. Improved method for the preparation of carboxylic acid and amine terminated self-assembled monolayers of alkanethiolates. Langmuir 21, 2633-6 (2005).

61. Akhremitchev, B.B. & Walker, G.C. Finite sample thickness effects on elasticity determination using atomic force microscopy. Langmuir 15, 5630-5634 (1999).

62. Popov, J. et al. Chemical Mapping of Ceramide Distribution in Sphingomyelin-Rich Domains in Monolayers. Langmuir 24, 13502-13508 (2008).

63. Ngwa, W. Nanoscale mechanics of solid-supported multilayered lipid films by force measurement. Thin Solid Films 516, 5045 (2008).

64. Uosaki, K., Shen, Y. & Kondo, T. Preparation of a Highly Ordered Au (111) Phase on a Polycrystalline Gold Substrate by Vacuum Deposition and Its Characterization by XRD, GISXRD, STM/AFM, and Electrochemical Measurements. J. Phys. Chem. 99, 14117-14122 (1995).

65. Brown, D.A. & London, E. Structure and Function of Sphingolipid- and Cholesterol- rich Membrane Rafts. J. Biol. Chem. 275, 17221-17224 (2000).

66. Litman, B.J., Lewis, E.N. & Levin, I.W. Packing characteristics of highly unsaturated bilayer lipids: Raman spectroscopic studies of multilamellar phosphatidylcholine dispersions. Biochemistry 30, 313-319 (1991).

67. Ladbrooke, B.D. & Chapman, D. Thermal analysis of lipids, proteins and biological membranes a review and summary of some recent studies. Chem. Phys. Lipids 3, 304-356 (1969).

68. Cronan, J.E. & Gelmann, E.P. Physical properties of membrane lipids: biological relevance and regulation. Bacteriol. Rev. 39, 232-256 (1975).

69. Limpert, E., Stahel, W.A. & Abbt, M. Log-normal Distributions across the Sciences: Keys and Clues. BioScience 51, 341 (2001).

- 109 -

70. Cross, S.E., Jin, Y., Rao, J. & Gimzewski, J.K. Nanomechanical analysis of cells from cancer patients. Nat. Nanotechnol. 2, 780-783 (2007).

71. Gerbal, F. et al. Measurement of the elasticity of the actin tail of Listeria monocytogenes. Eur. Biophys. J. 29, 134-140 (2000).

72. Dimitriadis, E.K., Horkay, F., Maresca, J., Kachar, B. & Chadwick, R.S. Determination of Elastic Moduli of Thin Layers of Soft Material Using the Atomic Force Microscope. Biophys. J. 82, 2798-2810 (2002).

73. Chachaty, C., Rainteau, D., Tessier, C., Quinn, P.J. & Wolf, C. Building Up of the Liquid-Ordered Phase Formed by Sphingomyelin and Cholesterol. Biophys. J. 88, 4032-4044 (2005).

74. Bruzik, K.S. & Tsai, M.D. A calorimetric study of the thermotropic behavior of pure sphingomyelin diastereomers. Biochemistry 26, 5364-5368 (1987).

75. Takahashi, H., Okumura, Y. & Sunamoto, J. Structure and thermal history dependant phase behavior of hydrated synthetic sphingomyelin analogue: 1,2-dimyristamido- 1,2-deoxyphosphatidylcholine. BBA-Biomembranes 1713, 40-50 (2005).

76. Schroeder, R., London, E. & Brown, D. Interactions between saturated acyl chains confer detergent resistance on lipids and glycosylphosphatidylinositol (GPI)- anchored proteins: GPI-anchored proteins in liposomes and cells show similar behavior. Proc. Natl. Acad. Sci. U.S.A. 91, 12130-12134 (1994).

77. Samsonov, A.V., Mihalyov, I. & Cohen, F.S. Characterization of Cholesterol- Sphingomyelin Domains and Their Dynamics in Bilayer Membranes. Biophys. J. 81, 1486-1500 (2001).

78. García-Sáez, A.J., Chiantia, S. & Schwille, P. Effect of Line Tension on the Lateral Organization of Lipid Membranes. J. Biol. Chem. 282, 33537-33544 (2007).

79. Smalley, J.F., Chalfant, K., Feldberg, S.W., Nahir, T.M. & Bowden, E.F. An Indirect Laser-Induced Temperature Jump Determination of the Surface pKa of 11-

- 110 -

Mercaptoundecanoic Acid Monolayers Self-Assembled on Gold. J. Phys. Chem. B 103, 1676-1685 (1999).

80. Qi, G. et al. Quantifying Surface Charge Density by Using an Electric Force Microscope with a Referential Structure. J. Phys. Chem. C 113, 204-207 (2009).

- 111 -

Chapter 4: Lipid bilayers on nanoaperture arrays

Biologically relevant cell membrane mimics represent a promising surface functionalization for biosensors. The deposition of model lipid bilayers on plasmonic nanohole arrays, nanoslit arrays, and annular aperture arrays is demonstrated. Force indentation is used to confirm coverage of the arrays with lipids, with characteristic breakthrough events observed throughout the scan areas on each array. Force properties were influenced by substrate topography, which in large scan areas exhibiting much variation in topography impeded the observation of lipid phase segregation. In smaller scans over single grains however, phase segregation is observed over hole arrays, and confirmed by separating force indentation data from the two phases and quantifying their Young’s moduli. Following the expected trend, moduli over the liquid ordered domains is higher than over the liquid disordered domains at 64MPa and 50MPa respectively. In large scans, modulus contrast can be seen between indentations over patterned apertures and over the gold substrate. Comparisons between these properties were discussed. Transmission spectra through the arrays were affected by the deposition of the lipids, with Au/solution Bloch modes of the hole-arrays in the visible and near-infrared registering shifts of 4nm and 6nm respectively due to the deposition of the lipids. As sensor surface functionalizations, these raft-forming model lipid bilayers are potentially valuable for characterizing the role of lipid rafts in the binding of drugs, proteins, or small molecules to cell membranes.

4.1 Introduction

Considerable interest exists in applying lipid bilayers as a fairly biocompatible1,2, and versatile surface functionalization platform and interface for a variety of biosensors3‐8. Such a sensor could be used to study ion-channel conductance9 or the binding of small molecules such as drugs or peptides to cell-membrane mimics3,4,10, depending on the

- 112 -

choice of transducer. Incorporation of membrane proteins, which represent a significant fraction of the proteome11, and are critical to many physiological processes such as nerve and immune function, potentially enables the detection of specific physiologically relevant analytes that may interact with these proteins12,13. Furthermore, the commercial availability of head-group functionalized lipids that bear reactive chemical species or biotin potentially increases the breadth of accessible targets8,14. In addition to versatility and biocompatibility, another potential advantage to lipid-based surface functionalization is the potential for targeting ligands to be mobile within the surface, which allows for flexible distribution of surface ligands, which can potentially aid multivalent attachment of a target analyte15. Despite their fluidity, supported lipid bilayers and even pore- spanning bilayers can remain stable for 24 hours7.

Although deposition of lipid bilayers on surfaces such as glass and mica is routinely reported16‐19, deposition of bilayers on metallic surfaces, which are more relevant for biosensing, often requires modification of surface chemistry6,20. Various methods exist to cast lipid bilayers on metallic surfaces, many of which rely on the bonding or tethering of some portion of the bilayer to the metal surface through specific headgroup charge or chemistry21‐24. An alternative method is presented in the previous chapter that enables the self-assembly of untethered zwitterionic lipids. In fact, mixtures of lipids that are known to phase segregate were cast on gold surfaces to mimic lipid rafts25. In this chapter, the methodologies discussed in the previous chapter are applied to cast lipid bilayers on plasmonic nanoaperture array26,27 sensors.

The nanoaperture arrays utilized here are metal films that feature a periodic arrangement of holes, slits or annuli that exhibit unique optical transmission properties that are sensitive to changes in optical index or dielectric constant at the array surface26‐28. Their unique optical properties are attributed primarily to the excitation of surface plasmon polariton Bloch-waves29, whose properties depend on the geometric parameters of the array, such as aperture size and period, as well as on the refractive index and/or dielectric

- 113 -

constant of the medium in which they propagate27. Like commercially available surface plasmon resonance sensors, controlled adsorption of analyte species to the sensor surface can change the effective refractive index of the interface, which affects the properties of the surface plasmon. The properties of the plasmon are observable in the transmission spectrum of the arrays, since the transmitted light is largely dominated by the scattered surface plasmons28,30‐32.

There are a number of reports in literature regarding the deposition of lipid bilayers on porous substrates that resemble nano-hole arrays in terms of scale7,14,33. A subset of these studies report their use in biosensing applications7,14; in one case as a component of an electrochemical sensor, and in another as a random arrangement of holes that relies on the localized surface plasmons for signal transduction. In the latter, a metal film with a random pattern of holes is coated with sputtered SiOx that acts to facilitate lipid fusion on the surface. A single lipid species was mixed with a small amount of biotinylated lipid, and a biotin-avidin interaction is detected and reported to produce a 1 nm shift in the position of the LSP resonance.

In this chapter, a model lipid mixture is deposited onto three types of nanoaperture arrays: nano-hole arrays, nano-slit arrays34‐36, and annular aperture arrays37‐39. These model lipid mixtures are known to phase-segregate into domains that mimic some aspects of lipid rafts. As with the previous chapter, a mercaptoundecanoic acid monolayer is first cast on the arrays to protect the surface from contamination, and to make the surface extremely hydrophilic, a property required for efficient fusion of vesicles on the surface. The integrity, and extent of coverage of the lipid bilayer is verified by atomic force microscopy and its mechanical properties measured by force mapping.

The mechanical characterization of the bilayer is used to observe and identify two coexisting lipid phases on a nano-hole array. These data are also used to characterize the

- 114 -

mechanical properties of the bilayer over the apertures to determine if the bilayer spans the apertures. Indentation of the AFM tip at regular intervals over a fixed area allows each force-indentation curve to be mapped to a (x,y) coordinate, and can also be used to collect statistics of mechanical data across the scan area. Maps of mechanical properties can be constructed from this data and correlated to features in the AFM image. Of particular interest in the force analysis of lipid bilayers is the breakthrough event. The breakthrough event is characteristic of indentation into a lipid bilayer, and occurs when enough force is applied to the lipid bilayer through the AFM tip so as to temporarily rupture the bilayer. The sudden rupture of the bilayer is associated with a discontinuity in the measured force. The breakthrough process is schematically illustrated in Figure 4-1.

The critical force at which the tip is able to breakthrough the bilayer depends on the mechanical properties of the lipid layer, and can be used to differentiate between coexisting lipid phases, as discussed in the previous chapter.

- 115 -

The presence of the lipid bilayer can be determined by the presence of the breakthough event, since indentation on a rigid surface not coated with the bilayer results in a different force-indentation profile that does not have a discontinuity. The presence of nanoapertures on the surface introduces other types of force-indentation curves that result when the tip interacts with an aperture during indentation. An automated computer algorithm is used to sort the force curves into three categories: Breakthrough events, Indentation without breakthrough, which are called Hard-contact events, and “Other” events that cannot be identified as either of the other two events. Example force curves for each event are shown in the lower half of Figure 4-1 for reference. This categorization enables the breakthrough events to be isolated from the other events. When displayed as a 2D map, the categorization allows the extent of lipid coverage to be assessed. Defects in the bilayer appear as clusters of hard-contact events in the map, while indentations over apertures typically result in “other” events.

- 116 -

4.2 Materials and Methods

Dioleoylphosphatidylcholine/Egg sphingomyelin/Ovine Cholesterol (DEC221) aliquots: Dioleoylphosphatidylcholine (DOPC) in chloroform, lyophilized egg-sphingomyelin (ESM), and lyophilized ovine cholesterol (Chol) were purchased from Avanti Polar Lipids (Alabaster, Al) and stored at -20 °C until use. For aliquots, DOPC (used directly), ESM dissolved in a 2:1 mixture of chloroform and methanol, Chol dissolved in chloroform were mixed to a 2:2:1 molar ratio and a final mass of 10.7 mg. The mixture was divided into 10 aliquots in glass vials and dried under a stream of Argon gas distributed by a manifold for 1h or until the lipids formed a film on the bottom of the vials. The vials were then dried under vacuum overnight to remove residual solvent. The dried lipid films were repressurized with Argon gas, capped, and sealed with tape and stored at -20 °C until use.

Nanoaperture Arrays: Nanohole arrays, nanoslit arrays and annular aperture arrays were fabricated by collaborators at the University of Pittsburgh using focused ion beam milling (FIB). Starting with flame-annealed 11 mm x 11 mm gold substrates (Arrandee, Wether, Germany) prepared as described in the previous chapter, patterns were directly milled near the middle of the substrate. Each pattern was approximately 40 µm x 40 µm. Several substrates were patterned with all three array types having around 100 µm edge-to-edge spacing to prevent cross talk and to ensure spectra can be spatially separated. In the data shown however, the nanohole and nanoslit arrays were on the same substrate, while the annular aperture array was on another. Nominal diameter and period for the nanohole array were 200 nm and 500 nm respectively. Nominal slit width and period for the nanoslit array were 200 nm and 400 nm respectively. The nominal period for the annular aperture array was 900 nm and the inner radius was 125 nm and the outer radius was 215 nm. FIB cross-sections of the films suggest that the thickness of the gold films were over

- 117 -

300 nm. For lipid deposition, AFM imaging and Vis-NIR spectrometry, samples were mounted to 30 mm round glass slides (Asylum Research, Santa Barbara, CA, USA) with a thickness similar to standard light microscopy slides, or to standard microscopy slides cut to 25 mm x 25 mm size, where the former was used to mount to a commercial AFM fluid cell (Asylum Research), while the latter was used to mount the sample in a custom fluid cell. Substrates were bonded to the glass slides using a UV cured optical adhesive (Norland NOA88, Norland Products, Cranbury, NJ).

Mercaptoundecanoic Acid (MUDA) SAM: Self-assembled monolayers of mercaptoundecanoic acid (95%, Sigma-Aldrich Canada Ltd, Oakville, ON, Canada) were deposited after the gold substrates were bonded to the glass slides, and following at least 2 hours of UV cleaning. MUDA SAMs were deposited as described in the previous chapter. The SAM quality was deemed sufficient when after rinsing with 18 MΩ-cm water, a thin sheet of water remained on the gold surface even after shaking. The substrates and attached glass slides were then mounted in the appropriate clean fluid cell and stored in 18 MΩ-cm water and covered until lipid deposition.

Lipid deposition: Aliquots of DEC221 were thawed and hydrated in 18 MΩ-cm water to a final concentration of 1mg/mL at 50 °C with periodic vortexing for 30 minutes. The resulting multilamellar vesicle (MLV) suspension was extruded through 1µm pores using commercially acquired lipid extruder (Mini Extruder, Avanti Polar Lipds, Alabaster, AL, USA) at 50 °C as described in the previous chapter. A portion of the resulting unilamellar vesicle (ULV) suspension was used immediately, and the rest was stored for later use in a 1 dram glass vial at 4 °C for up to one week. For deposition, 18 MΩ-cm water, lipids, and 100 mM CaCl2 solution were preheated to 50 °C. Substrates mounted in fluid cells were brought to temperature by repeatedly filling the fluid cell with 50 °C 18 MΩ-cm water and discarding. Finally, the ULV suspension, CaCl2 and water were applied to the warmed substrate to a final lipid concentration of 0.05 mg/mL and final CaCl2 concentration of 10 mM and incubated at room temperature for 30 min. The substrates

- 118 -

were then rinsed indirectly with about 100 mL of 18 MΩ-cm water form a squeeze bottle without allowing the substrate and bilayer to be exposed to air, or to cross an air/water interface.

AFM Imaging and Force Mapping: All imaging and force mapping was performed in 18 MΩ-cm water using an MFP-3D AFM (Asylum Research) and proprietary iDrive magnetic tapping mode accessory and specialized metal-coated tips. Substrates were equilibrated with the AFM tip in water for at least 1 hour and as much as 8 hours before imaging. A top-view camera accessory was used to align the tip and the patterned arrays, which are both clearly visible in the camera’s field of view. After a suitable location was imaged, force maps were collected using the built-in software function, and analyzed offline using custom software written in Igor Pro.

Vis-NIR Transmisson Spectra: Visible and NIR transmission spectra were collected using Ocean Optics HR2000+ and NIR512 fiber optic spectrometers (Ocean Optics, Dunedin, FL, USA) connected via optical fiber to the front port of a Nikon inverted microscope (Nikon Eclipse TE2000, Nikon Canada Inc., Mississauga, Canada). After lipids have been characterized by AFM, the fluid cells containing the arrays and lipids were moved to the microscope where spectra were collected directly through the fluid cell without disturbing the substrate.

The microscope’s Tungsten-Halogen lamp was used as an unpolarized broad-band source to illuminate the top-side of the substrate, while a 20x objective below the fluid-cell was used to collect transmitted light. While the field of view of the microscope is much larger than any given array, an iris mounted to the front port of the microscope via a C-mount- to-lens-tube assembly (ThorLabs, Newton, NJ, USA) was used to restrict the field entering the spectrometer to an area about the size of the array. A collimating lens (Ocean Optics) designed specifically for the optical fiber (Ocean Optics), couples light into the

- 119 -

fiber. The collimating lens is mounted after the iris in the optical path via a 2-axis adjustable fiber mount modified to fit the thread of the collimating lens (ThorLabs). The 2-axis enabled optimal positioning of the collimating lens in the center of the optical path. The two spectrometers were interchanged by attaching the free end of the fiber to one spectrometer or the other.

Nanoaperture arrays were positioned in the center of the field of view by adjusting the position of the microscope stage while visually aligning the array with a crosshair visible through the ocular lenses, which marks the center of view. A reference bright spectrum was collected through a single aperture, equivalent in size to the extents of the arrays, which was milled into the substrate during array fabrication for this purpose. The real- time spectrum through this aperture was also used to adjust the field of view of the spectrometer, via an iris. A reference dark spectrum is collected over an unpatterned, and defect-free region of the substrate. The %T transmission spectra were calculated by the spectrometer software. Multiple arrays were fabricated on each substrate, and their spectra were collected serially using one spectrometer, then again using the other. Only spectra for arrays upon which the lipid layers have been characterized are shown. Lipid- free spectra were also taken in this fashion under water after the MUDA-SAM deposition.

4.3 Results and Discussion

4.3.1 Unpatterned Gold

Figure 4-2 depicts AFM imaging and force mapping data over a 4µm x 4µm area that exhibits several noteworthy features regarding the nature of the substrate and the lipid bilayer. Within this area there were gold grains of various size and shape along with grain boundaries. There were also regions of good bilayer coverage as well as large

- 120 -

defects in the bilayer. Phase segregation of the lipids were observed upon close examination. Analysis if these images and the force maps in this unpattered region is helpful to identifying the contributions to the data from lipids, defects, grain boundaries and phases, and will assist in interpreting the AFM and force-map data from the aperture arrays, whose patterned topography introduces additional complications. The contribution of the patterned topography can better be understood when the contributions from the unpatterned gold are understood.

The extent of lipid coverage can be observed directly from the topographic image in panel A and the phase image in panel B. However, due to the large variations in topography that arose from grain boundaries and from topographic irregularities on the individual grains, and the height scale required to visualize them, the phase segregation of the lipid layer, which produces topographic contrast on the order of 1 nm was difficult to observe in panel A. Defects in the bilayer however, which provide topographic contrast on the order of 5 nm, was much more apparent on this height scale. The phase contrast image in panel B resolved both defects in the bilayer and also phase segregation, but provided less detail than the topography in identifying grain boundaries. Defects in the bilayer were identified by examination of both panel A and panel B side-by-side. Lipids appeared as medium to dark grey in panel B, while defects and grain boundaries appeared as light grey to white. For example, the large grain that occupies most of the top two- thirds of the image appeared to be mostly covered by a lipid bilayer as seen by the grey colour it took on in panel B. A defect in the layer was also detected near (x,y) coordinate (2.1,3.6) that appears as a nearly triangular light-grey patch in panel B. Examination of the same region in panel A revealed a dark grey patch of the same shape and location, which indicates a depression of uniform depth. The lower third of the image below y = 1 µm had many defects. The defect pattern was most noticeable in the lower left corner near the origin of the image. A round feature at the (x,y) coordinate of (0.75, 0.75) µm was recognized as a lipid patch in both the height and phase images.

- 121 -

Due to the Z-scale in both images, the two lipid phases were more difficult to resolve. The lipid phases were most noticeable in the phase image in panel B. Specifically, a lipid patch at the (x,y) coordinate (2.4, 0.75) exhibited the most easily resolvable phase segregation in panel B. The extent of the lipid patch was easily resolved in both the height and phase images. Within the patch, the boundaries of the lipid phases appeared as a lighter grey than the rest of the patch. In the height scale of the topography image, the height difference between the phases was not easily resolved. As discussed in a later section, images and force maps acquired over a smaller area resolved the phase segregation with greater ease.

The extent of lipid coverage was also reported in the force maps over the same area. To produce the image in panel C, each force curve in the map was analyzed for its features using an automated algorithm, and was categorized as either a breakthrough event (that can only occur over a lipid-covered region of the substrate), a hard contact event (an indentation where no breakthrough event is detected, and is commonly associated with indentation into the gold substrate), and other events, (which do not fall into either of the first two categories). The breakthrough events, which indicate the presence of a lipid bilayer, were coloured grey, while hard contact events, which are predominantly bare- gold were coloured black, and others were assigned the colour of white. The image in panel C is mostly grey, indicating that a breakthrough event was detected over most of the area in this map, and as such, the grey features in panel C correspond with features in the height and phase images that have been identified as lipid. Indentations identified as hard contact coincidedvwell with features in height and phase that have been identified as gold. Other, unrecognizable events in the force curve seem to have appeared nearly at random.

Panel D is the breakthrough force map, which in cases where a force indentation was identified as a breakthrough event, the maximum force acquired prior to the breakthrough was recorded in this map. In cases where a breakthrough was not detected, the force

- 122 -

value of the next point of interest in the curve was recorded instead. In the case of Hard contact events, the second point is typically the end of the indentation curve (typically the trigger point for the measurement), and appears as white in the image. Here, grain boundaries were clearly identified, and the lipid-covered areas exhibited variations in their breakthrough forces that are possibly indicative of phase segregation.

With this interpretation as a baseline, the analysis of the topography, phase and force data collected over the nanoaperture arrays can be understood. In addition to AFM data, visible and near IR transmission spectra were collected over these arrays to determine the

- 123 -

effect of the bilayers on the transmission spectrum. Figures 3, 4, and 5 summarize these data for nano-hole arrays, nano-slit arrays, and annular aperture arrays respectively. Imaging and force mapping were performed over 5 µm x 5 µm areas.

4.3.2 Nano-hole arrays

The topography and phase images of the bilayer-covered hole-array are shown in Figure 3 A and B respectively. The images were captured near one edge of the patterned area so as to include some of the unpatterned area in each image for comparison with the results shown in Figure 2. As in Figure 2, Figure 3 depicts a number of gold grains featuring a variety of sizes and surface characteristics. The majority of grains were flat, or exhibited the characteristic triangular pattern typical of the (111) gold surface. Additionally, the regular square array of holes was readily discernable in the images.

A close examination of the topographic image revealed that some holes appeared deeper than others. The apparent depth of the holes may depend on whether the hole is covered by the lipid layer or not. In these images, the tapping mode phase image revealed more about the extent of lipid coverage over these structures because the density of topographic features within the array were great enough to obscure the ~ 5 nm height contrast afforded by a defect in the lipid layer. The phase image on the other hand was less coupled to topography. In the phase image, the nano-holes appeared light grey, as do other large variations in topography such as grain boundaries and a small number of grains with irregular topography. On flat grains and between holes, the bilayer-covered regions appeared dark grey while defects appeared nearly white.

From the force map data, the classification map in Figure 3C shows that the majority of the surface produces breakthrough events associated with indentation of a lipid bilayer over a rigid support. Over the nano-holes however, the force curves were predominantly

- 124 -

classified as “other,” which results in a periodic arrangement of white pixels in the classification map. And, as expected, there were black pixels assigned to hard-contact events that correspond well with the defects in the bilayer seen in the phase image. Additionally, a fourth force curve category was added to account for some of the features in the force map, namely long-distance breakthrough events, which appeared much like breakthrough events but were differentiated by their longer breakthrough distances of between 15 nm and 20 nm. These were assigned the colour of light grey, and appear primarily in and near the edges of the holes. Some breakthrough-like events were categorized in the “other” category when the distance from contact to the end of breakthrough exceed 20 nm. These longer breakthrough-like events may have been caused by the indentation into a hole covered by a lipid bilayer, where the added indentation depth comes from deformation of the lipid over-layer into the hole. However, they may also have come from interactions between some portion of the tip and the edges of the holes, where at some critical force, the tip sliped into the hole. This interpretation is speculative since it was not possible to identify with certainty the origin of these long distance breakthroughs without further study.

- 125 -

Figure 4-3: DEC221 bilayers on nanohole arrays. Regular square array of holes is discernable by AFM topography and phase (A and B respectively). Lipids are obscured by large variations in topography, but hole depths suggest lipids are spanning many holes, and defects in the bilayer appear as white features in the phase image. Force classification map (C) shows most indentations over the gold result in breakthrough events (Grey) while hard-contact events (Black) correspond to defects detectable in (B). Indentation over holes produces “Other” events. Breakthrough force map (D) cannot resolve phase segregation. Likewise, breakthrough force histogram (E) shows a broad, single-mode distribution of forces. Two modes in indentation depths (F) represent indentation over holes and over the support.

Because of the diversity of force indentation curves collected over the array, only the breakthrough forces are displayed in the force map in Figure 3D. That is, contributions from long-distance breakthroughs, hard-contact, and other events are not shown. From the breakthrough force map, it is clear that a broad distribution of breakthrough forces existed. The histogram of the breakthrough forces (Figure 3E) does not show the bimodal distribution typically associated with phase segregated lipid bilayers, as was apparent in the previous chapter. Instead, it appears nearly Gaussian in profile. Given the short incubation time, it is possible that the lipid domains were too small to be clearly resolved in the force map. Furthermore, within this 5 µm x 5 µm area there were grain boundaries, grains of various sizes and topography, as well as the array itself, which contributed to variations in the recorded forces to the extent that the domains could be resolved. In fact, when the experiment was repeated, and the measurements were made

- 126 -

over a smaller 2 µm x 2 µm area that contains as few grain boundaries as possible, the domains structure became clearly distinguishable in the force map. These results are discussed later.

Force maps at this scale were utilized to examine the difference in mechanical properties between lipids supported directly by the gold substrate, and those that spaned the holes. The recorded indentation depth was used to separate their contributions to the force map: Indentation performed over the holes were deeper than those performed over the gold support. A histogram of indentation depths was plotted in Figure 3F. The histogram reveals at least 3 clear populations of indentation depths. The population of indentations with depths centered at 0 nm correspond to hard-contact events where no lipid is present. The population of indentations with indentations that ranged from 3nm to 11nm likely corresponds to indentation performed over a supported bilayer. Indentations greater than 11 nm are likely those that occured over holes or grain boundaries. By sorting the data based on the depth of indentation, indentation curves performed over holes can be separated from those performed over the gold support. The modulus maps and histograms recorded over holes and over gold are presented in Figure 4.

- 127 -

Figures 4A and C are the modulus map and histogram respectively for force curves with indentation depths between 3nm and 11nm, while Figures 4B and D are the respectively the modulus map and histogram for indentations that reached depths ranging from 11nm to 30nm. The limits of demarcation were determined from the histogram of indentation depths in Figure 3F. Figures 4A and 4B correspond well with the pattern of gold and of holes respectively. Log-normal fits of the histograms (black lines in 4C and 4D) were used to determine the average modulus value for these two populations of curves. The average modulus of lipid supported by gold was 84MPa, while the average modulus of the lipids spanning holes was 12 MPa. The factor of 7 difference in the average modulus value was attributed in part to the presence of the solid support, which according to the analysis in the previous chapter could account for a factor of at least 3. The remaining difference may be a result of differences in the modes of deformation available to the lipid layer in the two cases. In the case of the gold-supported lipids, the primary mode of

- 128 -

deformation under the applied load was compression of the alkane chains of the lipid molecules, while in the case of the lipid over the nano-holes, bending modes may have accounted for some portion of the deformation under load. The model used to calculate the modulus values here does not account for bending modes of the bilayer.

In order to observe the phase segregation of the lipids, the experiment was repeated with an emphasis on imaging and force analysis over a smaller area. Using the same number of points for force mapping over a smaller area leads to better spatial resolution. Furthermore, by focusing on a single grain, fewer topographic inconsistencies due to variations from grain to grain or from grain boundaries were observed. Figure 4-5 depicts AFM images and force map for a 2 µm x 2 µm area which predominantly contained a single grain, as well as some grain boundaries and aperture that were part of the nano-hole array. Close examination of the topography in Figure 4-5A revealed some contrast between co-existing lipid phases. The phase image in Figure 4-5B however did not bare much contrast between the two phases. The lipid domains were better resolved in the force maps, however. The Young’s modulus map in Figure 4-5C weakly revealed the domains, while the pattern of the domains was much clearer in the breakthrough force map of Figure 4-5D. Based on previous work, the light grey patches were identified as the lo domain with higher breakthrough forces, while the dark grey regions were identified as the softer, more fluid ld matrix. The identification of the lipid domains could be made over the holes since the force curves were rarely identified as breakthrough events.

Despite being able to differentiate between two domains in the modulus and force maps, neither histogram (Figure 4-5E and F) exhibited the well-separated peaks typical of well- separated phases. Perhaps the short 30 minute incubation time at room temperature was not sufficient for the lipid components to thoroughly partition. That is, a significant fraction of DOPC remained in the lo phase, and likewise, a significant amount of SM existed in the ld phase. This would serve to soften the lo phase and stiffen the ld phase to

- 129 -

the point where their peaks overlaped, obscuring their distinction in the histogram. This may also account for why defined populations were not clearly identified in histogams derived from the 5 µm x 5 µm scan.

- 130 -

breakthrough force map (D) clearly show phase segregation. Modulus histogram (E) appears as single mode. Breakthrough force histogram (F) shows two overlapping modes.

Nevertheless, it was possible to choose a threshold breakthrough force by which to sort the data so as to isolate the data that corresponded to each domain. It was found that a threshold breakthrough force value of 1.75 nN could be used to divide the data. The resulting Young’s modulus map for indentation curves with breakthrough force values greater than 1.75 nN is shown in figure Figure 4-6A, while the modulus map of events where the breakthrough force is less than 1.75 nN is shown in Figure 4-6B. The features in Figure 4-6A correspond well with the light grey lo features in the breakthrough force map while the features in Figure 4-6B follow those of the dark grey ld features.

- 131 -

Histograms of the modulus values are shown in Figure 4-6C and D. Log-normal fits to the distributions (black line) were used to find the average modulus value for each phase.

The average Young’s modulus for the lo domain was 64 MPa, while the average modulus of the ld phase was 50 MPa. As expected, the lo phase exhibited a higher modulus than the ld phase. In this case, the difference was small and is likely due to incomplete phase segregation. Overall, these modulus values were lower than those recorded on the 5 µm x 5 µm scan.

- 132 -

In the process of coating this nano-hole array, a nano-slit array on the same substrate was simultaneously coated. The same set of measurements were performed on the nano-slit array, and the results are discussed in the next section.

4.3.3 Nano-slit array

The topographic image of the slit-array reveals a large number of smaller grains, and therefore more grain boundaries over a 5 µm x 5 µm area than was the case for the hole- array in Figure 4-3. As a result, the presence of the lipid layer is difficult to ascertain from either the topography or the phase images in Figure 4-7A and B respectively. Accordingly, phase segregation was not immediately evident in either image. The classification map shown in Figure 4-7C, generated from the force map data, suggests that breakthrough events were recorded throughout the scan area, indicating that a lipid bilayer had been cast over the array. Breakthrough events occured primarily over the gold supports, while “other” force events occured predominantly over the milled slits, generating a periodic line pattern in the classification map that clearly follows the slit- array pattern seen in the topography, albeit with a slight tilt in scan-angle. This result was predicted from the pattern formed in the classification map taken over the hole-array, where breakthrough events were recorded when indentation occurs over a supported region of the bilayer, while non-breakthrough events occured over the milled features. There were also a large number of defects distributed across the scan area that may have resulted from the large number of grain boundaries in this region of the sample.

The breakthrough force map in panel D and associated histogram in panel E reveal a broad distribution of breakthrough forces as well as many irregularities in the map. It could be argued that there were some small patches in the map where breakthrough forces appeared higher than the surrounding regions, though the presence of coexisting phases could not be determined with certainty. As was the case with the hole-array, no distinct populations of breakthrough forces emerged in the histogram. Conversely,

- 133 -

distinct populations were identifiable in the indentation depth histograms as was the case with the 5 µm scan of the hole-arrays. As before, the narrowly distributed population of force curves with indentations between 2 and 10 nm were attributed to indentation of a supported bilayer, while the very broadly distributed population of curves with very large indentations were attributed to indentations over the slits.

By sorting the force curves based by indentation depth, populations of force curves performed over gold and over the slits were separated and analyzed independently. Figure 4-8 shows the results of the separation. Modulus values from force curves with

- 134 -

indentations between 2 nm and 7.5 nm were mapped in panel A with the corresponding histogram shown in panel C. The modulus map and associated histogram for force curves with indentation depths greater than 7.5nm were plotted in panels B and D respectively. As expected the pattern in panel A corresponds well with the pattern of gold in the slit array, while the pattern in panel B follows the slits themselves.

Figure 4-8: Modulus maps and associated histograms of nanoslit array separated by depth of indentation. Shallow indentations (A,C) indicate indentation over supported lipid; fitted peak at 29 MPa. Deep indentations (B,D) indicate indentations over apertures; fitted peak at 7.6 MPa.

As before, the peak modulus value from the histograms was greater when the bilayer was supported directly by the rigid substrate. The peak position in the distribution of moduli for the gold-supported bilayer was 29 MPa, while the peak position of the modulus over the slits was 7.6 MPa. Also, as in the case of the hole-array data, the distribution of modulus values in the supported bilayers was wider than those recorded over the slits. Since the presence of the substrate support affects the apparent modulus values, it follows

- 135 -

that variations in the support will lead to variations in the apparent modulus. Therefore, the variety of topographic features that resulted from the large number of grains and grain boundaries in this 25 µm2 area likely contributed to a broadening of the distributions compared to data collected over a single, smooth grain. Furthermore, large variations in the substrate were not accounted for in the modulus-fitting model. Over the slits, the distribution of modulus values was narrower.

Both the modulus values recorded over the supported bilayer and over the slits were significantly lower than those recorded over the hole arrays despite the bilayer having been deposited simultaneously. The difference in modulus values could have arisen from a difference in time between measurements: While the force map over the slit-array began about 9 hours after the deposition and rinsing of the bilayer, an additional 6 hours elapsed between the beginning of the slit-array force map and the beginning of the hole- array force map during which the slit array force map, and imaging took place. At the manufacturer’s stated operating temperature of the AFM, which is 27 °C, it is possible that some ripening and annealing of the bilayer occurred during this time.

4.3.4 Annular Aperture Arrays

The experiment was repeated on another substrate to obtain the same set of measurements on an annular aperture array. Figure 4-9A and B are the topography and phase images of the AAA respectively. The presence of the bilayer was difficult to confirm from either image due to the lack of any obvious defects aside from the features of the annuli themselves. The force classification map in panel C however, confirmed that many breakthrough events were recorded over this area. The breakthrough force map in panel D however did not reveal clear evidence of phase coexistence. Likewise, the histogram of breakthrough forces (Figure 4-9E) did not show a clear bimodal distribution that would be expected if distinct phases were present. However, it was shown above that separation

- 136 -

of the two phases on the breakthrough histogram is not necessary for coexisting phases to be resolved on the force map.

Figure 4-9: DEC221 bilayers on annular aperture arrays. Regular square array of annuli is discernable by AFM topography and phase (A and B respectively). Lipid phases are likely obscured by large variations in topography, but hole depths suggest lipids are spanning many holes. Force classification map (C) shows most indentations over the gold result in breakthrough events (Grey) while hard-contact events (Black) correspond to defects. Indentation over grain boundaries produces “Other” events. Breakthrough force map (D) cannot resolve phase segregation. Likewise, breakthrough force histogram (E) shows a broad, single- mode distribution of forces. Two modes in indentation depths (F) represent indentation over holes and over the support.

Unlike previous cases with slit-arrays and hole-arrays, the indentation depth histogram shown in Figure 4-9F did not clearly indicate two populations of force indentation depths that can be attributed to indentations over supported bilayers and over aperture-spanning bilayers. This was because the annuli in the AAA were larger than the holes in the holes arrays, and as such, the density of features in the same sized scan area is lower in the former. Hence, fewer indentations occured over aperture over the same sized area. Nevertheless, dividing the populations of force curves by indentation depth in the same

- 137 -

manner as before yielded the same basic result: the modulus map of shallow indentations (Figure 4-10A) followed the pattern of gold in the array, while the modulus map of deep indentations (Figure 4-10B), with its ring-like features, followed the pattern of the apertures in the array. Histograms of the modulus values were fitted to log-normal functions to yield mean moduli of 5 MPa and 20 MPa over apertures and over gold, respectively.

These values for moduli are on the same order as those derived from the slit-array force maps, even though this sample endured nearly 12 hours of scanning and force indentations before this force map was recorded. Although variations in the properties of the bilayer are likely to occur from one film to the next, this may suggest that factors such as time and temperature alone are insufficient to explain all of the differences in the

- 138 -

observed moduli for the bilayer on the hole-array and the slit-array discussed above. A systematic study of bilayer properties under controlled temperatures and times, would be necessary to determine more precisely, the role that these factors play in the properties of the bilayers. Such studies however are beyond the scope here.

The goals of this project are to ascertain whether the lipid bilayer is deposited on the substrate, show that they are present on the various nanoaperture systems prior to testing their transmission spectra, and demonstrate phase–segregation of this lipid layer on an array. The images and force maps over the nano-hole arrays, nano-slit arrays, and annular aperture arrays provide sufficient evidence of the bilayer formation over these arrays. Since centimeter-scale surfaces can be treated and coated by membranes, multiple arrays fabricated on the same substrate were functionalized simultaneously, as was demonstrated with the hole-array and slit-array above.

Phase segregation of the lipid layer, which is important for mimicking lipid rafts, was demonstrated over nano-hole arrays. These features were resolvable when a smaller scan area was found that contains a large flat grain. This was necessary because inconsistencies in topography that resulted from scanning a collection of gold grains and grain boundaries led to a broad distribution of breakthrough forces that obscured the distinction between the two lipid phases.

4.3.5 Transmission spectra

Visible and near IR spectra were collected for each of the arrays shown in Figures 3, 7 and 9 before and after lipid deposition. All spectra were recorded in water. Solid grey lines are the MUDA functionalized arrays in water, while the black solid lines are the spectra recorded after the lipid deposition.

- 139 -

Spectra for the hole array shown in Figure 3 are shown in Figure 4-11A and B. The full band spectra are shown in the inset for completeness. In the visible transmission spectrum, only the narrow band between 700 nm and 800 nm appeared to shift due to the presence of the bilayer (Figure 4-11A). This band was attributed to a higher order SPP- Bloch-wave mode on the Au/solution interface. The other features below 700 nm were attributed to modes of the Au/quartz interface, since they were unaffected by the bilayer. The wavelength of the peak was found by fitting the peak to a Gaussian (markers) and taking the centroid of the fit. The peak position shifted from 748 nm to 752 nm following bilayer deposition. The NIR spectrum in Figure 4-11B exhibited similar behaviour. Only the peak from 1400 nm to 1600 nm shifted due to the presence of the bilayer. This band can be attributed to the lowest-order SPP-Bloch wave. The bandwidth and position is nearly exactly double that of the band in the visible spectrum. In this case, the peak position was measured to shift from 1493 nm to 1503 nm.

Notably, structured transmission spectra were still obtained despite the relative roughness from the grain boundaries of the flame annealed gold over areas of tens of micrometers over which the arrays reside. Also, at over 300 nm thick, these gold films were much thicker than most reported in literature, and unsurprisingly, the transmission efficiency was low. Another consequence of the thick gold film was that the aspect ratio of features tends to suffer, since FIB-milled features tend to taper with depth.

- 140 -

Figure 4-11: Visible (left side) and NIR (Right side) transmission spectra through the same nanohole array (A,B), nanoslit array (C,D), and annular aperture array (E,F) used for force mapping before (grey line) and after (black line) bilayer formation (confirmed by force mapping). For nanohole arrays (A,B), peak position is determined by fitting to Gaussian (markers). The Vis peak (A) red-shifts 4 nm, while the NIR peak (B) red- shifts 6 nm. Broad-band transmission spectra shown in inset.

Figure 4-11C and D are the visible and NIR spectra for the slit-arrays. Unfortunately, without using a polarized light source, it was not possible to separate the contributions to the spectrum from the 1D SPP-BW and from the propagating modes allowed by the slits themselves. At the time of the experiment, a broadband polarizer was not available for integration with the microscope. Nevertheless, a peak was observed in the NIR spectrum of the slit-array without the lipid bilayer (Figure 4-11D, grey line) at a wavelength of 1590 nm that can be discerned above the transmission of the allowed propagating mode. Upon deposition of the bilayer (black line), this band is red shifted such that the peak was

- 141 -

no longer detectable by the spectrometer, and the peak was no longer discernable above the direct transmission mode.

For the annular aperture arrays, the lone peak observed in the visible transmission spectrum (Figure 4-11E) near 680 nm likely corresponds to a Bloch mode on the Au/glass interface, and was not observed to shift in the presence of the bilayer. Likewise, the peak at 1390 nm in the NIR spectrum (Figure 4-11F) is also likely to be a Bloch- mode on the Au/glass interface. There was a peak around 1590 nm in the absence of the bilayer (grey line) that likely corresponds to a Bloch mode on the Au/solution interface. Upon deposition of the bilayer (black line) the mode red-shifted beyond the maximum detectable wavelength of the spectrometer and thus the peak appeared to have disappeared. The slight upward slope of the transmission spectrum at wavelengths beyond 1560 nm resembles the blue edge of a peak that might be centered just beyond the maximum detectable wavelength of the spectrometer, and may well be the same mode that was observed before lipid deposition.

4.4 Conclusions and Perspectives

These results have confirmed the deposition of lipid bilayers on plasmonic nanohole arrays, nanoslit arrays and annular aperture arrays. The presence of the bilayer was confirmed by the presence of breakthrough forces over a large portion of each array, and by transmission spectra of the arrays that report changes in the refractive index in the environment of the array. At large scan sizes, variations in topography from the various gold grains, the grain boundaries, and the apertures themselves obscured the relatively small difference in height expected between the two co-existing lipid phases. At smaller scan sizes however, the coexisting lipid phases were resolvable in topography, and by contrast in Young’s modulus and breakthrough force maps.

- 142 -

A small scan area was chosen that included a single gold grain, and a smaller fraction of grain boundaries than was possible for larger scans. This ensured fewer and less extreme variations in topography that allowed the domains to be resolved. While only just noticeable in the topography, and nearly indistinguishable in the phase image, the domains were clearly discernable by force map: The lo domain had higher average breakthrough forces and moduli than the ld matrix. Histograms of the modulus and breakthrough force however did not show the characteristic bimodal distribution canonically expected for phase-segregated lipids. The breakthrough force histogram however appeared as a broad peak with a clear shoulder, indicating two overlapping populations of forces. Dividing the data into two bins by breakthrough force value yielded two populations of force curves, which when plotted as separate modulus maps each represented primarily one of the two phases visible in the full modulus map. Though the separation was imperfect, average moduli were still determined from the histograms.

The lo domain had an average modulus of 64 MPa while the ld matrix averaged 50 MPa.

The greater stiffness of the lo domains compared to the ld matrix was consistent with expectations both theoretical and practical.

Though phase segregation was not detected in larger scans, contrast existed between the modulus and force values collected over lipids that were directly supported by the gold substrate and those that may have spaned the apertures. These modulus contributions were separated by sorting the data into two bins depending on indentation depth, with deeper indentations indicating an aperture. Though modulus values varied from one array to another, moduli over the directly supported membranes were higher than aperture-spanning bilayers in all cases. This was not surprising since the supporting substrate is known to increase the apparent modulus. It is additionally notable that the validity of the Sneddon model, which does not account for bending of the bilayer, is questionable for the analysis of pore-spanning lipids.

- 143 -

Spectroscopy through the arrays further confirmed the deposition of the bilayers and provided insight into the effect of the bilayer on optical properties of the arrays. While the transmission spectra from the slit arrays proved difficult to interpret due to the strong presence of propagating modes of the array in the spectra, a possible Bloch mode resonance appeared in the NIR spectrum. This peak appeared to red-shift out of the range of the spectrometer after lipid deposition. A Bloch mode of the Au/solution interface of the annular aperture array was also visible in the NIR spectrum. The peak red-shifted out of the range of the spectrometer following lipid deposition – a shift of at least 10 nm. The spectra from the hole arrays were easier to interpret. Bloch modes from the Au/glass interface and the Au/solution interface were both visible, with the Au/glass modes identified by their immobility following lipid deposition. The visible Au/solution mode and the NIR Au/solution mode shifted 4 nm and 6 nm respectively following lipid deposition.

Taken together these data have demonstrated how the methods of the previous chapter have been applied to producing phase-segregated lipids on various plasmonic aperture arrays, as well as the ability of functionalize multiple arrays on the same substrate simultaneously. Observing phase segregation of lipids on arrays is an important to future applications of model lipid systems as more biologically relevant cell membrane mimics in biosensors that can be used to probe the interactions between soluble analytes and lipid rafts. For example, lipid rafts have been shown to play an important role in the amyloidosis of the Aß peptide and oligomers associated with Alzheimer’s disease10. A sensor consisting of a nanoaperture array coated with a model raft-forming bilayer could be valuable for studying the kinetics of Aß adsorption and aggregation. Another natural extension of this work would be to incorporate membrane-bound proteins, or lipid- anchored peptides to elicit and detect the specific interaction of a target analyte with the membrane. This is potentially valuable for example, in the discovery and characterization of pharmaceuticals, which inevitably interact with cell surfaces4, or for exploring the combined role of the lipid raft40, and cell surface proteins such as CD441, and CCR512 on the mechanisms of HIV infection.

- 144 -

4.5 Acknowledgements

The author acknowledges the contributions of Matthew Kofke and the University of Pittsburgh NanoScale Fabrication and Characterization Facility for the FIB milling of all the nanoaperture arrays shown. The author also acknowledges the University of Toronto Department of Chemistry Machine Shop for fabrication of the custom fluid cell.

4.6 References

1. Glasmästar, K., Larsson, C., Höök, F. & Kasemo, B. Protein Adsorption on Supported Phospholipid Bilayers. Journal of Colloid and Interface Science 246, 40- 47 (2002).

2. Kono, K., Ito, Y., Kimura, S. & Imanishi, Y. Platelet adhesion on to polyamide microcapsules coated with lipid bilayer membrane. Biomaterials 10, 455-461 (1989).

3. Lee, T., Mozsolits, H. & Aguilar, M. Measurement of the affinity of melittin for zwitterionic and anionic membranes using immobilized lipid biosensors. The Journal of Peptide Research 58, 464-476 (2001).

4. Cooper, M.A. Optical biosensors in drug discovery. Nat Rev Drug Discov 1, 515-528 (2002).

5. Watts, T.H., Gaub, H.E. & McConnell, H.M. T-cell-mediated association of peptide antigen and major histocompatibility complex protein detected by energy transfer in an evanescent wave-field. Nature 320, 179-181 (1986).

6. Nikolelis, D.P., Hianik, T. & Krull, U.J. Biosensors based on thin lipid films and liposomes. Electroanalysis 11, 7-15 (1999).

7. Reimhult, E. & Kumar, K. Membrane biosensor platforms using nano- and microporous supports. Trends Biotechnol 26, 82-89 (2008).

- 145 -

8. Sackmann, E. Supported membranes: scientific and practical applications. Science 271, 43-48 (1996).

9. Nikolelis, D.P. & Krull, U.J. Establishment and control of artificial ion-conductive zones for lipid membrane biosensor development. Analytica Chimica Acta 257, 239- 245 (1992).

10. Kremer J.J. & Murphy R.M.[1] Kinetics of adsorption of beta-amyloid peptide Abeta(1-40) to lipid bilayers. Journal of Biochemical and Biophysical Methods 57, 159-169 (2003).

11. Wallin, E. & Heijne, G.V. Genome-wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms. Protein Science 7, 1029-1038 (1998).

12. Hoffman, T.L., Canziani, G., Jia, L., Rucker, J. & Doms, R.W. A biosensor assay for studying ligand-membrane receptor interactions: Binding of antibodies and HIV-1 Env to chemokine receptors. Proceedings of the National Academy of Sciences of the United States of America 97, 11215 -11220 (2000).

13. Mossman, K.D., Campi, G., Groves, J.T. & Dustin, M.L. Altered TCR Signaling from Geometrically Repatterned Immunological Synapses. Science 310, 1191-1193 (2005).

14. Jonsson, M.P., Jonsson, P., Dahlin, A.B. & Hook, F. Supported lipid bilayer formation and lipid-membrane-mediated biorecognition reactions studied with a new nanoplasmonic sensor template. Nano Lett. 7, 3462-3468 (2007).

15. Nam, J., Nair, P.M., Neve, R.M., Gray, J.W. & Groves, J.T. A Fluid Membrane- Based Soluble Ligand-Display System for Live-Cell Assays. Chembiochem 7, 436- 440 (2006).

16. Richter, R.P. & Brisson, A.R. Following the Formation of Supported Lipid Bilayers on Mica: A Study Combining AFM, QCM-D, and Ellipsometry. Biophys. J. 88, 3422-3433 (2005).

- 146 -

17. Johnson, J.M., Ha, T., Chu, S. & Boxer, S.G. Early Steps of Supported Bilayer Formation Probed by Single Vesicle Fluorescence Assays. Biophys. J. 83, 3371-3379 (2002).

18. Johnston, L.J. Nanoscale Imaging of Domains in Supported Lipid Membranes. Langmuir 23, 5886-5895 (2007).

19. Sullan, R.M.A., Li, J.K. & Zou, S. Direct Correlation of Structures and Nanomechanical Properties of Multicomponent Lipid Bilayers. Langmuir 25, 7471- 7477 (2009).

20. Steinem, C., Janshoff, A., Ulrich, W.P., Sieber, M. & Galla, H.J. Impedance analysis of supported lipid bilayer membranes: a scrutiny of different preparation techniques. BBA-Biomembranes 1279, 169-180 (1996).

21. Naumann, C.A. et al. The Polymer-Supported Phospholipid Bilayer: Tethering as a New Approach to Substrate−Membrane Stabilization. Biomacromolecules 3, 27-35 (2002).

22. Naumann, R. et al. Tethered Lipid Bilayers on Ultraflat Gold Surfaces. Langmuir 19, 5435-5443 (2003).

23. Plant, A.L., Brighamburke, M., Petrella, E.C. & Oshannessy, D.J. Phospholipid/Alkanethiol Bilayers for Cell-Surface Receptor Studies by Surface Plasmon Resonance. Anal. Biochem. 226, 342-348 (1995).

24. Stelzle, M., Weissmueller, G. & Sackmann, E. On the application of supported bilayers as receptive layers for biosensors with electrical detection. The Journal of Physical Chemistry 97, 2974-2981 (1993).

25. Simons, K. & Vaz, W.L. MODEL SYSTEMS, LIPID RAFTS, AND CELL MEMBRANES1. Annu. Rev. Biophys. Biomol. Struct. 33, 269-295 (2004).

26. Ebbesen, T.W., Lezec, H.J., Ghaemi, H.F., Thio, T. & Wolff, P. Extraordinary optical transmission through sub-wavelength hole arrays. Nature 391, 667-669 (1998).

- 147 -

27. Krishnan, A. et al. Evanescently coupled resonance in surface plasmon enhanced transmission. Optics Communications 200, 1-7 (2001).

28. Brolo, A., Gordon, R., Leathem, B. & Kavanagh, K. Surface Plasmon Sensor Based on the Enhanced Light Transmission through Arrays of Nanoholes in Gold Films. Langmuir 20, 4813-4815 (2004).

29. Chang, S., Gray, S. & Schatz, G. Surface plasmon generation and light transmission by isolated nanoholes and arrays of nanoholes in thin metal films. Opt. Express 13, 3150-3165 (2005).

30. Gordon, R., Sinton, D., Kavanagh, K.L. & Brolo, A.G. A New Generation of Sensors Based on Extraordinary Optical Transmission. Accounts of Chemical Research 41, 1049-1057 (2008).

31. De Leebeeck, A. et al. On-Chip Surface-Based Detection with Nanohole Arrays. Analytical Chemistry 79, 4094-4100 (2007).

32. Tetz, K.A., Pang, L. & Fainman, Y. High-resolution surface plasmon resonance sensor based on linewidth-optimized nanohole array transmittance. Opt. Lett. 31, 1528-1530 (2006).

33. Steltenkamp, S. et al. Mechanical Properties of Pore-Spanning Lipid Bilayers Probed by Atomic Force Microscopy. Biophys. J. 91, 217-226 (2006).

34. Sun, Z., Jung, Y.S. & Kim, H.K. Dynamic evolution of surface plasmon resonances in metallic nanoslit arrays. Appl. Phys. Lett. 86, 023111 (2005).

35. Lee, K., Lee, C., Wang, W. & Wei, P. Sensitive biosensor array using surface plasmon resonance on metallic nanoslits. J. Biomed. Opt. 12, 044023-5 (2007).

36. Jung, Y.S. et al. High-sensitivity surface plasmon resonance spectroscopy based on a metal nanoslit array. Appl. Phys. Lett. 88, 243105 (2006).

37. Baida, F.I. & Van Labeke, D. Light transmission by subwavelength annular aperture

- 148 -

arrays in metallic films. Optics Communications 209, 17-22 (2002).

38. Kofke, M.J., Waldeck, D.H., Fakhraai, Z., Ip, S. & Walker, G.C. The effect of periodicity on the extraordinary optical transmission of annular aperture arrays. Appl. Phys. Lett. 94, 023104 (2009).

39. Orbons, S.M. & Roberts, A. Resonance and extraordinary transmission in annular aperture arrays. Opt. Express 14, 12623-12628 (2006).

40. Campbell, S.M., Crowe, S.M. & Mak, J. Lipid rafts and HIV-1: from viral entry to assembly of progeny virions. J. Clin. Virol. 22, 217-227 (2001).

41. Siliciano, R.F. The role of CD4 in HIV envelope-mediated pathogenesis. Curr. Top. Microbiol. Immunol 205, 159-79 (1996).

- 149 -

Chapter 5: Incorporation of membrane proteins extracted from cell cultures into artificial bilayers

The addition of membrane proteins to artificial supported bilayers more closely mimics cell membranes and augments potential analyte targets for biosensing applications. A method of preparing supported lipid bilayers containing natural membrane proteins is demonstrated. Membrane proteins are extracted from a lung cancer cell line using a commercial kit and incorporated into artificial lipid vesicles and cast onto mica substrates as bilayers. The protein-containing lipid bilayers are compared to protein-free lipid bilayers by AFM imaging and force mapping. Proteins are found to be abundant and mobile within the bilayer, and do not impede lipid phase segregation. However proteins appear to cause tip contamination. Nevertheless, an application is suggested whereby cancer-cell proteins are cast in a bilayer on a sensor surface and used to screen cancer- targeting drugs or peptides.

5.1 Introduction

The ability to incorporate membrane proteins into supported lipid bilayers can broaden their potential applications as functional interfaces in planar biosensors because in addition to presenting a more biologically relevant mimic of a natural cell membrane, the presence of membrane proteins increases the number of potential analytes that can be captured on the surface through interactions between the proteins and specific analytes. The previous chapters demonstrate the deposition of phase-segregated model lipid mixtures onto gold surfaces relevant for surface plasmon resonance and electrochemical sensing, and more specifically, onto plasmon-active nanoaperture arrays. Without proteins, these bilayers represent a model system that mimics the formation of lipid rafts1,2, and are potentially applicable to detecting the interaction between soluble

- 150 -

analytes such as drug molecules, peptides, viruses, etc, to the lipids themselves3‐6. Additionally, the lipid bilayer acts as a biocompatibility layer7‐10.

This chapter builds upon the work of the previous chapter by demonstrating the incorporation into artificial bilayers, of membrane proteins extracted from cells via a commercially available membrane protein extraction kit. Specifically, cultured cells from the lung cancer cell line A549 were used as a source for membrane proteins. The eluate from the kit, which according to the manufacturer, contains plasma membrane proteins bound to vesicles composed of plasma membrane lipids, are incorporated with DOPC/Egg Sphingomyelin/Cholesterol mixtures (DEC221) by repeated freeze-thaw cycles, and refined by extrusion. The resulting vesicles are cast onto mica substrates and characterized by AFM. When compared to pure DEC221 bilayers, the DEC-membrane- protein bilayers (DEC-MP) have small features throughout the bilayer that suggest the presence of proteins. Force maps of the DEC-MP bilayers suggest rampant tip contamination, possibly by the proteins themselves.

The A549 tumor cell line, from which the proteins are extracted, originate from a human alveolar cell carcinoma11,12. As models for Type II pulmonary epithelial cells and non- small-cell lung carcinoma, cultures of this cell line is often used in tests of peptide and drug molecule uptake13, as well as for the production of pulmonary surfactant11,12. Like many cancer cells, A549 cells over-express the epidermal growth factor receptor, and may be used as a model for targeting EGFR-over-expressing tumor cells in general13. The specific cells used in these studies have been mutated to co-express GFP (Green Fluorescent Protein) with EGFR as a fusion protein. The extension of this work would involve casting these bilayers onto planar biosensor surfaces, such as nanoaperture arrays. Biosensors functionalized with cancer-cell surface-markers could serve as a model system for testing the binding of peptide fragments or small molecules for potential use as cancer-targeting moieties for diagnostics.

- 151 -

5.2 Materials and Methods

Dioleoylphosphatidylcholine/Egg sphingomyelin/Ovine Cholesterol (DEC221) aliquots: Dioleoylphosphatidylcholine (DOPC) in chloroform, lyophilized egg-sphingomyelin (ESM), and lyophilized ovine cholesterol (Chol) were purchased from Avanti Polar Lipids (Alabaster, Al) and stored at -20 °C until use. For aliquots, DOPC (used directly), ESM dissolved in a 2:1 mixture of chloroform and methanol, Chol dissolved in chloroform are mixed to a 2:2:1 molar ratio and a final mass of 10.7 mg. The mixture was divided into 10 aliquots in glass vials and dried under a stream of Argon gas distributed by a manifold for 1h or until the lipids form a film on the bottom of the vials. The vials are then dried under vacuum overnight to remove residual solvent. The dried lipid films are repressurized with Argon gas, capped, and sealed with tape and stored at - 20 °C until use.

Cell cultures: Cultures of the alveolar carcinoma cell line A549 were a gift from the Gang Zheng lab. The cells have been transfected to co-express green fluorescent protein (GFP) and the epidermal growth factor receptor (EGFR) as a fusion protein.

Membrane Protein Extraction: Membrane protiens were extracted using the Qiagen Qproteome plasma membrane protein kit, following the kit instructions. Though few details about the kit’s contents or mechanism were provided by the manufacturer, the procedure involved cell collection from culture flask, cleaning by several centrifugation and resuspension steps, cell lysis, and collection of the cell lysate. The lysate, containing all cell components, was pelleted and resuspended in a buffer, which labels an unspecified membrane component (and associated vesicle required for solubilization). The labeled membrane component, was attached to magnetic beads and separated from other components by repeatedly magnetizing the beads to the wall of a microcentrifuge tube, and removing the remaining liquid, followed by resuspension of the beads and

- 152 -

bound-protein. Eventually, the bound-membrane proteins (and associated vesicle) were released from the bead by repeated elution steps where proteins were released from beads and the beads were separated to the walls of the microcentrifuge tube by magnetic field, and the remaining elution solution retained. According to the manufacturer, the eluate contains plasma membrane proteins in vesicles of plasma membrane lipids. For storage, the final eluate was divided into 10 aliquots and flash-frozen in liquid nitrogen and stored at -20 °C until use.

Membrane Proteins in DEC221 Vesicles: DEC221 aliquots were thawed and rehydrated in PBS buffer to form large multilamellar vesicles (MLVs). An aliquot of the membrane proteins was then thawed and mixed with the DEC221 MLV suspension. The quantity of PBS used was such that the final concentration of the DEC221 lipids in the PBS/Eluate suspension was 1mg/mL. The DEC221/Membrane Protein extracts (DEC-MP) were homogenized by repeated freeze-thaw cycles: The DEC-MP suspension wasvflash- frozen in liquid nitrogen and then allowed to thaw at room temperature (~20min) and repeated 3 to 5 times14. The resulting DEC-MP MLVs were extruded at 48 °C using the Avanti Mini Extruder (Avanti Polar Lipids), through 100 nm polycarbonate membranes (Whatman, Piscataway, NJ, USA). The suspension was passed back and forth through the polycarbonate filter at least 15 times, and always an odd number of times. This produces 100 nm unilamellar vesicles (ULVs).

Mica substrates: Mica substrates were prepared by gluing 25 mm diameter top-grade mica discs (Ted Pella, Redding, CA, USA) to 30 mm round glass cover slips (Fisher Scientific, Ottawa, ON, Canada) using Norland (Cranbury, NJ) optical adhesive formula 88 (NOA88), and cured under UV light. The substrates were then cleaved using sharp tweezers or a razor blade, and mounted in a fluorocarbon AFM fluid cell (Asylum Research, Santa Barbara CA), and immediately covered with PBS buffer.

- 153 -

Vesicle Deposition: Supported lipid bilayers were formed by vesicle fusion. The DEC- MP ULV solution was preheated to 44 °C. The mica substrates were brought to temperature by risning with several volumes of PBS buffer at 44 °C. The PBS was discarded, leaving a thin hydration layer on the mica, and 100uL (enough to cover the mica surface) of the DEC-MP ULV suspenstion was applied immediately to the mica surface. The suspension was left to incubate with the substrate covered at room temperature for 30 min, then rinsed gently with room temperature PBS buffer. For some measurements, the PBS was later exchanged for MilliQ water. Note that no divalent cations were used in deposition. Divalent cations can reverse the effective surface charge of the mica to be net-positive. In water, this allows charge to be induced in the zwitterionic lipid headgroup, which assists lipid deposition15. However, in PBS buffer, the surface charge reversal attracts the large phosphate groups, which sterically block adsorption of the vesicles to the surface16. In total, seven DEC-MP bilayers were deposited at various times on three different mica substrates, cleaving the mica before each deposition.

AFM imaging and force mapping: AFM imaging and force mapping were performed as before using an Asylum Research (Santa Barbara, CA, USA) MFP-3D. Images were collected using magnetic tapping mode with specialized metal-coated tips. Force mapping was employed as described in the previous chapter, and analyzed by custom software described previously. The software was optimized to better detect double breakthrough events, when they were found to be common among the data. Force curves are classified as hard-contact (no breakthrough), Breakthrough, Double Breakthrough, and Other, depending on the feature detected in the indentation data. The breakthrough force was recorded for each breakthrough event, and the first breakthrough force was recorded for each double breakthrough event. Force vs indentation curves were fit to the Sneddon model as described previously to obtain the Young’s modulus as a fitting parameter.

- 154 -

5.3 Results and Discussion

The A549 cells provided by collaborators had been transfected with a gene to express GFP-EGFR fusion proteins. Consequently, the extracted proteoliposomes appeared fluorescent, confirming the successful extraction of plasma membrane components. The fluorescence spectrum of the eluate is shown in Figure 5-1.

Figure 5-1: Normalized fluorescence excitation (grey) and emission (black) spectra of GFP-EGFR containing cell extracts

The proteolipisomes were mixed with the multilamellar DEC221 vesicles using the widely published freeze-thaw method and then extruded. The resulting unilamellar DEC221-membrane-protein (DEC-MP) vesicles were cast on mica substrates for imaging. Figure 5-2 compares the resulting supported lipid bilayers cast from pure DEC221 (panel A), with those cast from the DEC-MP (panel B). The scan size is the same in both images. The light-grey features are the lo domains. The proteins appear as uniformly distributed specks in panel B. The addition of the proteins does not appear to affect the phase segregation, with the lo domains in both images appearing as rounded, irregular shapes.

- 155 -

Figure 5-2: AFM magnetic tapping mode topography of pure DEC221 bilayer (A), and DEC + Membrane proteins (B, C). A and B are 5 µm x 5 µm scans and panel C is a 1 µm x 1 µm scan. Raft-like lo domains appear as light grey irregular, but rounded structures in a (dark grey) ld matrix in A and B. Proteins appear as small white spots in B and C.

A more detailed image of the proteins is shown in a 1 µm x 1 µm scan shown in panel C.

Domains of lo were deliberately avoided in this scan area. The uniform shape and orientation of the features in this image most likely indicate some form of tip- contamination rather than the structure of the proteins. The different heights of features however, might indicate different proteins on the surface.

During imaging of the DEC-MP lipids, long polymer-like constituents would appear in the images. These long fibers were observed to migrate in location from one scan to the next. This is highlighted in Figure 5-3, which shows three successive images of the same area. There is a dark fiber visible in the bilayer, which in panel A ascends from the lower middle of the frame and runs at a ~30 degree angle to the vertical towards the top left corner. In Panel B, this fiber is seen to have shifted towards the lower left corner, while features associated with the domains remain largely unmoved. By the third successive image (panel C), the fiber has vanished, perhaps having migrated out of the scan area.

- 156 -

Figure 5-3: From left to right, consecutive AFM topography scans over a 10 µm x 10 µm area tracking the migration of a protein fiber in the lower left portion of the image.

Though the interaction of the tip with the bilayer may have an impact on the motion of the fiber, the mobility of the fiber nevertheless indicates that the bilayer most likely remains fluid. Furthermore, after prolonged imaging over the course of 10 hours, the number of specks in the image is reduced, indicating that the smaller proteins are also mobile. The fluidity of the bilayer and of the embedded proteins could potentially impact the presentation of epitopes on the surface. Depending on the analyte of interest, multivalent attachment of the analyte to the surface may have an optimal length-scale or orientation17 that is more easily satisfied by mobile attachments than by covalently tethered ones. In fact, studies suggest that the lateral oraganization of membrane proteins affects cell-signaling18.

The mobility of the proteins in the bilayer is likely responsible for the contamination of the AFM tip that is evident in the force map data. In all, seven force maps were performed on three DEC-MP bilayers on three different mica substrates, using three different AFM tips. In all cases, the force maps, and classification maps exhibit horizontal banding, which indicates that during scanning, conditions remained stable only long enough for several lines of force curves to be collected. For example, the mean breakthrough force recorded may suddenly increase for several lines of scanning, or breakthrough events change to double breakthrough events events – where two breakthrough events occur for a single indentation – or vice versa. The change in

- 157 -

conditions can be triggered by adsorption of lipids or proteins to the tip, which can influence the interaction between the tip and the supported bilayer. For example, accumulation of hydrophobic proteins may result in lower, or absent breakthrough forces, while accumulation of charged species may change the tip’s interaction with the electrical double-layer, and a lipid coating on the tip may result in double breakthrough events.

Figure 5-4 shows a comparison of the force maps collected over pure DEC221 and a DEC-MP bilayer that remained consistent for half the scan. Panels A-C are the image, categorization map, and breakthrough force map for the DEC221 bilayer. The categorization map in panel B indicates breakthrough events were recorded over the ld matrix (grey), while indentations over the lo domains are identified as hard contact (black) due to the lack of a breakthrough event during indentation. This behaviour differs from force maps collected in previous chapters over both gold and mica that showed breakthrough events over both lipid phases. Where breakthrough events did occur, the high salt concentration from the PBS buffer used in the deposition, despite being largely removed by exchange with MilliQ water, likely increases the force required for

19 breakthrough . Some force curves collected over the ld matrix have also been categorized as hard-contact, and appearing as black pixels in panel B, indicating that breakthrough events can also be absent even over the ld matrix.

The breakthrough force map in panel C shows a similar pattern. Only the forces from breakthrough events are shown, while “hard-contact” and “others” are given a value of zero (black). The average breakthrough force value from the force histogram (not shown) is 1.9 nN. This is slightly higher than the average breakthough force for the ld domain on mica seen in previous chapters where the bilayers were deposited in water, with 10 mM calcium, and imaged in pure water, suggesting that the presence of salt can increase breakthrough values. The trigger value for the force map was 8 nN, meaning the indentation was stopped when a force of 8 nN was experienced by the tip, indicating that the breakthrough force of the lo domain is higher than 8 nN. A few individual force

- 158 -

indentations (not shown) over lo domains with trigger values as high as 18 nN also did not exhibit breakthrough.

The lack of breakthrough over lo domains is also observed for the DEC-MP bilayer shown in this figure. Consequently, the black features in panel E correspond reasonably well with lo features in the topography image. Over the ld domains however, the majority of events were recognized as double breakthrough and coloured light grey in panel E. The double breakthrough, in the absence of topographic features, is likely due to contamination of the tip either by proteins or by a bilayer, or both. It has been reported in literature that double breakthrough events can occur when a bilayer forms on a chemically-modified hydrophobic tip in high salt concentrations19. In this case, adsorbed and denatured proteins may make the tip hydrophobic, and the presence of salt stabilizes the bilayer on the tip. This may explain why double-breakthrough events were not detected on the pure DEC221 bilayer, where the tip cannot be contaminate by proteins. A double breakthrough would occur from the rupture of bilayer on the tip, followed by the rupture of the bilayer on the surface, or vice versa.

- 159 -

The breakthrough force map in panel F includes both single and double breakthrough events with only the first breakthrough value shown. The histogram of these forces (not shown) can be fit to a Gaussian with a center-value of 7.4 nN. These high breakthrough values are due to the presence of PBS buffer during force mapping. In another force map where the PBS was exchanged for MilliQ water (not shown) the average breakthrough force value was 2.7 nN, which is closer to the value observed for DEC221 in panel C. Despite exchanging the PBS for MilliQ water, banding still appeared in the force map, which suggests that residual salts are not likely to be responsible for the banding behaviour.

Because the lo domains are predominantly categorized as “hard-contact” events, and the ld domains as double breakthrough events, their contributions to the Young’s modulus can be separated by means of their respective event classification. In Figure 5-5 the modulus values corresponding to double-breakthrough events and hard-contact events are mapped separately in A and B respectively. The patterns seen in the maps bear a resemblance to the domain pattern seen in the topography image. Histograms of the modulus values for double-breakthrough and hard-contact events are shown in panels C and D respectively. The peak values obtained from the log-normal fits to the distributions are 29.4±0.3 MPa and 40.3±0.7 MPa respectively. As expected, the hard-contact events, which are associated with the stiffer lo domains result in modulus values that are higher than the double-breakthrough events that correspond to the softer ld matrix.

- 160 -

Similarly, the data from the pure DEC221 bilayers can be separated by event classification. Figure 5 A and B are the modulus maps of the breakthrough and hard- contact events respectively. Breakthrough events occurred almost exclusively over the ld phase. The majority of hard-contact events occurred over the lo domains, however a fair number of hard-contact events also occurred over the ld matrix. Although this indicates that separation of the two phases by this criterion is incomplete, the classification represents the best available criterion for automated separation. Log-normal fits to the histograms in panels C and D produce mean apparent moduli of 29.0±0.5 MPa and

31.5±0.3 MPa for the ld and lo phases respectively. These values are in reasonable agreement with those obtained from the DEC-MP bilayer, though the mean apparent modulus for the lo phase in this case is lower than that of the DEC-MP, perhaps due to the inclusion of some ld data.

- 161 -

Overall, these values for the moduli of the two phases are lower than those reported in previous chapters. This discrepancy may be a result of the difference in thermal history or the difference in ions that are present during deposition and measurement. Previous evidence suggests that thermal annealing has a dominant effect on the properties of the lo phase, while electrostatic coupling of the ld phase to the surface though divalent calcium ions has a dominant effect on the apparent modulus of the ld phase. In this case, in the absence of Ca2+, the coupling of the bilayer to the surface is low, which may reduce the extent to which the presence of the substrate can affect the apparent modulus, thereby leading to a low apparent modulus for the ld phase. The deposition of the bilayer occurred at room temperature, which is below the glass transition temperature of sphingomyelin, which is the majority constituent in the lo phase. This may explain the relatively low modulus values of the lo phase.

The similarity between the moduli and breakthrough forces of DEC221 and the DEC-MP is surprising considering that in the DEC-MP case, the tip is likely contaminated, as

- 162 -

evidenced by the double breakthrough events and measurement instabilities discussed above. Furthermore, the type and concentration of salts as well as the presence of the extracted proteins and plasma-membrane lipids in the bilayer are also likely to affect the mechanical properties of the bilayer. These data are insufficient to conclusively separate the contributions to the modulus by these factors. While modulus data was collected for DEC-MP after the PBS buffer was exchanged for MilliQ water (not shown), separation of the contributions to the modulus by the two phases was not possible due to the instabilities caused by contamination of the tip. However, without separating the contributions from the two phases, the histogram appears as a single-mode distribution with a log-normal fit peaked at 15 MPa, which is about half the modulus of the ld phase in the DEC221 case where the PBS is exchanged for MilliQ water. This suggests that other factors such as tip contamination, the presence of the proteins in the layer, or the presence of cellular lipids may have as much an effect on the apparent modulus as the buffer, though the contributions from these factors can only be determined by further experimentation with adequate controls, perhaps in the form of chemically modified tips to prevent contamination.

5.4 Conclusions and Future Perspectives

The incorporation of membrane proteins into supported lipid bilayers can potentially broaden the applications of lipid bilayers as functional layers in biosensors both by providing a more biologically relevant membrane mimic and by displaying membrane proteins as surface ligands potentially capable of capturing specific analytes. An example of the latter might be applicable to anti-cancer drug discovery, and may involve the incorpration of EGFR into supported bilayers and the detection of potential pharmaceuticals agents that target EGFR. The data presented here suggests that EGFR containing lipids can be extracted from cultured human alveolar carcinoma cell lines by means of a commercially available kit. The extracted plasma membrane proteins and lipids were mixed with the artificial model-membrane lipid mixture DEC221, and refined into unilamellar vesicles by extrusion.

- 163 -

These protein-containing vesicles were successfully cast onto mica supports and imaged by AFM. When compared to pure DEC221 bilayers, the DEC-MP bilayers had a homogeneously speckled appearance, which were attributed to the proteins. In some cases, protein fibers were observed to migrate across the bilayer from one scan to another, indicating mobility of the proteins within the bilayer. In fact, with prolonged scanning, the number of proteins identifiable in the image is reduced. This indicates not only mobility of these proteins, but also a means by which tip-contamination can occur.

Numerous force maps over multiple DEC-MP bilayers all exhibit horizontal banding in the force or categorization maps indicating instability in the conditions. This instability is most likely triggered by transient tip contamination since adsorption and desorption of proteins and/or lipids to the tip can alter its interaction with the bilayer during the scan. This banding and contamination is rarely seen with pure DEC221 bilayers, suggesting the presence of the membrane components is responsible. Nevertheless, regions of the force maps where the contaminants remain stable for large portions of the scan can be analyzed using the methods discussed in previous chapters. The propensity of double-breakthrough events in the example shown is also indicative of tip contamination. Separating the indentation curves over the lo and ld phases can still be achieved in the example shown, and apparent modulus values of the lo and ld phases were determined to be 40.3±0.7MPa and 29.4±0.3 MPa respectively.

Although these values appear comparable to the values obtained for the pure DEC221 example, which were 31.5±0.3 MPa and 29.0±0.5 MPa for the lo and ld phases respectively, the similarity is likely coincidental considering the multitude of factors that are likely to affect the apparent modulus. For example, in the pure DEC221 case, the PBS buffer used in the incubation step was exchanged for MilliQ water. When the same was done for a DEC-MP bilayer, the average apparent modulus value was 15 MPa, and

- 164 -

mean breakthrough forces were more comparable to the pure DEC221 case, though separation of the lo and ld contributions could not be readily achieved. In addition to the salt conditions, the addition of the proteins and associated plasma membrane lipids are also likely to affect the mechanical properties of the DEC-MP bilayer. The conditions of this study are not suitable for determining the contribution of these factors to the mechanical properties.

Nevertheless, these results demonstrate that fluid protein-containing bilayers can be cast into planar supports, and the presence of the cellular membrane components does not adversely affect the phase segregation of the model lipids. A natural extension of this work would be to cast these bilayers, containing alveolar cell carcinoma surface proteins onto metal surfaces suitable for SPR. The plasmonic properties of such a surface could be used to characterize the expression of EGFR on the artificial surface by detecting the binding of either EGF peptide or anti-EGFR antibodies to the supported bilayer. Ultimately, the method used here can be generalized to produce a variety of supported cell surface mimics by extracting membrane proteins from different cell cultures.

5.5 Acknowledgements:

Natalie Tam from the Gang Zheng Lab at the Toronto University Health Network is acknowledged for culturing the A549 cell line. Funding for this work is provided by the Natural Sciences and Research Council of Canada (NSERC) and the Biopsys NSERC strategic network grant.

5.6 References

1. Johnston, L.J. Nanoscale Imaging of Domains in Supported Lipid Membranes. Langmuir 23, 5886-5895 (2007).

- 165 -

2. Dietrich, C. et al. Lipid Rafts Reconstituted in Model Membranes. Biophys. J. 80, 1417-1428 (2001).

3. Ramsden, J.J. Partition coefficients of drugs in bilayer lipid membranes. Cellular and Molecular Life Sciences 49, 688-692 (1993).

4. Danelian, E. et al. SPR Biosensor Studies of the Direct Interaction between 27 Drugs and a Liposome Surface: Correlation with Fraction Absorbed in Humans. Journal of Medicinal Chemistry 43, 2083-2086 (2000).

5. Cooper, M.A. Optical biosensors in drug discovery. Nat Rev Drug Discov 1, 515-528 (2002).

6. Puu, G. An Approach for Analysis of Protein Toxins Based on Thin Films of Lipid Mixtures in an Optical Biosensor. Analytical Chemistry 73, 72-79 (2001).

7. Kono, K., Ito, Y., Kimura, S. & Imanishi, Y. Platelet adhesion on to polyamide microcapsules coated with lipid bilayer membrane. Biomaterials 10, 455-461 (1989).

8. Andersson, A., Glasmästar, K., Sutherland, D., Lidberg, U. & Kasemo, B. Cell adhesion on supported lipid bilayers. Journal of Biomedical Materials Research 64A, 622-629 (2003).

9. Glasmästar, K., Larsson, C., Höök, F. & Kasemo, B. Protein Adsorption on Supported Phospholipid Bilayers. Journal of Colloid and Interface Science 246, 40- 47 (2002).

10. Malmsten, M. Protein Adsorption at Phospholipid Surfaces. Journal of Colloid and Interface Science 172, 106-115 (1995).

11. Lieber, M., Todaro, G., Smith, B., Szakal, A. & Nelson-Rees, W. A continuous tumor-cell line from a human lung carcinoma with properties of type II alveolar epithelial cells. International Journal of Cancer 17, 62-70 (1976).

12. Smith, B.T. Cell line A549: a model system for the study of alveolar type II cell

- 166 -

function. Am. Rev. Respir. Dis 115, 285-293 (1977).

13. Foster, K.A. Characterization of the A549 Cell Line as a Type II Pulmonary Epithelial Cell Model for Drug Metabolism. Experimental Cell Research 243, 359- 366

14. Geertsma, E.R., Nik Mahmood, N.A.B., Schuurman-Wolters, G.K. & Poolman, B. Membrane reconstitution of ABC transporters and assays of translocator function. Nat Protoc 3, 256-266 (2008).

15. Richter, R.P. & Brisson, A.R. Following the Formation of Supported Lipid Bilayers on Mica: A Study Combining AFM, QCM-D, and Ellipsometry. Biophys. J. 88, 3422-3433 (2005).

16. Cha, T., Guo, A. & Zhu, X. Formation of Supported Phospholipid Bilayers on Molecular Surfaces: Role of Surface Charge Density and Electrostatic Interaction. Biophys. J. 90, 1270-1274 (2006).

17. Yu, J. et al. Polyvalent interactions of HIV-gp120 protein and nanostructures of carbohydrate ligands. NanoBioTechnology 1, 201-210 (2005).

18. Mossman, K.D., Campi, G., Groves, J.T. & Dustin, M.L. Altered TCR Signaling from Geometrically Repatterned Immunological Synapses. Science 310, 1191-1193 (2005).

19. Garcia-Manyes, S., Oncins, G. & Sanz, F. Effect of Ion-Binding and Chemical Phospholipid Structure on the Nanomechanics of Lipid Bilayers Studied by Force Spectroscopy. Biophysical Journal 89, 1812-1826 (2005).

- 167 -

Chapter 6: Summary of Part I and Future Perspectives

6.1 The Importance of Lipid Bilayers in Biosensors

The lipid bilayer is a construct that is ubiquitous in biology, allowing the organization and controlled compartmentalization of molecules, which is essential to fundamental aspects of life from the separation of charges that drive the machinery for ATP synthesis in mitochondria1, and action potentials essential to brain2 and heart3 function, to the presentation of an array of cell surface markers that are involved in immune recognition and complex cell signaling networks4. Because of their importance to many biological processes, membranes and membrane proteins are attractive targets for study in biology, pathology, and pharmacology. The membrane is of particular interest in the latter, since cell membranes play an important role in the uptake of nutrients and small molecules into the body, for example through the gastrointestinal epithelium5,6.

Combining lipid bilayers, usually in the form of supported bilayers with biosensors that transduce molecular binding events enables the measurement of the binding kinetics of proteins, peptides, drugs and small molecules to either the bilayer itself or to transmembrane proteins. The applications for drug testing and discovery are apparent7, as are the applications for studying the molecular mechanisms for such pathogens as Aß8 and the HIV virus9. In fact, several studies have explored these avenues, but is still an active area of research.

The importance of such a sensor for fundamental studies, and to some extent for the applications mentioned above, might depend on how well the supported membranes mimic natural cells membranes. For example, it would be important to assess the degree

- 168 -

to which the binding or interaction of some analytes with the membrane depends on the ability for membrane-bound receptors to mobilize, orient, or coalesce with rafts. Mimicry of these aspects of biological membranes in artificial systems is an ongoing challenge in the field. While some aspects such as mobility and integral membrane protein presentation have been emulated in literature, the inclusion of lipid domains or rafts and assessment of their function in such a biosensor has been rare.

The research presented in this thesis makes several important steps towards the incorporation of phase-segregated lipids and natural plasma membrane proteins into optical biosensing platforms. The characteristics and methods of preparation of these bilayers differ from existing methods in several ways.

6.2 Existing methods of lipid surface functionalization

The deposition of supported lipid bilayers is routinely reported on smooth, charged, hydrophilic substrates such as mica and glass. In many ways mica represents an ideal substrate for supported lipid systems prepared for scanned-probe or other microscopies. It is however not generally suitable for biosensing, especially for SPR or electrochemical methods that require conductive or metallic substrates.

Deposition of lipids onto metal surface however, is not as straight forward. With the exception of freshly cut metal wires, lipid bilayers do not tend to form directly on metal surfaces. This is likely because the surface charge and hydrophilicity can be easily affected by environmental contaminants. Typical contact angles required for the fusion of vesicles on surfaces is less than 10 degrees. Initial experiments with thermal vapour deposited gold substrates suggests a static contact angle of about 30 to 50 degrees is normal. Cleaning with aggressive agents such as piranha solution or UV ozone can bring the static contact angle down to below 15 degrees but the state is short lived and contact

- 169 -

angles rapidly return to the 30 to 50 degree range even if stored carefully in clean containers under MilliQ water.

To circumvent this behaviour of gold, many surface modifications have been employed in literature7,10. Often, hydrophobic self assembled monolayers are formed on gold, upon which can be formed, lipid monolayers, with the tail group of the lipid interacting with the hydrophobic SAM. This method can mimic the outer leaflet of a membrane, and can even host some lipid-anchored species, but cannot host transmembrane proteins that span the bilayer. Several hydrophilic surface modifications have been demonstrated. One such method utilizes a hydrophilic polymer cushion that allows self-assembly of bilayers that host transmembrane protein species. Another method utilizes long polymer tethers that allow the bilayer to form relatively far away from the gold surface. This and the former method both exhibit the advantage of accommodating large intra- or extracellular domains in the proteins they host. Another method involves sputtering of SiOx onto the gold and depositing the lipid onto the SiOx layer. The primary disadvantage of this method, as is often argued, is that there is insufficient space between the lipid and the substrate to accommodate large extra-membraneous domains. The combined Langmuir- Blodgett/-Schaefer method has also been employed. However, a possible advantage for vesicle-fusion is the potential to deliver different vesicles to different regions of the surface via microfluidics, potentially allowing multiple bilayers with different proteins to coexist on a the same surface with potential advantages for multiplexed sensing.

In these endeavours however, there are no reports to the best of the author’s knowledge that demonstrate the phase segregation of lipids on gold. Though in the context of membrane protein incorporation, cushioned or tethered systems are ostensibly the ideal methods to pursue, it is conceivable that both the polymer cushion and the extended water layer would be difficult to probe by AFM force map, which has proven to provide the most reliable contrast between the two phases on gold. Bilayers directly supported by

- 170 -

solid substrates are likely more appropriate for use in scanned probe microscopy, but they limited in their ability to host membrane proteins with large extra-membraneous domains.

The presence of the nanoapertures in the substrate however, may offer a reservoir in which to host such proteins, provided pore-spanning lipid bilayers can be formed. Several reports exist of pore-spanning lipid bilayers on non-plasmonic aperture arrays11,12.

6.3 Summary of results

The results presented in the previous chapters demonstrates the deposition of phase- segregated lipid bilayers on plasmonic nano-hole arrays and the characterization of the bilayer by force indentation, as well as the detection of the bilayer by the refractive-index sensitive plasmonic properties of the array. This was achieved by first demonstrating a method by which to deposit and characterize the lipid layer on gold, then applying this method to a trio of plasmonic nano-aperture arrays, confirming lipid deposition on each.

6.3.1 Phase segregation of lipids on gold

A model ternary lipid system containing DOPC, Egg sphingomyelin, and cholesterol in a 2:2:1 molar ratio was chosen because it is known to phase segregate into sphingomyelin and cholesterol rich liquid ordered domains in a liquid disordered matrix consisting primarily of DOPC. This model membrane system has been widely studied using scanned probe microscopy13‐16, and the mechanical properties of the coexisting phases has been previously reported15.

The general method employed to deposit lipid bilayers on gold involved controlling the surface charge and hydrophilicity via the deposition of a mercaptoundecanoic acid SAM. The carboxylic acid functional group presents a net negative surface charge similar to

- 171 -

mica. The hydrophilicity and surface charge density depended on the grafting density of the SAM. High quality acid-terminated SAMs require acidic conditions for deposition as reported by Wang et al.17.

Vesicle fusion occurred readily on MUDA modified gold, provided the correct buffer conditions were present as outlined by Cha et al.18. Buffers with large molecular anions such as phosphate, were found by Cha et al.18 to be incompatible with positively charged surfaces in terms of zwitterionc lipid deposition. The phosphate anions tended to form a charge balancing layer against cationic surface charges and sterically hinderd the fusion of vesicles. Likewise, in unpublished experiments, it was confirmed that the use of divalent cations such as calcium with negatively charged surfaces effectively reversed the net surface charge, which prevented the formation of a bilayer in PBS buffer. In water however, Ca2+ acted to bridge the surface charges to the lipid headgroups and stabilized the bilayer. The Ca2+ counter balanced the charge on the surface and also induced a dipole in the zwitterionc headgroups of the lipids that attracted them to the surface.

The deposited lipid bilayers were characterized by AFM tapping mode imaging and force mapping. Several optimizations were required for this characterization, including flame annealing of the gold to produce flat grains upon which to image and indent; a magnetic tapping mode accessory for the AFM was used, which enables more reliable phase images than traditional tapping mode in water; AFM force mapping was used to mechanically characterize the bilayer and identify the lipid phases; and an automated force map analysis algorithm was used to sort the force map data and create various maps derived from the force data such as the breakthrough force map and modulus map. The former was used to identify the phases: the lo phase was expected to have a higher breakthrough force than the ld phase due to the tight acyl-chain packing afforded by the fully saturated sphingomyelin. The breakthrough force histogram showed the typical bimodal distribution expected with the lo peak at 3 nN and the ld peak at 2.1 nN

- 172 -

The contrast between the breakthrough forces of the two domains was used to separate the data into two populations corresponding to the lo and ld domains, which were then analyzed separately. The fitted moduli from force curves collected from the lo and ld domains were then plotted in separate maps and histograms. The maps of the two phases clearly resembled the two phases seen in the phase image and the force map. The modal value of each modulus was extracted by fitting histograms to log-normal functions. The moduli were 100 ± 2 MPa and 59.8 ± 0.9 MPa for the lo and ld phases respectively. The presence of the solid support was found to contribute to an over estimation of the modulus by at least a factor of 3.

Comparisons to the same experiment on mica revealed the same trend with breakthrough forces and moduli being greater for the lo phase than the ld phase, and the modulus values were found to be overestimated by a similar factor due to the influence of the mica support. The modulus values of both phases were greater on mica than on gold. The difference in the modulus of the lo phase was large and attributed to the difference in annealing. The difference between the more fluid ld phases was small and could be accounted for by difference in the surface charge densities, and the subsequent effect on the Poisson’s ratio – a parameter whose value was assumed in the fit. The magnitude by which this parameter might be affected by the difference in surface charge density was found to affect the calculated modulus by a magnitude consistent with the observed difference.

Overall, the deposition and phase segregation of lipids was demonstrated. The mechanical properties of the two phases were found to be influenced by surface charges and thermal history.

- 173 -

6.3.2 Model lipids on nano-hole, nano-slit and annular aperture arrays

The methods used to deposit and characterize lipid phases on flat gold were applied to deposit phase segregated lipid bilayers on nano-hole, nano-slit, and annular aperture arrays. The arrays were fabricated on flame-annealed gold to aid the AFM characterization. Multiple arrays were fabricated on each substrate and were simultaneously coated by lipid bilayers in one step. In the data presented, the hole-arrays and slit-arrays were on the same substrate and the annular aperture arrays were on a separate substrate. For all three arrays, the presence of lipid bilayers was confirmed by identifying breakthrough events in force indentation maps. These breakthrough events were observed throughout the force-map area, indicating that the bilayers covered an extensive area of arrays. The patterned apertures, grain boundaries, and lipid defects accounted for the majority of non-breakthrough events. At 5 µm x 5 µm scan sizes, the phase segregation of the lipids was not discernable and this was attributed to the variations in the topography from multiple grains, grain boundaries and milled features that obscured the subtle variations in topography and phase between the lipid phases. The topography also broadened the distribution of recorded force and modulus values, which obscured the differences between the two phases in the force maps.

Using a smaller scan size on the hole array with a freshly redeposited lipid bilayer, an area could be isolated that contained mostly a single grain with only a few grain boundaries and a small collection of milled nano-holes. With fewer topography variations, phase segregation was faintly observable in the topography, and clearly observed in both the modulus and breakthrough force map. The breakthrough force histogram was deemed to exhibit better separation between the two phases, and used to sort the modulus data. The imperfectly separated moduli were 64 MPa and 50 MPa for lo and ld phases respectively. These values appeared much lower than those found in the previous chapter, especially for the lo phase. The imperfect separation of lo from ld force curves means that a portion of ld force curves are grouped with the lo phase, lowering the average modulus. Additionally, the vesicle suspension used in the previous study

- 174 -

contained PBS buffer. Although it was diluted in the deposition process with MilliQ water, enough that the phosphate did not adversely interfere with the bilayer deposition, it is possible that the small quantity of remaining salts increased the apparent modulus greatly, as has been reported elsewhere19. Even within this experiment, the average unseparated modulus values varied from one force map to another and ranged from 20 MPa to 84 MPa. Moduli even differed by more than a factor of 2 between the nano-slit array and nano-hole array maps that were collected over the same bilayer on the same substrate. Clearly, subtle variations in conditions can greatly affect the measured modulus.

Despite variations in the mechanical properties, the presence of the bilayers over the three types of arrays were confirmed by the force-indentation maps. Following the confirmation of the bilayers, transmission spectra were taken through the arrays, and compared to spectra with just the MUDA SAM. In both cases, the arrays were immersed in water for the spectroscopic measurement. Vis and NIR spectra were taken, which cover a wavelength range from 400 nm to 1625 nm (with a gap between 850 nm and 950 nm). The spectra from the slit arrays were dominated by the propagating modes of the array. A possible Bloch mode resonance appeared in the NIR spectrum near 1590 nm, which appeared to red-shift out of the range of the spectrometer after lipid deposition. In the annular aperture array spectra, a discernable Bloch mode resonance was also found at 1590 nm, which likewise red-shifted out of the range of the spectrometer following lipid deposition. The nanohole array spectra were more tractable. Bloch modes from the Au/glass interface and the Au/solution interface were both visible, with the Au/glass modes identified as those bands which did not shift following lipid deposition. The visible Au/solution mode and the NIR Au/solution mode shifted 4 nm and 6 nm respectively following lipid deposition. These results clearly demonstrated the deposition of model lipid bilayers on three plasmonic aperture array structures suitable for biosensing, with phase segregation being observable when fewer topography features were imaged.

- 175 -

6.3.3 Membrane protein incorporation into model lipid bilayers

Plasma membrane proteins were extracted from cultured A549 lung cancer cells, which over expressed a GFP-EGFR fusion protein. The commercial kit used to extract the proteins produced vesicles of plasma membrane lipids containing an assortment of plasma membrane proteins. Membrane extracts were found to be fluorescent, though the fluorescence spectra did not closely resemble GFP. Nevertheless, the membrane protein extracts were reconstituted with DEC221 vesicles by the freeze-thaw method and extruded. The protein-containing DEC221 (DEC-MP) vesicles were deposited onto mica substrates and analyzed by AFM imaging and force mapping.

Imaging revealed that the phase segregation occurred in the DEC-MP bilayer and in a control pure DEC221 bilayer, with both samples having similar domain shapes and sizes. The proteins appeared as a random dispersion of tiny flecks in the image. Zooming into a small group of proteins reveals that they varied in height, and that possible tip contamination occurs rapidly, as indicated by the regularity and orientation of the observed shapes. Some protein fibers were also observed to migrate from one scan to the next. After extended imaging, the protein counts appear to diminish, suggesting they might be contaminating the tip. Alternatively, proteins may have migrated to phase boundaries, or aggregated in regions outside the areas that were scanned.

In the interest of retaining protein structure, the lipids were hydrated, extruded, deposited and imaged in PBS. The presence of the PBS is known to cause unusual tip interactions in force indentations of bilayers, and to cause the contamination of hydrophobic tips with lipids. This kind of behaviour was observed in the force maps, even after PBS was exchanged for water. This is likely because the tip contamination occurred before the buffer exchange, and persisted afterwards. Despite possible tip contamination, average modulus values were within expected ranges. In the presence of proteins and PBS, the

- 176 -

moduli for the ld and lo phases were 29.4 MPa and 40.3 MPa respectively. If tip contamination has occurred, this might represent the mutual indentation of supported bilayer and the tip contaminants.

These results indicate that natural membrane proteins extracted by this commercial kit can be incorporated into artificial lipid systems and cast onto surfaces.

6.4 Future work

There are three type of experiments that arise immediately from this work, and each involves the detection of a soluble analyte. The first, would involve casting a bilayer containing biotinlyated lipids on nanoaperture arrays and detecting the binding of avidin. Since the binding constant of the biotin/avidin pair is known in literature, this experiment may be valuable to calibrate the system response. However, the use of biotin/avidin binding in many sensor applications usually involves immobile assemblies. In the case of a fluid lipid bilayer, calibration may prove to be challenging. The second experiment would involve using the lipid bilayer to directly bind other amphiphiles or hydrophobic species. The third would involve casting membrane proteins extracted from cells onto an aperture array surface and detect some interaction between a membrane protein and, for example, an antibody or peptide.

There are several interesting conceivable variations to the second experiment. In one variation, the lipid bilayer-coated sensor might be capable of directly measuring the partition coefficent of a molecule. The partition coefficient is typically used to determine the relative hydrophobicity of a molecule, and is determined by the relative partitioning of the molecule into an aqueous phase (water or buffer) and into a more oleaginous phase such as octanol at equilibrium. The water/octanol partition coefficient of a drug, peptide, or amino acid is often taken as an approximate measure of the tendency for these

- 177 -

molecules to reside in the lipid portion of the bilayer, but it is well known that using different oleaginous phases results in different partition coefficients. It would be interesting therefore, to directly measure the water/bilayer partition coefficient using model membranes. It may be possible to determine if the partition coefficient between water and a saturated liquid ordered phase would differ from that between water and a liquid disordered phase for any given molecule. Furthermore, the partition coefficient of a molecule between water and a two phase lipid system may not be a linear combination of the two pure phases, since the presence of the lo / ld interface may be favourable for binding of some molecules. Another variation of this experiment might involve detecting the insertion of the Aß peptide into model membrane systems and to compare the pure lo and ld phases to lipids with coexisting phases.

An example of the third type of experiment would be a direct extension of the work of the previous chapters: Casting the membrane proteins extracted from cancer cells onto nanoaperture arrays surfaces. In a proof of principle experiment, it may be interesting to take advantage of the EGFR over expression and detect the binding of either anti-EGFR antibodies or EGF peptides. Such a surface could also be used to test the binding of various drugs or other small molecules or peptide to these membranes, which might be useful for drug discovery, or for discovering other peptide sequences that can be used to target cancer cell surface markers. A possible complication may arise because there is currently no straightforward way to control the orientation of the membrane proteins.

It was alluded to earlier that the ability of untethered or uncushioned bilayers to host transmembrane proteins is reduced by the thinness of the aqueous space between the bilayer and the surface. By combining a lipid bilayer with a nano-hole array, and if the lipids can be made to span the holes, the holes themselves can provide the aqueous space below the bilayer to accommodate large extramembraneous domains. While the lateral mobility of such a protein is likely limited to within the holes, the 200 nm diameter of the holes is still large, even on the scale of multi-subunit membrane proteins20.

- 178 -

It was alluded to in the first chapter that supported lipid bilayers, especially those self- assembled via vesicle fusion are suitable candidates for delivery by microfluidics. Microfluidic delivery of surface functionalizations would allow different targeting species to be delivered and cast onto different regions of the surface for the simultaneous detection of multiple analytes. A possible fourth type of future experiment could therefore involve the delivery of lipid vesicles via microfluidics. Vesicles containing different membrane proteins could be delivered and cast onto different aperture arrays, and their respective targets detected simultaneously. Such an experiment would leverage the advantages of both hole-arrays, and of a lipid-bilayer functionalization platform to realize multiplexed detection.

6.5 Final Remarks

The results presented here represent important steps towards realizing lipid bilayers as functional biointerfaces in optical biosensors, with an emphasis on model membranes, which mimic aspects of raft formation, and their deposition on gold surfaces. Several challenges with regards to the self-assembly of lipid bilayers on gold, and the characterization of lipid phases have been addressed by combining several existing lessons from surface science, to provide an understanding of the processes involved.

Unlike many existing methods for depositing lipids on gold, the method presented does not require specially charged or functionalized lipids, or chemical anchoring of any portion of the bilayer, or thick polymeric surface layers and thus produces a supported lipid bilayer construct that most closely resembles bilayers supported by mica: a construct that has been optimized for study by scanned-probe microscopy.

- 179 -

Without the need for specific headgroup modifications and charges, a lipid layer of nearly any composition can in principle be deposited. This allows for the deposition of model membrane systems that phase segregate into lipid rafts, or potentially, the deposition of natural cell membranes on gold. Additionally, given the wide commercial availability of lipids with various headgroups containing biotin, dyes, and various reactive species or charges, the methods demonstrated in this thesis may constitute a flexible and adaptable platform technology for surface modification of biosensors.

Additionally, these untethered and uncushioned lipid bilayers are optimal for scanned probe microscopy and force indentation mapping, which has proven to be an important tool for the analysis of coexisting lipid phases, especially where topography (for example from the hole arrays) interferes with the traditional identification of lipid phases by height. Furthermore, since the deposition and phase segregation of lipids on hole arrays has been verified by a detailed physical characterization, and correlated to a specific shift in the spectra of the hole arrays, future bilayer depositions can be confirmed by the resulting shift in the spectrum of the arrays without requiring a detailed nanomechanical characterization.

Although the results focus primarily on the deposition of bilayers on aperture-array-type sensors, the techniques employed can be adapted to other metallic sensor surfaces as well. These lessons, for example, can be applied to coating metal nanoparticles with lipid bilayers, as is described in Part II.

For nearly as long as supported lipid bilayers have existed in literature, they have been suggested as biosensor functionalizations. The continued development of biosensor applications of lipid bilayers will provide an additional avenue by which to overcome some of the challenges of biosensing, such as the rapid delivery and self-assembly of multiple surface functionalizations for multiplexed sensing as discussed above.

- 180 -

Additionally, the potential for lipid-based functionalizations to be delivered and self- assembled in microfluidic channels makes them suitable biointerfaces for integrated lab- on-a-chip applications. Since the lipids are physisorbed to the surface, and can be removed by ethanol or isopropanol, they may be suitable for applications where the active portion of the sensor is reused.

These aspects make nanoaperture-based optical transducers ideal partners for lipid functionalized sensors. Like lipid bilayers, nanoaperture arrays are suitable for use in integrated devices due to their small size, linear sampling geometry and modest detection requirements. Additionally the apertures of the arrays can potentially provide a large aqueous reservoir for integral membrane proteins with large intra- or extracellular domains. Furthermore, it is within the holes where electromagnetic field strength is greatest. Thus, the detection of interactions that occur near the holes is potentially favoured. Furthermore, recent developments in the fabrication of free-standing nanoaperture arrays21 may make them suitable for combined surface- plasmon/electrochemical sensing, where the activity of transmembrane transporter proteins can be measured electrochemically11.

The applications for biosensors with lipid interfaces are broad, due to the ubiquity, and importance of lipid membranes in biology. Sensors of this nature have the potential to become important tools in fundamental biology, pathology, and drug discovery.

6.6 References

1. Boyer, P.D. The ATP Synthase - A Splendid Molecular Machine. Annu. Rev. Biochem. 66, 717-749 (1997).

2. Stuart, G., Schiller, J. & Sakmann, B. Action potential initiation and propagation in rat neocortical pyramidal neurons. J. Physiol. (Lond.) 505 ( Pt 3), 617-632 (1997).

- 181 -

3. Luo, C. & Rudy, Y. A model of the ventricular cardiac action potential. Depolarization, repolarization, and their interaction. Circ Res 68, 1501-1526 (1991).

4. Grakoui, A. et al. The Immunological Synapse: A Molecular Machine Controlling T Cell Activation. Science 285, 221-227 (1999).

5. Fromm, M.F. Importance of P-glycoprotein at blood-tissue barriers. Trends Pharmacol. Sci 25, 423-429 (2004).

6. Amidon, G.L., Lennernäs, H., Shah, V.P. & Crison, J.R. A Theoretical Basis for a Biopharmaceutic Drug Classification: The Correlation of in Vitro Drug Product Dissolution and in Vivo Bioavailability. Pharmaceutical Research 12, 413-420 (1995).

7. Cooper, M.A. Optical biosensors in drug discovery. Nat Rev Drug Discov 1, 515-528 (2002).

8. Kremer J.J. & Murphy R.M.[1] Kinetics of adsorption of beta-amyloid peptide Abeta(1-40) to lipid bilayers. Journal of Biochemical and Biophysical Methods 57, 159-169 (2003).

9. Conboy, J.C., McReynolds, K.D., Gervay-Hague, J. & Saavedra, S.S. Quantitative Measurements of Recombinant HIV Surface Glycoprotein 120 Binding to Several Glycosphingolipids Expressed in Planar Supported Lipid Bilayers. Journal of the American Chemical Society 124, 968-977 (2002).

10. Steinem, C., Janshoff, A., Ulrich, W.P., Sieber, M. & Galla, H.J. Impedance analysis of supported lipid bilayer membranes: a scrutiny of different preparation techniques. BBA-Biomembranes 1279, 169-180 (1996).

11. Reimhult, E. & Kumar, K. Membrane biosensor platforms using nano- and microporous supports. Trends Biotechnol 26, 82-89 (2008).

12. Steltenkamp, S. et al. Mechanical Properties of Pore-Spanning Lipid Bilayers Probed by Atomic Force Microscopy. Biophys. J. 91, 217-226 (2006).

- 182 -

13. Johnston, L.J. Nanoscale Imaging of Domains in Supported Lipid Membranes. Langmuir 23, 5886-5895 (2007).

14. Popov, J. et al. Chemical Mapping of Ceramide Distribution in Sphingomyelin-Rich Domains in Monolayers. Langmuir 24, 13502-13508 (2008).

15. Sullan, R.M.A., Li, J.K. & Zou, S. Direct Correlation of Structures and Nanomechanical Properties of Multicomponent Lipid Bilayers. Langmuir 25, 7471- 7477 (2009).

16. Yuan, C., Furlong, J., Burgos, P. & Johnston, L.J. The size of lipid rafts: an atomic force microscopy study of ganglioside GM1 domains in sphingomyelin/DOPC/cholesterol membranes. Biophys. J 82, 2526-2535 (2002).

17. Wang, H., Chen, S., Li, L. & Jiang, S. Improved method for the preparation of carboxylic acid and amine terminated self-assembled monolayers of alkanethiolates. Langmuir 21, 2633-6 (2005).

18. Cha, T., Guo, A. & Zhu, X. Formation of Supported Phospholipid Bilayers on Molecular Surfaces: Role of Surface Charge Density and Electrostatic Interaction. Biophys. J. 90, 1270-1274 (2006).

20. Rubinstein, J.L., Walker, J.E. & Henderson, R. Structure of the mitochondrial ATP synthase by electron cryomicroscopy. EMBO J 22, 6182-6192 (2003).

21. Eftekhari, F. et al. Nanoholes As Nanochannels: Flow-through Plasmonic Sensing. Analytical Chemistry 81, 4308-4311 (2009).

- 183 -

Part II. Lipid Bilayer Functionalization of Plasmonic Nanoparticles

Part II of this thesis is concerned with the encapsulation of metal nanoparticles by lipid bilayer vesicles to produce surface enhanced Raman probes potentially suitable for medical diagnostics. While the methods of functionalizing planar surfaces presented in Part I differ from the methods of functionalizing particle surfaces in Part II, lessons from Part I with respect to the phenomenological behaviour of lipids nevertheless apply. In particular, the interplay between charges on the lipid head-groups and on surfaces is important to consider, as well as the melting temperature of the lipids. This is explored in Chapter 7 – the experimental section of Part II. Chapter 7 details the fabrication and characterization of three SERS probes each utilizing a different dye molecule, and each produced using a unique variation of the same method. The lipid encapsulation acts to protect the particle from aggregation in various salt and acid environments. A novel feature of two of the three methods presented is the ability to simultaneously functionalize the surface with both a protective lipid coating and a dye. Strategies for incorporation of targeting moieties are discussed. Chapter 8 provides a review of literature on the topic of nanoparticle-based probes for biological labelling. Both metallic SERS nanoparticles and fluorescent quantum dot probes are reviewed in terms of surface functionalization strategies, and recently published findings regarding their use as in vitro and in vivo reporters. Also reviewed are recent in vitro and in vivo toxicity studies for both particles. Chapter 9 concludes Part II with a summary of experimental findings, and perspectives on possible future studies.

- 184 -

Chapter 7: Model lipid membrane encapsulation of SERS nanoparticles

7.1 Introduction

The diagnostic and therapeutic applications of metal nanoparticles1,2 requires three basic functional elements in addition to the particle itself: (1) a layer to control surface chemistry for biocompatibility, biodistribution, and/or colloidal stability3 (2) a targeting moiety4-6; and (3) in the case of surface enhanced Raman scattering (SERS) nanoparticles, Raman-active molecules7-10. Advantages of SERS particles over florescent probes include: (i) resistance to photobleaching, (ii) the ability to provide chemical information about the probe or particle environment (since SERS is a vibrational spectroscopy), and (iii) relatively narrow peak widths in spectra, which allow the spectra of many more probes to be distinguished in the same spectral range.

There are many methods of modifying the surface of nanoparticles to include these elements that involve both chemisorbed and physisorbed molecules or polymers and combinations therein. Alternatively, liposomes designed for drug-delivery11,12 often share the first two functional requirements mentioned above, and various methods of engineering their stability and targeting properties have been widely investigated in literature13. In some cases the first requirement is provided by the properties of the vesicle itself14-17, while the second may come from peptides, antibodies, glycans, or folate covalently bound to lipid anchors12,18,19. Many of the strategies to address these functions in the context of liposome-drug-delivery20,21 can potentially be applied to metal nanoparticles for diagnostics, or combined diagnostics/therapeutics, provided that the third functional element, namely a Raman reporter, can be incorporated. The encapsulation of metal nanoparticles in liposomes has been increasingly adopted for the

- 185 -

purposes of colloidal stability, biocompatibility, and the potential of biological targeting22-24.

This chapter reports the encapsulation of metal nanoparticles in model lipid vesicles, which can potentially serve as a platform by which to impart these three functional elements. Two of these elements, namely the stability of lipid-encapsulated particles, and three avenues by which to include Raman-active species are demonstrated. Encapsulation of the particles by a lipid bilayer is confirmed directly by TEM and indirectly by comparing the particles before and after the encapsulation process with respect to their localized surface plasmon resonance (LSPR, via UV-Vis absorption spectra), hydrodynamic radius (via dynamic light scattering), and colloidal stability in acidic and high ionic-strength conditions. SERS spectra of the lipid-encapsulated particles are reported for three dyes, namely malachite green isothocyanate (MGITC), rhodamine lissamine DSPE (Rho-DSPE), and L-tryptophan (Trp), each incorporated by one of the three methods. The SERS spectra are monitored over a time period of several weeks to demonstrate stability of the particles. Proposed, are various means of incorporating targeting moieties. Additionally, the use of ternary lipid mixtures that are known to phase-segregate, strengthen the bilayer25 and mimic some of the functional aspects of lipid rafts.

The encapsulation of nanoparticles by lipids is achieved by sonicating nanoparticles in a suspension of multilamellar vesicles (MLV) for 45-60 min at 50 °C. Sonication of the MLV under these conditions in the absence of particles has been shown to produce unilamellar vesicles (ULV)< 100nm in diameter26,27. This transformation can be observed visually; The MLV suspension appears cloudy, while the ULV suspension appears almost clear, since the size of the vesicles in the ULV suspension has become smaller than the diffraction limit of visible light, and consequently scatters significantly less light than the MLV suspension. Here, it is demonstrated that sonication of the MLV suspension in the

- 186 -

presence of metal nanoparticles results in nanoparticles being encapsulated by a lipid bilayer.

The incorporation of dye molecules was achieved via three methods as illustrated in Figure 7-1A: (Method 1) conjugation of dye molecule (MGITC) to the nanoparticle prior to mixing with MUV suspension; (Method 2)employing a lipid species with a dye covalently-bound to the head-group of a lipid (Rho-DSPE); and (Method 3) addition of an aromatic, hydrophobic amino acid (Trp) to the lipid suspension prior to mixing with the particles. In all cases, the particles were separated from excess lipids by two centrifugation steps, and SERS spectra of the dyes are obtained, confirming the incorporation of the dye species, since unbound dye remains in the supernatant. In the case of the second method, the SERS spectrum also confirms the presence of the lipid layer since the dye is anchored to the bilayer. The chemical structures of the majority components of the bilayer are illustrated in Figure 7-1B. The next section describes the TEM, UV-Vis, DLS, and Raman characterization of the particles produced by the three methods. The stability of the particles is described in the section following that.

- 187 -

Figure 7-1: A) Schematic illustration of three methods of incorporating dye molecules to produce lipid- encapsulated gold SERS particles. The solid circle represents the gold nanoparticle, the rings represent lipid bilayer vesicles, and the hexagons represent small dye molecules. B) Chemical stuctures of the lipid mixture used to encapsulate the nanoparticles.

7.2 Materials and Methods

Dioleoylphosphatidylcholine/Egg sphingomyelin/Ovine Cholesterol (DEC221) aliquots: DEC221 alliquots were prepared as described in Part I. Dioleoylphosphatidylcholine (DOPC), egg-sphingomyelin (ESM), and ovine cholesterol (Chol) (Avanti Polar Lipids, Alabaster, Al, USA) were mixed to a 2:2:1 molar ratio and a final mass of 10.7 mg in a 3:1 chloroform/methanol solution. The solution was divided into 10 aliquots in glass vials and dried under a stream of Argon gas for 1 hr or until the lipids form a film on the bottom of the vials. The vials are then dried under vacuum overnight to remove residual solvent. The dried lipid films are repressurized with Argon gas, capped, and sealed with tape and stored at -20 °C until use.

- 188 -

Particle preparation:

Lipid-coated, dye-labeled particles were prepared using 3 different methods schematically illustrated in Figure 7-1A. The three methods differed by the means with which the dye is added.

Method 1 – Malachite Green: An aliquot of DEC221 was thawed and hydrated in MilliQ water to a final concentration of 1 mg/mL and incubated in a 50 °C circulating water bath for 30 minutes with brief vortexing every 10 minutes to form an multilamellar vesicles (MLV) suspension. Meanwhile, aqueous malachite green isothiocyanate was added to Au colloid and stirred for 10 min to facilitate the adsorption of the positively charged ionic dye to the negatively charged, citrate-capped 60 nm Au (forming MGITC-AuNPs), prior to lipid encapsulation. Equal parts DEC221-MLV and MGITC-AuNPs were mixed, while retaining a quantity of MUV suspension in a separate vial for comparison. The nanoparticle/MUV suspension and the retained MUV suspension were sonicated in a bath sonicator for 45-60min at 50 °C, or until the retained MUV suspension became clear, signifying the formation of small unilamellar vesicles. The vesicles and lipid coated particles were stored at 4°C until use.

Method 2 – Rhodamine-PE: Rhodamine – Lissamine- Phasphatidylethanolamine (Avanti Polar Lipids) in chloroform was divided into aliquots corresponding to 1 mol% of the DEC221 aliquots. One aliquot of the Rhodamine-PE was dissolved in chloroform and combined with one aliquot of DEC221, and dried under a stream of Argon for 1 hour, or until the solvent has visibly dried. The vial was then dried under vacuum overnight, repressurized with Argon, capped, and stored at -20 °C until hydration. The Rhodamin- PE was protected from light by reducing the room lighting as much as practical, wapping the vials in aluminum foil, and keeping the vials covered whenever practical.

- 189 -

Rhodamine-PE/DEC MUVs were made as described in Method 1. The final Rhodamine-

PE concentration was 5 mol% total lipid. In some cases, CaCl2 in water was added to the MUV suspension to a final concentration of 0.2 mM. The MUV was divided into two aliquots and placed in the sonicator at 50°C. An equal volume of the AuNP suspension was added to one vial, and sonication was continued until the lipid vesicle suspension was clear. Both the vesicles and the lipid coated particles are then sealed and stored at 4°C until use.

Method 3 – Tryptophan: The DEC221 MUV solution was prepared as described in method 1, except the lipid aliquot was hydrated in a 1 mM L-tryptophan solution. Equal parts Trp-MUV and AuNP were combined in a vial, and a portion of the MUV was retained in a separate vial. As before, both were sonicated at 50 °C for 45-60 min or until the pure lipid suspension became clear. Vesicles and lipid-coated particles were sealed and stored at 4 °C until use.

7.3 Characterization of Lipid-encapsulated SERS particles

Method 1: The lipid encapsulation serves to protect the dye/particle conjugate from flocculation or dissociation. The lipid layer was observed directly by TEM, as shown in Figure 7-2B, where it appeared as a diffuse ring around the dark nanoparticle, with an average thickness of 4.8 nm, consistent with expected size increase for the formation of a bilayer14. The hydrodynamic radius of the lipid coated particles, compared to the stock citrate-coated particles measured by DLS (Figure 7-2C) were consistent: The centroids of Gaussian fits to the histograms suggest that the hydrodynamic radius of the particles have increased from 30.5 nm to 36.4 nm due to the presence of the lipid bilayer. The presence of the lipid bilayer was further confirmed by the UV-Vis absorption spectrum of the particle whose main localized surface plasmon resonance (LSPR) absorption red-shifted from 534 nm to 538 nm when compared to the lipid-free particles (Figure 7-2D) due to

- 190 -

the addition of the dielectric lipid layer. This shift was significant compared to the 2nm shift widely reported for the addition of bulky polyethylene glycol to the particle surface7. Figure 7-2E is the SERS spectrum of the lipid-encapsulated particles, which can be identified as the SERS spectrum of MGITC reported elsewhere7,28,29, indicating that the MGITC remains associated with the metal particle following the sonication with lipids. The peaks in the SERS spectra were assigned to specific vibrational modes of MGITC, by comparison with assignments made in literature28. These assignments are summarized in Table 7-1.

Band (cm-1) Chemical Group Mode

447* Benzene ring deformation (out of plane)

530 Benzene ring deformation (in plane)

791* C-H (Benzene) wagging

1173 C-H rocking

1289* C-C; C-C-H rocking; rocking

1364 -N stretch

1584* ring stretch

1620 -N; C-C stretch; stretch

- 191 -

Figure 7-2: Characterization of Malachite Green isothiocyanate (MGITC) lipid-encapsulated SERS nanoparticles produced using Method 1. A) The chemical structure of MGITC, B) TEM image of a bilayer- encapsulated nanoparticle. The ring around the particle is the lipid bilayer, which measures 4.8 nm averaged from 5 measurements of 5 isolated particles. C) DLS histograms of hydrodynamic radius for stock citrate- coated particles (hatched,grey) and MGITC-lipid-coated SERS nanoparticles (outline,black). Gaussian centroids report average Rh = 30.5 nm and 36.4 nm for stock particles and SERS particles respectively. D) LSPR absorption in UV-Vis spectrum of stock (grey) and MGITC-lipid-coated particles (black). LSPR absorption peaks at 534 nm for stock particles and 538 nm for MGITC-lipid-coated particles. E) SERS spectrum of MGITC-lipid-coated nanoparticles showing strong SERS spectrum recognizable as that of MGITC.

Method 2: A small amount (5 mol% total lipid) of an appropriate dye-labeled phospholipid (Rhodamine-PE), was incorporated into the DEC221 vesicle, thus, surface

- 192 -

encapsulation and dye conjugation of the particle was achieved simultaneously. By extension, lipid species with targeting moieties may in principle be incorporated simultaneously in much the same way. Here, a phospholipid with a Rhodamine- lissamine modified headgroup (structure: Figure 7-3A) was chosen because it is commercially available and because Rhodamine dyes have been shown to have strong SERS cross-sections30-32. Note that the Rhodamine labeled lipids have a net-negative charge. The use of dye-labeled lipids was advantageous because the dye was covalently bound to the lipid, and thus the acquisition of a SERS spectrum of the dye following sonication and centrifugation of the particles served as an additional means of confirming the presence of the bilayer next to the particles. This is because dissolved lipids and even vesicles can only be settled from suspension under very high centrifugation speeds compared to the dense gold nanoparticles33. Separating the particles from the supernatant leaves left unbound lipids among the particle fraction. Furthermore, only those lipids bound to the particle were close enough to the particle’s surface to contribute to a detectable SERS signal.

Nevertheless, TEM (Figure 7-3B), DLS (Figure 7-3C) and UV-Vis(Figure 7-3D) were used to characterize the particles produced by Method 2. TEM image analysis revealed that the average bilayer thickness was 5.8 nm. Gaussian fits to the DLS histograms reported an average hydrodynamic radius of 36.8 nm, compared to 30.5 nm for the stock citrate-coated particles. Both the TEM image analysis and the DLS data suggested a larger bilayer than the particles produced by Method 1. This may have been due to the bulkier headgroup of the Rhodamine-labeled lipid. On the other hand, the shift in the LSPR band in the UV-Vis spectrum revealed a red-shift from 534 nm to 536 nm that was more modest than that found for Method 1 products, but similar in magnitude to the shift produced from the 5 KDa PEG functionalization of these particles7.

- 193 -

Figure 7-3: Characterization of lipid-encapulated SERS nanoparticles with 1 mol% Rhodamine-Lissamine- DSPE (Rho-PE) produced using Method 2. A) The chemical structure of Rho-PE, B) TEM image of a bilayer-encapsulated nanoparticles. The ring around the particles is the lipid bilayer. Clustering is caused by drying of the sample on the TEM grid. The bilayer measures 5.8 nm averaged from 10 measurements of 5 clusters of particles. C) DLS histograms of hydrodynamic radius (bars) and fits to Gaussian distributions (curves) for stock citrate-coated particles (grey) and Rho-lipid-coated SERS nanoparticles (black). Gaussian centroids report average Rh = 30.5 nm and 36.8 nm for stock particles and SERS particles respectively. D) LSPR absorption in UV-Vis spectrum of stock (grey) and Rho-lipid-coated SERS particles (black). LSPR absorption peaks at 534 nm for stock particles and 536 nm for Rho-lipid-coated particles. E) Fluorescence spectra of Rho-lipid vesicle supernatant (dashed) and of Rho-lipid-coated SERS particles (solid). F) UV-Vis absorption of Rho-lipid-coated SERS particles prepared with Ca2+. Grey solid line is the abs. spectrum of stock Au particles, black line is the abs. spectrum of Rho-lipid-coated particles, and grey dashed line is the abs. spectrum of Rho-PE/DEC vesicles. G) SERS spectrum of Rho-lipid-coated nanoparticles prepared with Ca2+ showing strong SERS spectrum recognizable as that of Rhodamine.

- 194 -

In addition to these data, fluorescence data of these particles and of the lipids were analyzed, since rhodamine exhibits a strong fluorescence. The normalized fluorescence excitation and emission spectra of the first supernatant and the isolated particles were overlayed in Figure 7-3E. The first supernatant exhibited broad excitation and emission bands with peaks at 548 nm and 596 nm respectively, and was nearly identical to the fluorescence spectra of the rhodamine-lipid samples prior to mixing with gold. This suggests very few particles remained in the supernatant, as expected. The raw fluorescence spectra of the particle fraction after two centrifugation steps were weak, and generally not detectable on the same scale as the supernatant (not shown). Thus, the normalized spectra are shown to emphasize the features. At first, it appeared that the Stokes shift had narrowed via a red-shift of the excitation peak and a blue-shift of the emission peak. This was attributed to the effect of quenching of fluorescence due to the overlap between the dye’s excitation spectrum and the LSPR resonance of the particle. The overall weakness of the raw fluorescence spectra suggested significant quenching by the particle was occurring. The strongest LSPR absorption occured at 536 nm (UV-Vis data), which coincided with the rhodamine excitation band (Supernatant and vesicle fluorescence data). The apparent peak near 576 nm for the rhodamine-lipid-particles’ excitation spectrum results from the weak LSPR absorption at longer wavelengths. Particles prepared in the same fashion with NBD-labeled lipids did not show the same quenching behaviour, since the dye excitation did not overlap with the particle LSPR.

At first, the SERS spectra of the rhodamine-labeled lipid-coated particles was weak (not shown). It is conceivable that the negative charges on both the citrate layer on the particle surface, and on the rhodamine-lissamine labeled lipid headgroup had caused the rhodamine-labeled lipid to preferentially sit in the outer leaflet of the bilayer, away from the particle surface. In this orientation, SERS excitation is expected to be less efficient. This postulate was supported by experiments on planar supported lipid bilayers that found that electrostatic interactions between a charged lipid headgroups and charged surfaces (or counter ions) resulted in asymmetric segregation of charged lipid species between the two leaflets of the bilayer34-37. It was therefore hypothesized that the

- 195 -

addition of divalent calcium may act to bridge these charges and allow the rhodamine- conjugated lipid to reside in the inner leaflet.

Indeed, strong Raman bands were observed when calcium was employed during sonication, however, the UV-Vis absorption spectrum (black line in Figure 7-3F) showed two distinct LSPR resonances. There are two possible interpretations.

With the first interpretation the second peak at 575 nm was attributed to particle aggregates while the peak at 536 nm was attributed to the LSPR absorption of individual lipid-coated particles, which is shifted relative to the absorption spectrum of stock citrate- coated particles (grey solid line Figure 7-3F); relatively narrow width of the second (575 nm) band suggests uniformity in number and geometry of aggregation. The mechanism by which particle aggregation was controlled is not known. DLS data from the Rho- PE/DEC-particles encapsulated in the presence of Ca2+ was inconclusive (not shown), since the software was unable to fit the data to a hydrodynamic radius.

Alternatively, with the second interpretation, the UV-Vis absorption at 575 nm was attributed to the absorption of light by the Rhodamine moiety, while the intepretaion of the first peak at 536 remained unchanged. The fluorescence excitation peak of the Rhodamine-DEC vesicles was found to be 576 nm from the fluorescence data (Figure 7-3E). To test this possibility, the UV-Vis absorption spectrum of the Rhodamine-DEC vesicles was measured. The spectrum of the Rhodamine-DEC vesicles (grey, dashed line in Figure 7-3F) was found to be superimposable over the second peak in the spectrum of the Rho-PE/DEC-coated particles. This interpretation of the origin of the second peak appears more plausible than the controlled-aggregation interpretation in lieu of a mechanism by which aggregation could be controlled.

- 196 -

The Raman spectrum of the Rhodamine particles produced with calcium is shown in Figure 7-3G. The Raman spectrum of the Rhodamine-containing vesicles alone was subtracted from this spectrum to remove the fluorescence background, although several Raman bands were strong enough to be discernable over the fluorescence. The observed peaks are consistent with the SERS spectra of Rhodamine variants incorporated by other means29,38-40. Assignment of the peaks to specific vibrational modes was accomplished by comparison with assignments made in literature for the SERS of Rhodamine B38 and the normal Raman scattering of Rhodamine 6G39, as shown in Table 7-2.

Band (cm-1) Chemical Group Mode

1207† C-H (ring) bending

1283* Xanthene ring breathing

1358 Xanthene ring stretching

1438 Xanthene ring stretching

1517 Xanthene; C-N ring stretching; stretch

1647 Xanthene; C-H ring stretching; rocking

Method 3:Soluble dyes were also co-encapsulated with the particles during sonication. In this case, the amino acid L-tryptophan (Structure: Figure 7-4A) was dissolved along with the lipids prior to forming the MUV solution. Upon sonication with the particles, Trp likely co-encapsulated with the particles. Although, due to the partition coefficient of

- 197 -

Trp41, the possibility of Trp partitioning into the lipid layer cannot be ruled out. While tryptophan has been found to localize at the lipid-water interface when incorporated into membrane proteins42, the exact nature of the Trp in this construct has not been studied. Nevertheless, the nanoparticles prepared in this fashion exhibited a 3.6 nm shift in the UV-Vis/LSPR spectrum, which was comparable to particles made by the other two methods. The Raman spectrum of the Trp particles is shown in Figure 7-4C. The spectrum of the lipid-coated particle is consistent with the Raman spectrum of Trp reported in literature43,44. By comparison to SERS spectra in ref 43, many peaks can be attributed to specific vibrational modes, which are summarized in Table 7-3.

Figure 7-4: Tryptophan as a Raman active molecule in a lipid-coated SERS nanoparticle. A) Chemical structure of Tryptophan. B) UV-Vis absorption spectrum of bare gold (grey) and Trp-Lipid-coated SERS particle (black). Peak shifts form 535nm to 538.6nm. C) SERS spectrum of Trp-lipid-coated SERS nanoparticle. Significant bands are marked by a vertical line and the wave number.

It is worth noting that a solution-phase Raman spectrum of Tryptophan alone, at the same concentration at which it was added to the lipid suspension was not detectable using the same laser power and integration time. This suggests that although the tryptophan SERS signal was weaker than that of the MGITC-lipid-particle sample, the observed spectrum nevertheless indicates a significant enhancement of the Raman signal.

- 198 -

Table 7-3: Assignment of observed SERS bands based on ref 43. “Band” column is the observed wave- number of a peak. (*) indicates that correlation of band position with ref 43 is > 5 cm-1, and that the assignment is the closest match. “Chemical group” and “Mode” columns indicate the chemical group and type of vibrational mode assigned by ref 43. Multiple assignments to a band are separated by semicolons. Parentheses indicate location of the chemical group within the molecule.

Band (cm-1) Chemical Group Mode

- + 926 C-COOH ; NH3 ; CH2 stretching; rocking; rocking

999 Indole ring breathing

1071 NH3+; (C-)H rocking; bending

1151 H (benzene) scisoring

1218 pyrole, C-COOH- stretching; stretching

1266* H(-indole); CH rocking; bending

H(-methyl) bending

1357* CH; H(-methyl) bending, bending

1375* CH2; CH wagging, bending

1392 COO- stretching (symmetric)

1475* CH2 scissoring

1550 indole stretching

1581 NH3+ scissoring

1605 indole stretching

1616* COO- stretching (antisymmetric)

7.4 Stability of lipid encapsulated SERS particles

The SERS spectra collected over time was used to assess the stability of the particles, since precipitation of the particles or dissociation of the dye from the particles would result in a reduction in the Raman intensity. The Raman spectra of the MGITC-lipid- coated-particles were recorded after 12 and 25 days of storage at 4 0C, and show no

- 199 -

significant signs of signal change, indicating the stability of the particles over time. These spectra, offset for clarity, are shown in Figure 7-5 A.

Surprisingly, even the spectra for Rhodamine-lipid-coated-particles prepared in the presence of Ca2+ exhibited stable SERS spectra over the course of 7 days. Stability over longer times have not yet been tested. The spectra of Rho-lipid-particles at 1 and 7 days are shown in Figure 7-5B, with the spectra offset for clarity.

- 200 -

The colloidal stability was also tested by subjecting both the stock citrate-coated nanoparticles and the lipid-encapsulated particles to several ionic and pH conditions, using the same particle concentrations. An additional sample, containing just the ULV suspension was also subjected to the same conditions as a control. For these tests, a fluorescent Nitrobenzoxadiazol (NBD)-tail-group-labeled phosphocholine lipid was incorporated into the lipid mixture at 0.2 mol% because this species imparts a bright green fluorescence that helps visualize the ULV suspension. It was also included in the particle-lipid suspension at the same concentration for consistency. Because the dye is conjugated to the tail-group, the chemical properties of the head-goups remain the same as the other components of DEC221. Table 1 summarizes of test conditions and observations. For each test, small aliquots of each suspension were tested independently, with the final concentration of the lipids in the ULV suspension, and the nanoparticles in the “stock” solution being nominally the same as those in the lipid-encapsulated particle suspension.

The suspensions were subjected to 5% acetic acid, 10mM CaCl2, 10mM NaCl, and PBS (50 mM monvalent salts), to test the resistance to acidic pH, divalent cations, and various monovalent salt concentrations.

In the first test (condition b), the pH of the environment for each of the three suspensions was lowered by adding acetic acid to a final concentration of 5% v/v. At neutral pH, the stock particles are protected from aggregation by mutual electrostatic repulsion provided by the anionic citrate coating. The addition of acid to the environment is expected to cause the protonation of citrate, and subsequently the flocculation of the nanoparticles. Indeed this was observed; the colour of the suspension of citrate-coated particles changed from pink to colourless, indicating that the nanoparticles had flocculated. Conversely, the lipid-coated particles were not visibly affected by the acid, suggesting that the particles

- 201 -

were protected from flocculation by means other than protonatable electrostatic repulsion, such as steric repulsion of the lipids, or mild induced dipole interactions between zwitterionic headgroups.

Baseline descriptions of the stock particles, lipid encapsulated particles, and ULV suspension were taken in ultra-pure water. See Table 1, treatment (a).

Lipid-Encapsulated Control 1: Particle Control 2: Lipid only Particles Only Treatment a) Water Clear, pink solution Clear, pink solution Greenish-yellow (NBD), faintly cloudy Treatment b) Acetic Clear, pink solution Clear, Colourless Less colourful Acid Treatment c) CaCl2 Paler pink Clear, Colourless Greenish-yellow, faintly cloudy Treatment d) NaCl Clear, pink solution Clear, pink solution Greenish-yellow, faintly cloudy Treatment e) PBS Clear, pink solution Clear, pink solution Greenish-yellow, faintly cloudy

In principle, increasing the ionic strength of the environment of the particles can shield the electrostatic interactions that stabilize particle suspensions. Specifically, divalent cations are known to bridge two anionic charges. Ca2+ ions, even in modest concentration, are expected to cause aggregation of the citrate-coated particles by this bridging mechanism. Similarly, Ca2+ is also known to electrostatically bridge lipid headgroups, which has been exploited to induce phase segregation45 and also fusion of vesicles on anionic surfaces46. Ca2+ was therefore expected to affect the stability of lipid- coated particles as well. The findings for condition-c were consistent with these expectations. Ca2+ caused the citrate-coated particles to aggregate as evidenced by the loss of colour of the suspension. In the case of the lipid-coated particles, the effect was less dramatic. The pink colour became less intense, but unlike the citrate-coated

- 202 -

particles, the colour of the lipid-coated particles was not lost completely. The reduced aggregation effect might have been due to weaker interaction between the lipid headgroups and calcium, than between the negatively charged citrate and positively charged calcium, which is expected because the charges on the zwitterionic lipids are only induced transiently. Predictably, the observabed cloudiness of the ULV suspension increased.

Similar concentrations of Na+ ions had little effect on the stock nanoparticles. Moderate concentrations of Na+ would not be expected to affect either the ULVs or the lipid-coated particles. Likewise, the addition of PBS, a higher concentration of salt with additional large molecular counter ions (phosphate) did not have an observable affect on any of the test or control suspensions.

7.5 Conclusions

Lipid bilayers represent a modular surface functionalization route for SERS nanoparticles, provided dye molecules can be incorporated. The encapsulation of commercially available, citrate-coated gold nanoparticles by model lipid bilayers was demonstrated, along with three methods of incorporating dyes, each with a different dye species, namely malachite green, rhodamine-lissamine, and tryptophan. SERS spectra of the three dye-lipid-particle constructs were observed, and lipid bilayer encapsulation was demonstrated by TEM imaging, dynamic light scattering, and the endogenous UV- Vis/LSPR spectroscopy of the particles.

The first method of dye incorporation involved associating MGITC with the nanoparticles before biayer encapsulation. The second method used a small fraction of Rhodamine-labeled lipids to combine encapsulation and dye-association in one step. The

- 203 -

third method involved combining tryptophan, which is known to partition into lipid bilayers, with the lipid suspension before encapsulating particles.

Lipid bilayers surrounding the particles were directly observed by TEM. Images of MGITC-lipid-coated-particles and Rhodamine-lipid-coated particles show a discernable ring around the particles with average thicknesses of 4.8nm and 5.8nm respectively, indicating the likey formation of a bilayer. The polar nature of the particles would, in principle, favour bilayer formation over monolayer formation.

UV-Vis absorption data for particles produced by the three methods showed a red-shift in the range of 3-4 nm, indicating a relatively significant deposition of material, while indicating that the majority of particles remained singular following the lipid treatment, with the notable exception of particles prepared in the presence of calcium ions. The UV- Vis absorption spectrum for the calcium containing Rhodamine-lipid-particles exhibited two peaks, the first was the familiar LSPR peak of individual particles, while the second peak at 575 nm was interpreted as either arising from controlled aggregation of particles, or from the absorption by the Rhodamine dye. The DLS data showed increases in hydrodynamic radii of the lipid coated particles in the range of 1-2 nm for all dye-lipid- coated particles, with the exception of those prepared in the presence of calcium.

SERS spectra were obtained for encapsulated particles prepared by methods 1 and 3, confirming the successful incorporation of the dye. SERS for the Rhodamine-lipid- particles was weak in comparison. It was postulated that negatively charged Rhodamine- lipids were preferentially segregated to the outer leaflet by repulsive interaction with the negatively charged citrate coating on the particle. It was hypothesized that calcium could act to bridge the charges between the citrate and the rhodamine-PE headgroup. While great improvement to the SERS intensity was observed for these particles, one interpretaion of the UV-Vis data suggested that aggregates had formed, as evidenced by

- 204 -

the appearance of two distinct LSPR bands. The longer wavelength band can be attributed to the aggregates. However, the relatively narrow width of the LSPR band associated with the aggregate suggests a single population of aggregates, indicating a fairly controlled aggregation. A mechanism by which aggregation number was controlled has not yet been identified.

The stability of the lipid-coated particles over time was demonstrated by the unchanged SERS spectra of the MGITC particles over the course of 25 days. Additionally, the stability of lipid-encapsulated particles to environmental factors such as high divalent salt concentration and acidic conditions was found to be greater than citrate-coated particles. Furthermore, a test against flocculation in acidic or high Ca2+ conditions provides a rapid and simple means of confirming lipid encapsulation. Surprisingly, Rho-lipid-coated particle aggregates prepared by Method 2 in the presence of Ca2+ also showed stable spectra for at least 7 days. Spectra after longer periods have as yet been unmeasured.

It is concluded that lipid encapsulation of gold nanoparticles constitutes a flexible platform by which to control the surface properties of metal nanoparticles for diagnostic and/or therapeutic applications. The use of a ternary mixture offers the possibility of enhanced mechanical stability; control of the membrane modulus also controls the driving force for NP-lipid mixing. Furthermore, Rafts offer a route to concentrate antibodies and an opportunity for creating anisotropic binding. A large body of knowledge exists in literature for the protection and targeting of vesicles and liposomes in the context of drug delivery. Combining existing targeting strategies with these encapsulation and dye-association techniques constitutes a new system for the development of SERS nanoparticles for medical diagnostics and therapeutics.

- 205 -

7.6 Acknowledgements

The author acknowledges the following people: Christina MacLaughlin for her contributions to the preparation and characterization of particles in this work; Dr. Nikhil Gunari for the preparation of TEM samples and Imaging; Iliya Gourevich and Niel Coombs of the Center for Nanoscale Characterization for advice on TEM sample preparation, and for TEM imaging.

7.7 References

1. Loo, C., Lowery, A., Halas, N., West, J. & Drezek, R. Immunotargeted Nanoshells for Integrated Cancer Imaging and Therapy. Nano Lett. 5, 709-711 (2005).

2. Huang, X., El-Sayed, I.H., Qian, W. & El-Sayed, M.A. Cancer Cell Imaging and Photothermal Therapy in the Near-Infrared Region by Using Gold Nanorods. J. Am. Chem. Soc. 128, 2115-2120 (2006).

3. Giljohann, D.A. et al. Gold Nanoparticles for Biology and Medicine. Angew. Chem. Int. Ed. NA-NA (2010).doi:10.1002/anie.200904359

4. Zhang, C., Yeh, H., Kuroki, M.T. & Wang, T. Single-quantum-dot-based DNA nanosensor. Nat Mater 4, 826-831 (2005).

5. Oyelere, A.K., Chen, P.C., Huang, X., El-Sayed, I.H. & El-Sayed, M.A. Peptide- Conjugated Gold Nanorods for Nuclear Targeting. Bioconjugate Chem. 18, 1490- 1497 (2007).

6. Kumar, S., Harrison, N., Richards-Kortum, R. & Sokolov, K. Plasmonic Nanosensors for Imaging Intracellular Biomarkers in Live Cells. Nano Lett. 7, 1338-1343 (2007).

7. Qian, X. et al. In vivo tumor targeting and spectroscopic detection with surface- enhanced Raman nanoparticle tags. Nat Biotech 26, 83-90 (2008).

8. Sha, M.Y., Xu, H., Natan, M.J. & Cromer, R. Surface-Enhanced Raman Scattering

- 206 -

Tags for Rapid and Homogeneous Detection of Circulating Tumor Cells in the Presence of Human Whole Blood. J. Am. Chem. Soc 130, 17214-17215 (2008).

9. Kneipp, J., Kneipp, H., Rajadurai, A., Redmond, R.W. & Kneipp, K. Optical probing and imaging of live cells using SERS labels. J. Raman Spectrosc. 40, 1-5 (2009).

10. Nguyen, C.T. et al. Detection of chronic lymphocytic leukemia cell surface markers using surface enhanced Raman scattering gold nanoparticles. Cancer Lett. 292, 91-97 (2010).

11. Hofheinz, R., Gnad-Vogt, S.U., Beyer, U. & Hochhaus, A. Liposomal encapsulated anti-cancer drugs. Anti-Cancer Drugs 16, 691-707 (2005).

12. Wu, J., Liu, Q. & Lee, R.J. A folate receptor-targeted liposomal formulation for paclitaxel. Int. J. Pharm. 316, 148-153 (2006).

13. Torchilin, V.P. Recent advances with liposomes as pharmaceutical carriers. Nat Rev Drug Discov 4, 145-160 (2005).

14. He, P. & Urban, M.W. Phospholipid-Stabilized Au−Nanoparticles. Biomacromolecules 6, 1224-1225 (2005).

15. Thaxton, C.S., Daniel, W.L., Giljohann, D.A., Thomas, A.D. & Mirkin, C.A. Templated Spherical High Density Lipoprotein Nanoparticles. J. Am. Chem. Soc. 131, 1384-1385 (2009).

16. Cormode, D.P. et al. Nanocrystal Core High-Density Lipoproteins: A Multimodality Contrast Agent Platform. Nano Lett. 8, 3715-3723 (2008).

17. Bakshi, M.S. et al. Surfactant Selective Synthesis of Gold Nanowires by Using a DPPC−Surfactant Mixture as a Capping Agent at Ambient Conditions. J. Phys. Chem. C 111, 5932-5940 (2007).

18. Chen, W.C. et al. In vivo targeting of B-cell lymphoma with glycan ligands of CD22. Blood 115, 4778-4786 (2010).

- 207 -

19. Park, J.W. Tumor targeting using anti-her2 immunoliposomes. J. Controlled Release 74, 95-113

20. Allen, T.M. & Moase, E.H. Therapeutic opportunities for targeted liposomal drug delivery. Adv. Drug Delivery Rev. 21, 117-133 (1996).

21. Park, J.W. et al. Anti-HER2 Immunoliposomes. Clin. Cancer Res. 8, 1172-1181 (2002).

22. Thaxton, C.S., Daniel, W.L., Giljohann, D.A., Thomas, A.D. & Mirkin, C.A. Templated Spherical High Density Lipoprotein Nanoparticles. J. Am. Chem. Soc. 131, 1384-1385 (2009).

23. Cormode, D.P. et al. Nanocrystal Core High-Density Lipoproteins: A Multimodality Contrast Agent Platform. Nano Lett. 8, 3715-3723 (2008).

24. Bakshi, M.S., Possmayer, F. & Petersen, N.O. Role of Different Phospholipids in the Synthesis of Pearl-Necklace-Type Gold−Silver Bimetallic Nanoparticles as Bioconjugate Materials. J. Phys. Chem. C 111, 14113-14124 (2007).

25. Sullan, R.M.A., Li, J.K. & Zou, S. Direct Correlation of Structures and Nanomechanical Properties of Multicomponent Lipid Bilayers. Langmuir 25, 7471- 7477 (2009).

26. Lapinski, M.M., Castro-Forero, A., Greiner, A.J., Ofoli, R.Y. & Blanchard, G.J. Comparison of Liposomes Formed by Sonication and Extrusion: Rotational and Translational Diffusion of an Embedded Chromophore. Langmuir 23, 11677-11683 (2007).

27. Maulucci, G. et al. Particle Size Distribution in DMPC Vesicles Solutions Undergoing Different Sonication Times. Biophys. J. 88, 3545-3550 (2005).

28. Lueck, H.B., Daniel, D.C. & McHale, J.L. Resonance Raman study of solvent effects on a series of triarylmethane dyes. J. Raman Spectrosc. 24, 363-370

- 208 -

29. Rule, K.L. & Vikesland, P.J. Surface-Enhanced Resonance Raman Spectroscopy for the Rapid Detection of Cryptosporidium parvum and Giardia lamblia. Environmental Science & Technology 43, 1147-1152 (2009).

30. Pettinger, B. & Krischer, K. Comparison of cross-sections for absorption and surface- enhanced resonance Raman scattering for rhodamine 6G at coagulated silver colloids. J. Electron Spectrosc. Relat. Phenom. 45, 133-142 (1987).

31. Michaels, A.M., Nirmal, M. & Brus, L.E. Surface Enhanced Raman Spectroscopy of Individual Rhodamine 6G Molecules on Large Ag Nanocrystals. J. Am. Chem. Soc. 121, 9932-9939 (1999).

32. Pieczonka, N.P.W. & Aroca, R.F. Single molecule analysis by surfaced-enhanced Raman scattering. Chem. Soc. Rev. 37, 946 (2008).

33. Zeidel, M.L., Ambudkar, S.V., Smith, B.L. & Agre, P. Reconstitution of functional water channels in liposomes containing purified red cell CHIP28 protein. Biochemistry 31, 7436-7440 (1992).

34. Richter, R.P., Maury, N. & Brisson, A.R. On the Effect of the Solid Support on the Interleaflet Distribution of Lipids in Supported Lipid Bilayers. Langmuir 21, 299-304 (2005).

35. Rossetti, F.F., Textor, M. & Reviakine, I. Asymmetric Distribution of Phosphatidyl Serine in Supported Phospholipid Bilayers on Titanium Dioxide. Langmuir 22, 3467- 3473 (2006).

36. Richter, R.P. & Brisson, A.R. Following the Formation of Supported Lipid Bilayers on Mica: A Study Combining AFM, QCM-D, and Ellipsometry. Biophys. J. 88, 3422-3433 (2005).

37. Käsbauer, M., Junglas, M. & Bayerl, T. Effect of Cationic Lipids in the Formation of Asymmetries in Supported Bilayers. Biophys. J. 76, 2600-2605 (1999).

38. Zhang, J., Li, X., Sun, X. & Li, Y. Surface Enhanced Raman Scattering Effects of

- 209 -

Silver Colloids with Different Shapes. J. Phys. Chem. B 109, 12544-12548 (2005).

39. Jensen, L. & Schatz, G.C. Resonance Raman Scattering of Rhodamine 6G as Calculated Using Time-Dependent Density Functional Theory. J. Phys. Chem. A 110, 5973-5977 (2006).

40. Chen, J. et al. Gold-aggregated, dye-embedded, polymer-protected nanoparticles (GDPNs): A new type of tags for detection with SERS. Colloids Surf., A 294, 80-85 (2007).

41. Machatha, S.G. & Yalkowsky, S.H. Estimation of the ethanol/water solubility profile from the octanol/water partition coefficient. Int. J. Pharm. 286, 111-115 (2004).

42. Blaser, G., Sanderson, J.M. & Wilson, M.R. Free-energy relationships for the interactions of tryptophan with phosphocholines. Org. Biomol. Chem. 7, 5119-5128 (2009).

43. Chuang, C. & Chen, Y. Raman scattering of L-tryptophan enhanced by surface plasmon of silver nanoparticles: vibrational assignment and structural determination. J. Raman Spectrosc. 40, 150-156 (2009).

44. Aliaga, A.E. et al. Surface-enhanced Raman scattering study of L-tryptophan. J. Raman Spectrosc. 40, 164-169 (2009).

45. Ross, M., Steinem, C., Galla, H. & Janshoff, A. Visualization of Chemical and Physical Properties of Calcium-Induced Domains in DPPC/DPPS Langmuir−Blodgett Layers. Langmuir 17, 2437-2445 (2001).

46. Richter, R.P., Bérat, R. & Brisson, A.R. Formation of solid-supported lipid bilayers: an integrated view. Langmuir 22, 3497-3505 (2006).

- 210 -

Chapter 8: Review of nanoparticle optical probes: Biological labelling and toxicity

(Adapted with permission from Ip, S., MacLaughlin, C. M., Nguyen, C. T., Walker, G. C. “Photonic Nanoparticles for Cellular and Tissular Labeling”, in Biological Nanoscience, Vol. 2 , Walker, G. C ; Krueger, B., eds., of Comprehensive Nanoscience and Technology, Andrews, D.; Scholes, G.; Wiederrecht, G, eds. (Academic Press, London), 2010, 0000-0000. Copyright © 2011 Academic Press. Contributions form Nguyen, C.T. do not appear in this chapter. C.M. MacLaughlin identified and gathered references on metallic nanoparticles. G.C. Walker guided the planning of the original publication, supervised information gathering, and contributed in an editorial capacity)

8.1 General Introduction

Over the last decade, there has been an enormous amount of research into the use of nanoparticles as reporters in biological and medical diagnostics1‐5. This is chiefly because advances have been made, in controlling their size, shape, and surface chemistry, enabling tunable optical properties, adaptability to the variety of chemical environments, and targeting to specific biomolecular markers. Additionally, reliable, and affordable commercial sources for particles are increasingly available.

This chapter will provide general background information on the use of nanoparticles in cell and tissue labeling, along with general strategies for associating nanoparticles to specific targets. Synthesis of various nanoparticles will not be reviewed because there are many varied techniques reviewed elsewhere6‐13, and because commercial sources of many particles are readily available. Instead, in two sections dedicated to noble metal nanoparticles, and fluorescent semiconductor quantum dots (QDs) respectively, this chapter will: review the general physical principles that govern their unique properties,

- 211 -

explain how these properties are used in detection and imaging modalities, review specific examples from literature of surface modifications, provide a cross section of both in vitro cellular and in vivo tissular assays from recent literature, highlighting areas where further development would be valuable, and review recent literature on in vivo and in vitro toxicity studies.

8.2 Background

Interest in developing nanoparticle biodiagnostics stems from 2 main advantages: (A1) labeling of biological targets with high specificity can be achieved through well understood surface chemical modifications14‐33, and (A2) their enhanced optical properties6,9,34‐37 may improve detection sensitivity, or throughput. In addition to these advantages, there are three themes frequently cited in recent literature that have motivated studies reviewed herein: (T1) tuning optical properties into the near infrared (NIR), where water, and proteins scatter and absorb light less intensely, to enable signal collection in much deeper tissue (e.g.: liver in mice38) with lower background interference39‐43; (T2) taking advantage of narrow, tunable optical emission properties to make spectrally distinct probes with similar or identical surface modification routes, to simultaneously probe multiple targets15,23,44‐51; (T3) leveraging their potential as agents in photo-assisted therapy to develop combined, non-invasive detection and treatment modalities39,52‐57. T3 is beyond the scope of this chapter, but is mentioned biefly. The advantages (A1, A2) over molecular dyes, and these three themes (T1-T3) will be discussed with respect to semiconductor quantum dots (QDs), and noble metal nanoparticles (NPs) in their respective sections using specific examples from the literature. This section however, will discuss basic principles related to A1, then highlight how enhanced spectral properties (A2) motivates themes T1-T3. Themes T1 and T2 will then be discussed in terms of challenges, while leaving examples to the particle-specific sections.

- 212 -

8.2.1 Surface functionalization

Although the relative ease of surface modification of nanoparticles is considered one of the primary advantages of this technology (A1), it is also generally a necessity because unmodified particles either are not intrinsically water soluble, or are prone to aggregation. Thus one purpose of surface modifications are to impart solubility in aqueous environments, or prevent aggregation and preserve their properties under various salt, temperature, and pH conditions relevant to their intended application. For example, a common biocompatible surface coating is poly (ethylene glycol) (PEG), which has been shown experimentally to resist aggregation, reduce cellular uptake, and increase blood circulation times14,24,52,53,38,58‐63.

Another purpose for surface modifications is to enable nanoparticles to associate with specific molecular targets, thereby reporting the presence or locations of a target, or providing imaging contrast between the target, and non-targets. For instance, in cellular imaging, fluorescent nanoparticles can be directed to attach specifically to mitochondria, allowing them to be positively identified in a fluorescence image64. Similarly, NPs can be functionalized to accumulate in specific tissues such as tumor sites, after being injected into the bloodstream17,54,38,65‐68. Unique to tumor targeting, nanoparticles can accumulate passively in tumors due to the enhanced permeability and retention (EPR) effect that arises from the inherently leaky vasculature and poor lymphatic drainage of tumors, without the need for specific targeting moieties69.

There are common strategies for associating particles with specific targets that have carried over from existing strategies for targeting other markers commonly used in biological assays. In general, these methods make use of specific molecular interactions to promote the association of the particle with the target: Single stranded DNA is targeted by attaching copies of the target’s complementary strand to the particle’s surface, and hybridization binds the particle to the target strands70,71; proteins and peptides are

- 213 -

often targeted by grafting monoclonal antibodies against the peptide sequence to the NP surface, and the specific antibody/antigen captures the target peptides to the NP surface14,49,50,52,53,55,56,38,58‐60,72‐77; some molecular targets may also be targeted through known, specific ligand/receptor binding such as biotin/avidin25. In principle, any known association between two molecules can be used to target the nanoparticle to a specific ligand. However, to be practical, the recognition and complementation between targeting molecules and their targets must be reliable, and remain monogamous in the presence of myriad other molecular species. Furthermore, affinity between the complementary species must be long lived enough for the probe to be associated with the target for the duration of the measurement. Also, the targeting moiety must remain functional after being subjected to the variety of environments between the surface conjugation and the target.

Synthetic routes for surface chemical modification of semiconductor QDs and noble metal NPs tend to differ because stabilizing surface ligands used in their respective synthesis procedures differ drastically. Commonly, QDs are synthesized in organic solvents from inorganic precursors, and stabilized by hydrophobic surface ligands such as trioctylphosphine oxide (TOPO)6, whereas metal nanoparticles are synthesized in aqueous solutions from aqueous precursors, and stabilized by polar ligands such as CTAB and citrate7849. Surface modification of QDs usually requires the exchange of hydrophobic surface ligands with amphiphilic or hydrophilic ligands, followed by transfer to an aqueous solvent. Examples of surface modifications for each type of particle will be presented in their respective sections.

Despite these differences, there are two notable similarities in surface chemical modification routes. The first exploits the high affinity between sulfur atoms and metal atoms to attach thiol-terminated molecules to the surfaces of semiconductors and metals alike. For instance, a mercaptoacetic acid monolayer can form on a CdSe/ZnS QD with the thiol group binding to the Zn atoms on the surface and the carboxylate group

- 214 -

interacting with the solvent environment14. The details of the method will be reviewed in section 8.4.2. The second common strategy involves reacting an amine group, such as the N-terminus of a protein or antibody, to a free carboxylate group on the surface via ethyl dimethylaminopropyl carbodiimide (EDC) and N-hydroxysulfosuccinimide (Sulfo-NHS) chemistry. The details of EDC/Sulfo-NHS chemistry will be reviewed in section 8.3.4.2. Alternatively, bioconjugation may also be achieved through electrostatic interaction between a charged protein and a pre-existing surface ligand of opposite charge26. Although these methods are used frequently in literature, there are other surface modification routes that will be reviewed for each particle.

8.2.2 Enhanced optical properties

The physical origins of enhanced optical properties are discussed in detail in the sections dedicated to the two types of particles. This section deals instead with how enhanced optical properties (A2) give rise to two important themes in recent literature, namely NIR optical emissions (T1) and multiplexing (T2).

QDs with fluorescence emission in the NIR, and metal NPs with surface plasmon resonances in the NIR are desirable because water and proteins tend to have the lowest absorption and scattering at wavelengths between 650 and 900 nm79. Therefore, the penetration of excitation light, or of signals emitted from the particles is greater in this window than for other wavelengths. Practically, in an in vivo imaging construct, NP probes active in the visible are useful for detecting tissue close to the skin but would require additional surgical methods to gain optical access to deep tissue or organs. NIR- active probes on the other hand, are optically accessible in deeper tissue or organs without the need for surgical intervention. Additionally, the ability to tune the spectral features of both QDs and metal NPs into the NIR through changes in size or shape with little or no change in surface chemistry preserves advantage A1.

- 215 -

Both size tunable emissions, and narrow optical emission bands contribute to the use of QDs and metal NPs in multiplexed detection formats. Size tunable emissions allow spectrally distinct particles to be made easily, while often preserving the general method for surface functionalization. Narrow emission bands allow multiple distinct probes to be detected simultaneously within a reasonable detection bandwidth. Alternatively, multiple NIR probes can be made within the low scattering and absorption window of water and proteins between 650 and 900 nm, for multiplexed detection in deep tissue. While QDs tend to have narrower fluorescence emssion bands than molecular fluorophores, Raman scattered light from metal nanoparticles have, by far, the narrowest emission bands (up to 30 times narrower than QD emissions80).

The third theme, combining detection and theraputics (T3) is also largely enabled by the enhanced optical properties of nanoparticles, namely the tunable absorption band. This will be briefly discussed.

8.3 Plasmonic noble metal nanoparticles

8.3.1 Introduction:

The plasmonic properties of noble metal nanoparticles have enabled their use in a variety of diagnostics aimed at capturing and detecting soluble analytes such as viruses, specific DNA sequences, proteins, or other small molecules2. Other specific targets of detection include anthrax spores81, biomarkers for Alzheimer’s disease82, and glucose or lactose concentrations83. Their most common application however is in commercially available pregnancy testing devices, which report the presence of soluble hormones that are produced during pregnancy via the change in colour that occurs when labeled nanoparticles aggregate around the target analyte in soution.

- 216 -

In additon to colorimetric transduction of analyte detection, the localized surface plasmon resonance of the particles also contributes to intense light scattering and enhanced electromagnetic fields at the surface of particles. These properties make them suitable as contrast agents in darkfield microscopy 53,59,54,55,57,49,73,74,84,66, as soluble probes for (surface enhanced) Raman spectroscopy and imaging 85,38,75,60,86, and as contrast agents in optical coherence tomography39,87,88, and optoacoustic imaging89‐91. These properties make metal nanoparticles promising candidates for in vitro and in vivo optical labeling for the detection, identifiation, and imaging of cells or tissue. For example, to detect cellular surface markers that can be used to label and identify cancer cells at early stages. These applications will be the primary focus of this section.

There are several evident advantages to develping plasmonic nanoparticle probes for in vitro and in vivo diagnostics. First, their size-tunable optical properties enables the LSPR to be tuned to the NIR pass-band of water and proteins79, enabling signal readout in deep tissue. Secondly, compared to fluorescence, the relatively narrow optical resonances of SERS nanoparticles allow the detection of more SERS probes than fluorescence probes over the same optical bandwidth. This enables the simultaneous, multiplexed detection of more cell surface markers than would fluorescence. Finally, in addition to scattering of incident light, these nanoparticles are efficient absorbers that convert absorbed energy to heat. Local temperatures have been reported to reach those capable of hyperthermal destruction of cells53‐55,57,39,52,76,56,92‐95, fueling studies into their use as combined diagnostic/therapeutic agents39,57,52,72.

8.3.2 Localized Surface Plasmon Resonance:

Metal nanoparticles derive their unique optical properties from their localized surface plasmon resonance: a highly localized oscillation of surface electron-density driven by an external optical field 96,97. The LSPR resonance is known to depend on the NP’s size, shape and composition97349698. Typically, anisotropic particles with diameters in the tens

- 217 -

of nanometers composed of noble metals such as gold, silver and copper, have LSPR resonances in the visible range97. The decay of excited surface plasmons can take place through heat evolution, or by Rayleigh or Raman scattering99.

Experimentally, the LSPR is manifest as a strong peak in the absorption spectrum of a colloidal metal nanoparticle suspension, which is responsible for the observed colour of the suspension. LSPR also manifests as an enahnced electromagnetic field surrounding the nanoparticle, with field strength that decays exponentially from the particle surface. The field enhancement is partially responsible for Surface Enhanced Raman Scattering spectroscopy, to be discussed in Section 8.3.4.

8.3.3 Tuning Metal Nanoparticle Surface Plasmon Properties:

As mentioned above, some of the advantages of metal NPs stem from their size and shape tunable properties, which contribute greatly to their adaptability and versatility as probes for a variety of applications. For example, NPs with LSPR resonances in the NIR band- pass window of tissue can avoid the compliations that arise from the autoflorescence of cellular components, and improve the depth at which optical excitation and signal collection can be achieved. In addition to wavelength, the size, shape and composition of the particles also affect the scattering and absorption components of the their optical extinction, which can also have a profound impact on their suitability for different applications. For example, efficient scattering is desirable for imaging and optical-probe applications, while efficient absorption (and heat evolution) is desirable in photo-assisted therapeutics. This sub-section summarizes the tunable properties of several nanostructures including nanosperes, nanorods, and nanocages.

Perhaps the most commonly studied nanoparticle, spheroids are easily synthesized and available through numerous commercial sources. They however have the most limited tunability among the various shapes, with diameter being the only adjustable parameter

- 218 -

for a given composition. For example, a change in diamter from 20 nm to 80 nm results in a shift in the wavelength of the LSPR peak from 520 nm to 580 nm100. This limited tunability limits their applicability where NIR LSPR wavelengths are desirable. Furthermore, given the width typical of LSPR bands, the number of different LSPR resonances that can be discerned by simultaneous detection is also limited over this range. While this range of 60 nm in the spectral shift is small considering the width of the plasmon band, a change of 60 nm in the particle diameter is significant in terms of its impact on the potential applicability of the particle, for example, its in vivo fate. The diameter however , can have a significant impact on the extinction coefficient, especially the scattering component101,100. This effect is clearly seen in Figure 8-1 c).

Nanorods are a natural extention of nanosperes that adds an additional length dimension and consequently, aspect ratio as geometric parameters for tuning. Their anisotropy results in two LSPR resonances that correspond to their short and long axies, where the latter corresponds to a stronger resonance at a longer wavelength than the former. The dominant long-axis resonance is widely tunable via the aspect ratio of the rod, allowing for resonances from the visible to the NIR, as shown in Figure 8-1 a). Additionally, nanorods exhibit size-dependent extinction coeffiecients that are independent of appect ratio. Also, at much smaller effective sizes, nanorods have comparble absorption and scattering properties as isotropic nanoparticles. This is a consequence of having extinction coefficients that are an order of magnitude greater per unit effective length than isotropic particles such as nanospheres and nanoshells100.

- 219 -

Figure 8-1: a) Tunability of Au nanorod surface plasmon resonance with increasing nanorod aspect ratio. B) Tunability of resonant wavelength of silica core-gold shell nanoshells with altering core-shell thickness ratio. C) Ratio of scattering to absorbance of incident light as a function of nanoparticle diameter for gold nanospheres99. Reproduced with permssion from Jain, P.K., Huang, X., El-Sayed, I.H. & El-Sayed, M.A. Noble Metals on the Nanoscale: Optical and Photothermal Properties and Some Applications in Imaging, Sensing, Biology, and Medicine. Accounts of Chemical Research 41, 1578-1586 (2008). Copyright © 2008, American Chemical Society.

Nanoshells are either bimetallic60,80or dielectric/metallic102 core/shell structures, which in addition to diameter, have the thickness of the shell as a tunable geometric parameter. Nanoshells behave similarly to nanospheres with respect to their diameter-dependent

- 220 -

properties100. Variations in shell thickness or core-to-shell ratio allows the LSPR to be tuned into the NIR102,100. In the case of silica/gold core/shell structure having a core diameter of 60 nm, varying the shell thickness from 5 nm to 20 nm results in a decrease of 300 nm in the resonance wavelength, as demonstrated in Figure 8-1 b).

- 221 -

Gold nanocages are hollow metallic or bimetallic cubes that offer two geometric parameters that influence their LSPR 13,56,87,88,103, namely edge length and shell- thickness, as demonstrated in Figure 8-2. As with all particle geometries, extinction, and especially scattering, scale with particle size72. Depending on synthetic parameters, nanocages can also have varying porosity, which does not affect LSPR wavelength, but does reduce scattering cross section72.

8.3.4 Surface Enhanced Raman Spectroscopy (SERS):

Raman spectroscopy is a method that probes vibrational and rotational modes of molecules. When incident light is scattered by a sample, it is either scattered elastically (Rayleigh scattering) or inelasticlally via phonon interaction (Raman scattering). If monochromatic light is used (or nearly monochromatic such as that from a laser source), the energy of the Raman scattered light can be either increased or decreased by a magnitude equal to that of the phonon. Typically these vibrational and rotational modes are orders of magnitude lower in energy than electronic modes, and thus the wavelength of Raman scattered light is shifted only slightly to the blue or red from the incident light. Consequently, notch filters or Bragg filters can be used to separate Raman scattered light from the much more abundant Rayleigh scattered light.

For any given scattering event, the probability of Raman scattring is much lower than Rayleigh scattering. Surface enhanced Raman scattering can amplifiy Raman scattered signals by14 to 15 orders of magnitude104‐106. The SERS effect can be attributed to two phenomena that contribute independently to the increase in Raman scattered light, namely electromagnetic enhancement and chemical enhancement mechanisms105,35. The electromagnetic enhancement arises from the increase in local field strength at or near the surface of a metal nanoparticle or on surfaces curved or roughened on similar lengthscales104,107. LSPR resonances clearly contribute to the enhancement of localized

- 222 -

electromagnetic fields. When a Raman active molecule enters the enhanced localized fields, the number of photons it can interact with increases dramatically. That is, the electromagnetic enhancement does not necessarily increase the probability of a Raman scattering event to occur, but rather increases the total number of Raman scattering events by increasing the number of photons the molecule interacts with. Further field enhancement can occur when plasmon fields of neighboring particles overlap, creating a “hot spot” 130,108,109,106. Chemical enhancement occurs via an independent pathway: Raman active molecules that chemisorb to the surface of a metal nanoparticle can form charge-transfer states between the molecule and the metal. The Fermi energy of the metal falls at an intermediate state between the HOMO and LUMO states of the chemisorbed species, and acts as a intermediate state for a HOMO-LUMO transition, thereby lowering the energy of such a transition by about half105. Furthermore, if there is an overlap between the absorption resonances of the nanoparticle and the Raman active species, Surface Enhanced Resonant Raman Scattering (SERRS) can occur, providing further enhancement of the Raman signal104106.

8.3.4.1 Potential Advantages of SERS as an imaging modality:

Compared to fluorescence probes, SERS nanoparticles have threee primary advantages that can be understood in the context of identifying diseased cells by detection of specific cell surface markers. While most cells from a given multicellular organism carry nearly identical copies of nuclear DNA, differentiation of cell types occur by differences in expression of these genes, and consequently, different cell types can be identified by the differential expression of various cell surface proteins. Some diseases –most notably cancer- can have a profound effect on the expression gene products (i.e. phenotype) both cytoplasmic and membrane-bound, and consequently afflicted cells often have different expression of surface protiens than healthy cells. However, because the genotype of the healthy and diseased cells are the same, accurate and reliable identification of the disease state requires the simultaneous differential detection of multiple cell surface markers.

- 223 -

While the antibody-based labeling of cell surface markers can be achieved using fluorescent probes, the width of the fluorescence emissions (~ 10s of nm) from a variety of probes ultimately limits the number of probes that can be simultaneously identified within a given bandwidth. Raman bands on the other hand are one-thirtyith the width of QD fluorescence bands on average110, meaning that many more SERS probes can be identified over the same spectral bandwidth. In fact, some reports in literature have demonstrated the simultaneous detection of 5 different Raman probes51 targeted to mouse tissue.

An additional advantage is that unlike fluorescent probes, SERS probes do not photobleach, which would result in the loss of signal over long imaging or data collection times. The final advantage is the greater availability of SERS probes in the NIR window region, where scattering and absorption of incident light by tissue, proteins and water are minima, which reduces the noise and increases the penetration depth of signals in biological samples. There are comparatively few fluorescence probes that can be both excited and detected within this window79. Thus, for deep tissue applications, SERS probes potentially offer a multiplexing advantage.

8.3.4.2 Surface modification and functionalization:

There are several aspects to functionalizing noble metal nanoparticles that differ considerably from the techniques described for semiconductor QDs. First, as mentioned, Ag and Au nanoparticles are often synthesized with polar citrate111 or CTAB surface- capping ligands78, and thus, unlike QDs, do not require ligand exchange and solvent transfer steps. Accordingly, surface functionalization reactions are performed in aqueous solution. Secondly, direct protein-particle bioconjugation is possible, where a cysteine-

- 224 -

terminated protein or peptide is adsorbed to the surface through the thiol-metal interation. Finally, if SERS is required, a Raman reporter must be bound to the surface10685. In the case of Nguyen et al.112, the Raman reporter is not chemically bound to the surface, which introduces a potential complication; The Raman reporter must remain physisorbed following subsequent bioconjugation steps.

The majority of published studies utilize very similar strategies for surface functionalization. The general strategy can be illustrated by example: Starting with the stock, citrate coated gold nanoparticles, a charged dye species such as malachite green isothiocyanate (MGITC) is added to the paricle suspension, where the positively charged dye associates with the negatively charged citrate. The particle is then protected using a thiolated polyethylene glycol (PEG-SH) that binds to the particle surface through a thiol- gold interaction. At this junction, the PEGylation of the particle can be observed via electron microscopy and dynamic light scattering. Additionally, the SERS spectrum of MGITC can measured as well.

For targeted SERS nanoparticles, antibodies against the target of interest can be conjugated to the surface. Starting with the PEGylated particles, a heterobifunctional carboxy-PEG-thiol is reacted with the particles. The thiol group bonds to the surface, leaving a mixed monolayer of PEG and COOH-terminated PEG. Surface vacancies are backfilled with excess PEG-SH. Activation of the coboxy groups is initiated using the ubiquitous ethyl-dimethylaminopropyl-carbodiimide(EDC)/N- hydroxysulfosuccinimide(SulfoNHS) chemistry, which which conjugates the N-terminus of proteins and peptides, including antibodies to free carboxy groups. Added monoclonal antibodies against the intended target become conjuaged to the HS-PEG-COOH and particles are separated from unbound antibodies by repeated centrifugation and resuspension in fresh buffer.

- 225 -

8.3.4.3 In vitro and in vivo SERS labeling

There are primarily three methods of utilizing metal nanoparticles as in vitro cellular probes: (1) deliberately labeling the particles with a Raman active dye species and detecting the spectrum of the dye85, (2) in the absense of a dye, monitoring the Raman scattering of the surface functionalization species such as CTAB or targeting antibodies73, or (3) using dye-free particles that enhance the Raman scattering of molecules in its immediate environment74.

The first modality is of particular relevance for this thesis, and is therefore reviewed in more detail by highlighting an example from literature. In a paper by Qian et al.38, SERS nanoparticles featuring diethylthiatricarbocyanine (DTTC) as a raman reporter, are targeted to EGFR-overexpressing tumor cells (the human head and neck carcinoma cell line Tu686) using single chain fragment antibodies (ScFV) against EGFR. The ScFV is conjugated to carboxy-PEG via the same EDC/NHS strategy discussed in the previous section. The experiment included several controls: (1) Control particles were prepared with non-specific IgG antibodies, (2) targeted particles were incubated with cells from a carcinoma cell line that does not expess EGFR (NCI-H520 cells). The results are summarized in Figure 8-3. The anti-EGFR labeled DTTC-SERS particles were detected on the cell line that overexpressed EGFR, producing a spectrum similar to the pure DTTC-SERS particles. No particles were detected on EGFR-negative cell cultures. Likewise, control experiments where cells have been incubated with particles that have not been conjugated with ScFV, or have been conjugated with non-specific IgG do not show the spectrum of DTTC SERS, suggesting no particles were bound.

- 226 -

In that same study, in vivo targeting was demonstrated using xenografts of the same Tu686 cell line on nude mice. In this case, malachite green-labeled particles were injected into the tail vein. Raman spectra were collected from a live mouse using a 785 nm laser focused through a microscope, with backscattered signal collected though the same objective, without surgical procedure. Spectra collected at the tumor site, which was near the skin surface showed the distinct SERS spectrum of malachite green. A control group that received particles without targeting moieties showed significantly less signal at the tumor site. For control experiments, spectra were also taken over the leg and above the liver and spleen. Accumulation of SERS particles were detected in both the liver and spleen for both targeted and non-targeted particles.

- 227 -

8.3.4.4 Toxicity of metal nanoparticles

With their potential as medical diagnostic tools, the potential hazardous and toxic effects of nanoparticles will require detailed characterization and quantification. While, the safety or toxicity of nanoparticles depends ultimately on their properties113,114 (size, composition, surface charge, and surface chemistry) the consequences of these properties on cells, tissue, and on an organism as a whole requires both in vitro and in vivo characterization. In vitro tests focus primarily on the cellular and tissular response to the particles; the uptake, distribution, and breakdown of the particles by cells, and the eventual fate of all products, and their effect on the viability of cells and tissue. In vivo toxicity tests focus on the effect of the particles on organisms as a whole; the absorption, distribution, metabolism, excretion, and immunological response within a live organism115,116,117. However, in vivo uptake, distribution, and accumulation do not necessarily correlate directly to toxicity. Additionally, accumulation and biodistribution of the particles can potentially impact their diagnostic accuracy.

8.3.5 In vitro toxicity of metal nanoparticles:

In vitro toxicity studies typically involve exposing confluent cultures of established cell lines to nanoparticles suspensions, and evaluating the uptake and compartmentalization within the cells, and cell morphology. Frequently, cell death rates are measured using standard assays. Occasionally, gene expression is monitored as well. In literature, nanoparticles of various shapes, sizes, and surface coatings have been tested to evaluate the influence of these factors on cellular toxicity. Various trends emerge among different studies.

In the case of nanorods118,119 and nanospheres116, surface charge appers to affect uptake of the particles, with positively charged particles (coated with CTAB for example), and negatively charged particles (carboxylate coatings) being the most and least efficiently internalized respectively. One study reports the efficiency of uptake of anionic particles

- 228 -

to be 6% that of cationic particles118. It has been reported that a combination of the homeostatic membrane potental and the relatively high concentration of anionic lipid species creates an electrifield that attracts cationic proteins and peptides120. This phenomenon may favour the adsorption of cationic particles to the membrane, thereby increasing the probability of uptake compared to other particles.

Uptake, however, does not necessarily correlate with toxicity. In most cases, despite significant particle uptake, most cells remained viable for at least several days118,119. Longer-term studies would be required to determine long-term effects. Furthermore, the presence of the particles within the cell did not appreciably alter gene expression, with one study reporting that 0.35% of some 10000 genes monitored were down regulated, while the rest remained at base line levels.

After uptake, particles appear to enter the endocytic pathway116 where degredation of particle components such as surface coatings and the particle itself might occur. Interestingly however, even the uptake of known toxic materials used in the surface coatings of the particles had only a small effect on toxicity119. Large amonts of unbound CTAB, however, was found to cause acute toxicity. After several centrifugation and wash steps to remove unbound CTAB form the particles, the apparent toxicity of the particles is dramtically reduced116.

In summary, it would appear from these studies, that surface charge is the primary determining factor for cellular uptake. While the size dependence of the particles was not adequately studied, it appears that trends in the surface-charge effects appear consistent among nanorods and nanospheres alike. With the surface charge dependence in mind, a systematic study of particle size with tightly controlled surface charges would be required to properly characterize the size dependence. Additionally, long-term cell viability tests are required to determine the chronic effects of particle uptake.

- 229 -

8.3.5.1 In vivo toxicity of metal nanoparticles:

In vivo toxicity studies are typically aimed at characterizing the effects of systemic injection of nanoparticles into model organisms to assess the fate of the particles. The distribution of the particles within an organism, as well as the eventual elimination and/or accumulation of particles have important consequences for not only potential toxicity, but also the effectiveness of the particles as probes for disease. Non-specific accumulation of particles in healthy tissue will lead to false positive identification of disease, and may eventually lead to inflammation of the tissue.

Thus evaluating the biodistribution of nanoparticles in model organisms. In such a study, the fate of PEG-coated 13 nm gold nanoparticles injected into the tail vein of mice was evaluated63. The particles remained in circulation for 30h, and accumulated in the liver and spleen over a 7-day period. The particles were also detected in the brain, kidneys, lungs, and testis but baseline levels were restored within a 7-day period. Accumulation in the liver and spleen resulted in an inflammatory response that induced cell apoptosis at rates that depend on dosage. Histopathological data confirms the accumulation of gold in the phagocytic cells of the liver and spleen, and not in other cell types within these organs, and no penetration of nuclei was observed in any histology samples. While, many claims are made about the non-toxic and biocompatible properties of colloidal gold, these results clearly show toxicity in the form of acute inflammatory response and cell apoptosis. Clearly toxicity studies using particles of various size, composition and surface functionalizations are required for a through understanding of the impact of these other parameters.

- 230 -

The effect of size on the biodistribution of particles was studied using dendrimer- encapsulated particles with different surface charges injected into the tail vein of mice114. In that study, the smallest cationic particles at 5 nm diameter had the longest blood- circulation time. Comparing particles of like surface charge, the smallest (5 nm) particles exhibited the least recognition from the host immune system and the least organ-specific accumulation. Interestingly, they were also excreted in the largest volumes through urine and feces. This suggests that the larger particles tend to accumulate in organs, which reduces their circulation half-lives. On the other hand, accumulating in certain organs also precludes efficient excretion. Given the findings of the previous example, the conclusion might be that smaller particles are less likely to be toxic since they are less likely to accumulate in organs and therefore less likely to elicit an inflammatory response.

In addition to size, nanoparticle composition can greatly affect toxicity. A study involving zebra fish embryos evaluated the morphological changes to gill tissue due to various sizes of gold and silver nanoparticles and gold and silver ions in their environment121. The study found that regardless of size, the gold particles had very low toxicity compared to silver, with gold being non-toxic for all concentrations studied. Silver nanoparticles of all sizes and ions on the other hand, showed toxicity scaled with concentration. This implies that choice of nanoparticle composition can have a drastic impact on toxicity.

8.4 Fluorescent semiconductor quantum dots

As mentioned in section 2, fluorescent semiconductor quantum dots (QDs) have primarily two advantages over molecular dyes that make them desirable as labels for cells and tissue, namely, fairly well understood surface modification routes that enable targeting specificity (A1), and enhanced optical properties (A2). Although these

- 231 -

advantages were discussed in general for nanoparticle probes, this section will highlight these advantages for quantum dots in particular.

Following a section that reviews the physical principles that lead to A2 and A1, will be a section dedicated to surface funtionalization and trends in literature therein. Followed by sections on in vitro cellular imaging and in vivo tissue imaging which bring to light, the two themes in recent literature (T1, T2). QDs are also widely used in solution phase assays to detect molecular analytes44, but this application will not be discussed. The final section addresses toxicity of QDs, both in vitro and in vivo.

8.4.1 Background

Fluorescent semiconductor quantum dots (QDs) are desirable probes in bioassays because they exhibit higher quantum yields, narrower emission bands, and better resistance to photobleaching than molecular fluorophores6,27. Additionally, they exhibit size dependent emission spectra allowing their emission to be tuned across most of the visible6,122‐124 and near IR spectrum40,41, while maintaining a broad absorption. These properties are desirable for the simultaneous labeling of several different targets: several quantum dots with distinct emission spectra can be chosen to minimize spectral overlap. Such control would be difficult with molecular probes where custom optical properties are not as easily engineered.

These unique optical properties (A2) come about as a result of both their size and their material properties. Semiconductor quantum dots (QDs) are typically composed of a binary II-VI compound such as PbS, CdSe, or ZnSe and may also be composed of two binary compounds synthesized in a core-shell structure. In the bulk, semiconductors behave differently than nanocrystals. A bulk semiconductor has a bandgap: the difference between the lowest edge of the conduction band and the highest edge of the

- 232 -

valence band. Direct bandgap semiconductors such as ZnSe, tend to have light emitting properties and are used in light emitting diodes (LEDs)125. Valence band electrons are excited into the conduction band either by an applied voltage or by absorption of a photon. The excited electron leaves behind an electron vacancy in the valence band called a hole, and the resulting electron-hole pair is often referred to as an exciton. When the exciton recombines, it may release a photon with a wavelength determined by the bandgap energy (which depends on the composition of the crystal). The exciton may also decay non-radiatively, by coupling with a phonon and releasing the energy as heat.

As the physical size of the semiconductor crystal approaches the size of the Bohr radius of an exciton, typically 1-5 nm depending on the material14, an exciton can be confined to the particle in 3 dimensions, and energy levels available to the electron above and below the bandgap are no longer continuous, but become discrete. This was first observed in colloidally grown semiconductor nanocrystals by L. Brus122,123, who drew an analogy between the discrete energy levels of the QD, and those of molecular orbitals. Figure 8-4122 schematically shows the transition from bulk electronic properties to those of a QD. The energy difference between the lowest discrete conduction band state and the highest discrete valence band state increases with decreasing particle size. This changes the energy that must be released, and thus the wavelength of light emitted, upon recombination. All things being equal, smaller QDs have greater difference between the lowest excited state, and highest ground state, and therefore have shorter emission wavelengths than larger ones. In fact, Brus’s calculations123 indicate that the excited state energy goes as R-2 following the particle-in-a-box solution, but accounting for the effective mass of the electrons and holes in the crystal and describing their interaction as a shielded coulomb interaction. Therefore, the fluorescence emission can be tuned through the particle size. Additionally, the confinement of the exciton can also increase the probability of radiative recombination in indirect bandgap materials such as Si126,127.

- 233 -

Figure 8-4: Comparison of electronic energy levels of bulk semiconductors having a continua of valence and conduction states (left), and semiconductor QDs having discrete conduction and valence sates (right).122 Reproduced with permission from Brus, L. Electronic wave functions in semiconductor clusters: experiment and theory. The Journal of Physical Chemistry 90, 2555-2560 (1986). Copyright © 1986, American Chemical Society. Reprinted with permission.

Also as a consequence of the quantum size effect, emission spectra of QDs can be broadened by polydispersity in QD size. Consequently, the interest in pursuing QDs in a variety of applications has been driven by the ability to reliably produce monodisperse QD samples of desired sizes. Many techniques are based on the method described by Murray et al.6. The details of the synthesis are beyond the scope of this review, and syntheses have been extensively reviewed in literature6,8‐10. Their results, however, demonstrate the ability to produce monodisperse samples with well-controlled size. Figure 8-5, excerpted from their paper, demonstrates both the quantum size effect, and

- 234 -

the control over the absorption peak afforded by their synthesis. They report that the dependence of the absorption energy on the particle size deviates somewhat from calculation predicted by Brus. They also demonstrate the ability to adapt the method to produce CdSe, CdS, or CdTe QDs.

Figure 8-5: Illustration of quantum size effect. The absorption peak appears at lower wavelengths for smaller particles.6 Reproduced with permission from Murray, C.B., Norris, D.J. & Bawendi, M.G. Synthesis and characterization of nearly monodisperse CdE (E = sulfur, selenium, tellurium) semiconductor nanocrystallites. Journal of the American Chemical Society 115, 8706-8715 (1993). Copyright © 1993, American Chemical Society.

As a consequence of their large surface area to volume ratios, the surfaces of QDs can have a profound effect on their optical properties. Dangling bonds on the physical edges of the crystal or the defects in the crystal structure several atoms deep, caused by

- 235 -

restructuring of the crystal surface to satisfy these dangling bonds, can lead to the presence of surface trap states6,122‐124. These states have energies within the bandgap of the QD as shown in Figure 8-4 and assist in the non-radiative recombination of the exciton leading to lowered quantum yield. These surface states can be mostly eliminated by growing a core-shell structure, where the shell material has the same lattice dimensions as the core so as to minimize strain in the crystal at the interface, and a wider bandgap than the core to maintain quantum confinement9,124. A common example is a CdSe core with a ZnS shell.

As will be discussed in detail in subsequent sections, the tunability and sharp optical features have generated interest in recent literature to develop both NIR (T1) and multiplexed (T2) cellular and tissular imaging modalities. The former can be achieved by using CdTe as the core material, while the latter can be achieved in part by developing a series of monodisperse QDs of different size.

Originally, as a consequence of their high quantum yields, and resistance to photobleaching, quantum dots were attractive alternatives to organic dyes for noninvasive in vivo probes, and as imaging contrast agents. There are however, several barriers to using QDs as either in vivo, or in vitro probes. Chan et al.14, in a highly cited paper on the topic, highlight some of these barriers, and demonstrate a QD-based in vitro labeling system that overcomes these barriers. The two barriers identified were: 1. Water solubility: Colloidal QDs are inorganic materials that are synthesized

and soluble in non‐polar solutions

2. The need to chemically modify the QD surface with targeting molecules while

preserving the optical properties of the QD

Chan et al. achieved this by starting with a CdSe/ZnS core-shell QD, and employing a mercaptoacetic acid monolayer to coat the QD surface (Figure 8-6). The mercapto group

- 236 -

forms a strong, high-affinity bond with the Zn on the surface, while the carboxyl end acts both to make the particle water soluble, and as a precursor to the addition of amine species such as proteins, peptides, and antibodies through EDC chemistry.

The surfaces are modified by suspending the hydrophobic particles in chloroform and reacting it for 2 h with glacial mercaptoacetic acid. PBS is then added to 1:1 volume ratio and shaken with the chloroform/QD solution. When the chloroform and buffer separate, the aqueous portion, containing the QDs, is retained. Ensemble fluorescence spectra show that the water-soluble, and bioconjugated QDs show no change in optical properties as a result of the surface modifications. SEM images indicate that the QDs remain unaggregated after modification.

- 237 -

Figure 8-6: Illustration of functionalized quantum dot used by Chan et al., featuring hydrophilic surface groups, and bioconjugated protein.14 Reproduced with permission from Chan, W. & Nie, S. Quantum dot bioconjugates for ultrasensitive nonisotopic detection. SCIENCE 281, 2016-2018 (1998). Copyright © 1998 The American Association for the Advancement of Science. Reprinted with permission.

The biocompatibility and labeling were demonstrated in an in vitro assay, whereby transferin-conjugated QDs were applied to HeLa cell cultures. Observations under a fluorescence microscope showed the cells had internalized the dots through receptor- mediated endocytosis. As a control, unconjugated mercaptoacetic acid QDs were not endocytosed, and no fluorescence was observed. Their utility in an in vitro immuno-assay was also tested by conjugating immunoglobulin G (IgG) to the particles. The addition of a polyclonal antibody recognizing the FAb portion of the IgG caused the IgG-QDs to

- 238 -

aggregate, while the presence of BSA as a control did not, indicating that antibodies can be conjugated to QDs and display a recognizable epitope.

The two barriers Chan et al.14 identified are areas that continue to garner significant interest as they apply broadly to the application of most nanoparticles in biological contexts. Consequently, innovations continue to be made towards more robust or more versatile surface modifications. These innovations are perhaps motivated by other trends in recent literature such as T1 and T2 discussed at the beginning of this section. In the following sections, some examples of recent innovations in surface functionalization will be reviewed, followed by recent examples of in vitro and in vivo applications that follow T1 and T2. Since there have been a number of recent reviews128‐130 that overlap with these topics, the focus of the following sections will be on literature published in 2007 to 2009.

8.4.2 Surface functionalization

As stated before, one of the advantages (A1) of QDs is the ability to modify their surface chemical properties. This allows QDs to be transferred from organic solvents, in which they are typically synthesized, to aqueous solvents where they are most often used. Additionally, it allows QDs to be conjugated with different functional groups, such as PEG to increase in vivo circulation times or antibodies to impart selective labeling. The versatility of applications of quantum dots seen in literature is correlated with the variety of surface functionalizations available. Thus, one of the research trends in surface chemical modifications of QDs involves developing a generalized surface modification technique that can be easily adapted to attach a variety of different functional or targeting groups to the surface. A generalized surface functionalization route is highly desirable for multiplexed detection (T2), because it allows the same technique to be applied to functionalize a variety of QD of different sizes, with different molecular markers. Additionally, versatile strategies for surface functionalization are also desirable because

- 239 -

there are many molecular interactions that can be used to target nanoparticles. Although antibodies are commonly chosen for their high specificity to their target, there are cases where peptides or small molecules are more desirable. A very important example with clinical significance is pointed out by Diagaradjane et al. (2008), which is reviewed in section 8.4.4. This section illustrates strategies for versatile surface functionalizations by way of examples from recent literature.

The simplest surface functionalization methods tend to follow the model presented by Chan et al., where the intrinsic organic monolayer is first exchanged for heterobifunctional ligands with one thiolated end, which are subsequently used as sites to covalently link PEG or targeting moieties following exchange to an aqueous solvent. Direct cunjugation of thiol-containing proteins may not be possible. Although the direct conjugation of a cystiene-terminated protein to noble metal NPs is possible, the intrinsic hydrophobic monolayer requires that a ligand exchange be performed in an organic solvent such as chloroform, making direct bioconjugation unsuitable, unless the protein can be reversibly transferred between water and chloroform.

In another early example, Bruchez et al.27, coat hydrophobic CdSe/CdS core/shell QDs with SiO2 to make them water soluble, precluding the use of thiol-linked surface functional species. Instead, they demonstrate two methods of targeting these particles. In the first, silanized SiO2 coated QDs are shown to preferentially label the nucleus of cells. In the second, they covalently link sulfo-NHS-terminated biotin groups to the surface and use this to selectively stain streptavidin functionalized actin in fibroblasts.

In general, the surface functionalization serves not only to make the particle water- soluble but also to promote the association of the particle with a desired target. In many cases, the more challenging of these two goals is the targeting of the particle because it requires a chemical or robust physical interaction between the targeting moiety and the

- 240 -

surface coating molecule. The targeting moiety and the surface-attachment moiety can be reacted first to form a complex that can then be attached to the particle, or the targeting moiety can be attached to the surface coating after it has associated with the particle. Often the latter is necessary, as it is often necessary to apply the surface coating to the particle in non-polar solvents before transferring the coated particle to an aqueous buffer solution: targeting molecules in the form of peptides, proteins and antibodies are often incompatible with non-polar solvents, necessitating their conjugation in biological buffer.

Generally, the covalent attachment of proteins can be achieved through the reaction of the amine terminus to a surface attachment group with a free carboxylate, through EDC chemistry. Many surface functionalization schemes, therefore, involve the adsorption of a carboxylate containing species to the particle surface.

A recent article by Dubois et al.28 describes a versatile system in which carbon disulfide is spontaneously reacted with a range of amine containing targeting species, forming a dithiocarbamate on the targeting species, which can then be adsorbed to a QD surface. Dubois et al.28 claim there are three main advantages to their method:

1. It is modular: many amine containing species, including some peptides can

be conjugated to the carbon disulfide, and they demonstrate this with a

variety of species

2. The chemical reaction is easy to perform: unlike the EDC chemistry, this

reaction between the amine, carbon disulfide, and QD can be performed in

one step, albeit in an organic solvent such as chloroform.

3. The dithiol linker forms a more robust interaction with the surface than a

single thiol can, making QDs modified in this fashion more stable

- 241 -

They demonstrated the attachment of the ligands shown in Figure 8-7, which includes some small hydrophilic amino acids, and some short (4 amino acids) hydrophilic peptides. A slightly modified preparation technique was required for making hydrophilic

QDs: the hydrophilic peptides, and CS2 are still mixed together in organic solvent, but in the presence of tetramethyl ammonium hydroxide, which presumably acts as a surfactant. It is thus unclear how effective this method would be for attaching larger water-soluble proteins.

- 242 -

Dubois et al.28 point out that more chemistry might be possible to attach functional groups to the free carboxylate of a protein, or C-terminus of the peptide. This would likely involve the well-known EDC chemistry, which annuls advantage ii.

Dubois et al.28 highlight several interesting observations. First, the replacement of the TOPO coating, observed by detecting the 31P NMR signal that arises from freely dissolved TOPO that has been exchanged by the dithiocarbamide, occurs only when both the CS2 and amine precursors are mixed with the QDs; nither precursor alone is insufficient for ligand exchange. Second, QDs with thin to no ZnS shell, exhibited quenching after ligand exchange, whereas QDs with thick or multilayered shells showed little change in photoluminescence spectra after the exchange.

Another recent example of a versatile QD functionalization system employs an amphiphilic polymer that wraps around the QD29,30. A schematic illustration of the system is shown in Figure 8-8. A reactive polymer backbone enables side chains bearing hydrophobic, biocompatibility, and targeting moieties to be added by chemical reaction. Starting with a commercially sourced poly(isobutylene-alt-maleic anhydride), amine terminated hydrophobic and hydrophilic branches can be added simultaneously via spontaneous reaction with the maleic anhydride29. When mixed with the QDs, the polymer wraps around the particle, with the hydrophobic side chains interacting with the QD and intercalating with the pre-existing TOPO coatings, while the biocompatible and targeting side chains interact with aqueous solvent, thereby solublizing the particles. This system allows for the addition of a wide variety of amine terminated side-chains that can be used to tailor the surface properties of the dot, or the interaction between the particle and amphiphilic polymer. Each reaction between a maleic anhydride species and an amine species makes a COOH group available on the backbone, to which other molecules can be attached. Additionally, a diamine can be used to cross-link the polymer coating after it associates with the particle. Alternatively, a commercially sourced poly(maleic

- 243 -

anhydride –alt-1-octadecene) can be used as a starting material, which has hydrophobic side chains already present30.

In both cases, the amine precursors for the hydrophilic, biocompatible, or targeting side chains must be compatible with an organic solvent such as THF or chloroform because the reaction with the backbone must be performed in the absence of water; in water, the maleic anhydride quickly hydrolyzes to form two COOH groups. Yu et al.30 demonstrated the attachment of antibodies to the free COOH groups via EDC chemistry.

- 244 -

8.4.3 In vitro assays

As indicated at the beginning of this section, narrow emission lines, and tunable emissions (A2) make quantum dots ideal for simultaneously labeling multiple targets. Thus, improvements in QD synthesis and functionalization (A1), have enabled many studies demonstrating the labeling of multiple targets.

Roullier et al.25 demonstrated the simultaneous two colour labeling of two interacting proteins using two different targeting strategies on a live cell. Among the variety of cellular imaging examples in literature, this example is interesting because it clearly demonstrates not only the successful labeling of two different surface proteins with distinct QDs (T2), but also demonstrates how this can be used to observe the interaction of the two proteins on the cells surface. Additionally, it demonstrates two rather different targeting strategies, neither of which involves antibodies.

The two proteins of interest, Infar1 and Infar2, are the subunits of an interferon receptor. Infar1 carried a decahistidine tag used for affinity purification, and was labeled by a single red QD that carried Tris-NTA functional groups attached to the particle via amphiphilic micelle encapsulation. Infar2 carried a biotin label attached by enzymatic reaction, and was labeled with a green QD functionalized with streptavidin. The micellular method used to functionalize the red QD is another example of a surface functionalization route found elsewhere in literature. The labeling of the two proteins and their interaction are schematically illustrated in Figure 8-9A.

- 245 -

Figure 8-9: Demonstration of labelling and tracking membrane proteins by two QDs with distinct fluorescence emissions and different targeting moieties.25 Reproduced with permission from Roullier, V. et al. High-Affinity Labeling and Tracking of Individual Histidine-Tagged Proteins in Live Cells Using Ni2+ Tris- nitrilotriacetic Acid Quantum Dot Conjugates. Nano Letters 9, 1228-1234 (2009). Copyright © 2009, American Chemical Society.

Many copies of both proteins were observed on the cell surface as seen in Figure 8-9B and C; however the x,y position of one copy of each protein was tracked. The paths of the red-QD-labeled Infar1 and the green-labeled-Infar2 were observed to coincide for a period of time (Figure 8-9D), during which, the diffusion coefficient of each probe was observed to simultaneously drop (Figure 8-9E). Upon dissociation, the diffusion coefficients simultaneously returned to baseline levels, and their paths ceased to coincide.

8.4.4 In vivo assays

Although in vitro studies can clearly provide insight into cellular processes, and in some cases disease states, in vivo studies can provide insight into bodily processes that are

- 246 -

inaccessible by in vitro studies, where a variety of bodily processes such as digestion and immune response, may hinder the delivery of QDs to their intended target. Additionally, as a general limitation of fluorescence probes, the absorption and scattering of a broad spectrum of visible light by tissue, proteins, and water, interferes with the excitation and detection of fluorescence markers with spectral features in the visible. This is largely the motivation behind the development of NIR emitting QDs (T1).

Interference can come from molecular sources as well, in the form of nonspecific binding. As a trivial example, nonspecific accumulation of QDs in organs such as the liver and spleen may result in the false-positive detection of a target. As will be addressed later, even the use of highly specific antibodies can lead to labeling of nontarget tissue, because diseased tissue may express the same surface markers as surrounding healthy tissue, just to a different extent. Thus, the detection of diseased states in that case, depends on detecting differential expression of proteins between normal and diseased tissue65. The problem of nonspecific labeling can be partially mitigated by multiplexed detection of a variety of surface markers (T2).

Nevertheless, there is great interest in developing in vivo diagnostics, motivated in part by the desire to combine noninvasive diagnostics with noninvasive therapy (T3), coined “theranostics”. Therapeutic applications are beyond the scope of this review, but there is a great body of literature available to interested parties131,132.

Diagnostics modalities tend to involve the use of QDs as imaging contrast agents, where most, but not all, make use of targeted quantum dots to label a region of interest, relying on fluorescence (and lack of fluorescence) to provide contrast. There are also many studies that involve characterizing the bodily distribution of QDs insofar as they depend on the properties of the QD functionalizaiton such as surface charge, type and length of hydrophilic ligands, and QD hydrodynamic radius. There are also some studies on the

- 247 -

toxicity of QDs and their in vivo fate, which are discussed in detail in the next section. Like many of the in vitro studies, many in vivo studies are aimed at detecting tumors, usually in the form of human cancer xenografts on mice17,65.

Imaging deep tissue is difficult because proteins and water scatter and absorb a large fraction of the visible spectrum, especially in the absorption band of most QDs. To overcome this, multiphoton excitation using longer wavelength lasers can be employed. Bateman et al. (2007)61 imaged deep skeletal muscle vasculature using 5000 Mw-PEG coated CdSe QDs using laser scanning multiphoton microscopy, which has the additional advantage of combining sections at different focal depths to produce three dimensional images. This example is related to T1 insofar as the use of a NIR source allows excitation of QD in deep tissue. Although they use QDs with emissions centered at 655 nm (red) this wavelength can still penetrate tissue more effectively than shorter wavelengths. As mentioned before, the low-scattering-and-absorption window of tissue lies roughly between 650 nm and 900 nm.

Although signals from QDs with 655 nm emission can still penetrate tissue, their absorption band occurs over a wavelength band that is highly scattered and absorbed by tissue. Using a NIR laser with a wavelength within the transmission window of tissue, a QD can be excited by a two-photon process. One NIR photon does not have enough energy to excite the QD, but if two photons arrive within the absorption cross-section of a QD simultaneously, they linearly combine to impart twice the energy of either photon alone. The multiphoton process can only occur when the photon density is high enough that the probability of two photons arriving in the absorption cross section of a QD is high enough; that is, in a tightly focused volume. As a consequence, fluorescence from other QDs outside of the focal plane is virtually eliminated. And, by scanning the plane of focus, images can be taken at different depths and combined to form a 3D image. To reduce background autofluorescence, they chose a wavelength where the autofluorescence signal was lowest (~ 900 nm).

- 248 -

- 249 -

Figure 8-10 shows the in vivo two-photon fluorescence image of deep skeletal muscle vasculature in a mouse hind leg at a depth of 150 to 200 µm. They observed that QDs may become trapped at bifurcations in microvasculature. Aggregates of QDs were observed being extravasated, presumably by the lymphatic system, possibly correlating with observations of QD accumulation in lymph nodes. In very fine vasculature, red blood cells can be observed as dark, non-fluorescent spots in a background of fluorescent plasma. Treating the mice with a lipopolysaccharide to simulate bacterial infection conditions led to the faster clearance of QDs form circulation.

In summary, there are many in vivo examples of the themes (T1 and T2) outlined in previous sections. Future areas of focus would follow from these themes. For example, by using a variety of size-tunable QDs in the NIR, multiplexed detection of multiple targeted NIR QDs could be realized. Additionally, combing these with multiphoton excitation will allow both excitation and collection of signals from deep tissue without the need for surgery.

To facilitate the development of multiplexed imaging, it would be beneficial to develop a library of markers and their relative expression levels in various tissue that could be used to detect diseased states. In addition, a library of targeting species against these markers would also be required.

8.4.5 Toxicity

As in vivo applications mature, there is a growing need for toxicity studies, both in vitro and in vivo, which assess the effects of acute and chronic exposure to QDs in cells and in whole animal models. Many in vitro studies focus on the uptake and trafficking of various QDs, and consider the various modes of toxicity such as the generation of free radicals and the toxicity of Cd2+. Many in vivo studies focus on the biodistribution of

- 250 -

QDs, accumulation in various tissues and organs, as well as pathways to clearance of QDs with various compositions and coatings. In this section both in vitro and in vivo toxicity studies are addressed in separate sections. Several reviews on the topic, as well as important findings are highlighted with a focus on recent literature.

8.4.5.1 In vitro toxicity

In a recent review, Rzigalinski et al.133 present the state of knowledge on the toxicity of Cd-chalcogenide QDs. The review is extensive, focusing primarily on in vitro toxicity, but with a brief section on animal toxicity that will be revisited in the next section. For in vitro toxicity studies, they compare published toxicity studies of core-only, core/shell structures with ZnS shells, core/shell structures with organic coatings.

They identify 3 major sources of toxicity: Cd2+ toxicity, spontaneous free radical generation, and oxidative stress. In the first case, Cd2+ is released into the cytoplasm when the QD core is degraded in the cell by lysosomes, and leads to in apoptotic cell death. In the second case, excited state electrons can be transferred to molecular oxygen, producing singlet oxygen that can initiate the generation free radicals. In the last case, the presence of the QD can catalyze the generation of reactive oxygen species134, upsetting the reducing environment maintained by cellular homeostasis.

Coatings, and capping layers were found to reduce Cd2+ toxicity. They point out several sources that independently reported that core/shell QDs show reduced Cd2+ toxicity compared to core-only QDs. Presumably, the ZnS shell protects the core from oxidative degredation. However, they also refer to evidence that ZnS can be oxidized in the presence of air and water, resulting in SO2- radicals. Organic capping layers tended to increase the concentration of QDs required to induce toxicity. Although not covered in

135 that review, Z. Chen illustrated that CdS/Se QDs coated with SiO2 with OH surface

- 251 -

modifications subjected to buffer solutions of pH 7.4 and pH 4.8 (to simulate the pH conditions of blood and the renal tube respectively), Cd2+ was not found to be released.

In another good review on the topic, Stone et al.136 emphasize the need to develop rapid in vitro toxicity screens to test known in vivo toxicity mechanisms such as inflammation and oxidative stress. Stone et al. detail the many challenges in current in vitro assays, identifying evidence that the surface catalytic, optical, and electronic properties of the QDs can interfere with standard cellular toxicity assays which largely involve fluorescence, or color-change. Additionally, the large surface-area to volume ratio of QDs may cause the adsorption of molecular probes or enzymes used in some assays, leading to an underestimation of the probe quantity or enzymatic activity. Careful positive and negative controls are recommended to determine the extent of this interference. Rzigalinski et al.133, also shared similar concerns.

There are also other shortcomings in many in vitro and in vivo toxicity assays. As Chang et al.137 point out, many studies have compared the toxic effects of QDs in terms of the concentration of particles administered or known to be in the extracellular environment. However, they provide evidence that in many cases, it is the intracellular concentration is more relevant. The mechanism of Cd2+ toxicity involves the uptake and lysosomal degredation of QDs to release toxic Cd2+ into the cytoplasm. Therefore, considering the widely reported differences in cellular uptake efficiency of variously coated QDs, the intracellular concentration is more indicative of the potential toxic effects due to Cd2+. In fact, Chang et al.137 provide evidence that the reduced cytotoxicity of PEG coated QDs arises entirely from reduced uptake, and not from any protective feature of the layer itself. This understanding is especially relevant to interpreting findings that show PEG length determines the degree of toxicity. Figure 8-11 shows that when cytoplasamic concentrations of bare core-only QDs and core QDs with different length PEG coatings (750, and 6000 Mw) are consistent, cell death rates are also the same. This of course, does not account for QDs with a ZnS coating, or with SiO2 coatings.

- 252 -

In addition to the mechanisms of toxicity identified by Rzigalinski et al.133, there have been studies published since then that elucidate different mechanisms of toxicity. For example, Frohlich et al.138 tested the toxicity of carboxylate functionalized polystyrene QDs, both fluorescently labeled, and non fluorescent. PS nanoparticles are not known to catalyze either the generation of free radicals or reactive oxygen species - agents that contribute oxidative stress in cells. These polystyrene particles are reasonable models for general polymer-coated nanoparticles. They found that size of the particles was the determining factor for toxicity, with the smallest ones, (20 nm diameter) were the only ones that induce toxic effects on EAhy926 endothelial cells, showing evidence of membrane damage. Various particle sizes from 40 nm to 500 nm did not show appreciable cytotoxicity. Toxicity was assayed using several different standard cytotoxicity assays, and they all produced results consistent with one another. Adding

- 253 -

BSA at 1%, 5%, and 10% reduced toxicity of the 20 nm particles, but also increased the hydrodynamic radius of the particles to nearly 40 nm. Uptake efficiency also depended on particle size, with the 20nm particles being the most efficient, and the 200nm particles the least. In light of findings by Chang et al.137, a study is warranted to compare the toxicity of these differently sized PS particles under the same intracellular concentration.

Another mode of toxicity has recently been identified by Linse et al.139, whereby various nanoparticles including coated carbon nanotubes, inorganic nanoparticles, and polymer coated QDs catalyzed the formation of amyloid fibrils of ß2 microglobulin (ß2m). The amyloid fibrils of ß2m are found in the skeletal tissue of all patients undergoing hemodialysis, leading to chronic joint inflammation140. Other diseases linked to amyloidosis of other proteins include Alzheimer’s disease141, Creutzfeld-Jakob disease142, and type-II diabetes143.

The precise mechanism of how nanoparticles catalyzed amyloidosis cannot be determined from this study, however some insight was still obtained. The catalysis is characterized by a reduction in lag-time for the formation of amyloid fibrils, as measured by the onset of thioflavin-T fluorescence. Engineered particles of block-copolymers of different size, with different ratios of hydrophobic to hydrophilic block lengths reveal that smaller, more hydrophilic particles reduce the lag-time more than larger, or more hydrophobic particles. This suggests surface-area dependent catalysis, where the adsorption of protein to the particle surface enhances the local concentration of monomers, reducing diffusive barriers to fibril nucleation. This is supported by indirect observations that suggest the first monolayer of proteins on the surface of the particles are misfolded. These results are summaried in Figure 8-12.

- 254 -

Koeneman et al.144 investigated an in vitro assay aimed at determining mechanisms of QD absorption if they were accidentally ingested (for example, due to environmental contamination of water or food sources). Their assay utilizes epithelial cells grown in culture that form tight junctions with their neighbors, similar to epithelial layers in the digestive tract. They evaluate the effect if the QDs on the integrity of the epithelial layer by measuring the electrical resistance across the cell layer. Colloidal and aggregated QDs were applied to the apical surface of the cell layer. Colloidal QD suspensions at extracellular concentrations of 0.1 and 1.0 mg/L caused a decrease in electrical resistance,

- 255 -

indicating a breakage of the epithelial layer, and a live/dead assay indicates the presence of dead cells in the layer. Concentrations of 0.01 mg/L (extracellular) and below did not cause a drop in resistance over a 24 hr period. As a control, they tested the effect of the capping ligands alone and found no drop in resistance either, even at concentrations of 1000 mg/L. As another control, they exposed the layer to Cd2+ ion in PBS buffer, and did observe a drop in resistance across the epithelium. On the other hand, aggregated QDs did not show any toxicity. They also determined the extent to which QDs crossed the disrupted epithelial layer by way of mass balance: at a starting concentration of 1 mg/L, 34.1% of the QDs crossed the epithelial layer after 24 hrs, with higher starting concentrations yielding similar percentages.

8.4.5.2 In vivo toxicity

Generally, in vivo toxicity studies often involve quantifying the accumulation levels of QDs in various organs, and evaluating clearance or circulation times. Lacking in literature, as Rzigalinski et al.133 point out, are thorough in vivo studies to determine the potential toxicity induced by nonspecific localization of QD in non-target tissue. Further details can be read in their review.

135 Chen et al. reported that their OH functionalized SiO2 coated CdSe/S QDs had blood plasma half lives of 19.8 hours. 33.3% of the initial dose was cleared through feces, with the maximum amount cleared around 6-12 hours post injection. 23.8% was cleared by urine with a peak around 24-36 hrs hours post injection.

The urinary clearance of QDs was studied in more detail by Soo Choi et al.145. They argue that the toxicity of rapidly cleared QDs are much lower than that of accumulated QDs because the latter can invoke inflammation, and over time, are degraded, releasing toxic components. They found that with QDs, as with globular proteins, renal clearance

- 256 -

depends on hydrodynamic diameter, with QDs and proteins of ~5-6 nm diameter being more efficiently cleared than larger QDs or proteins. The larger dots instead, accumulated in the liver, lungs and spleen. The blood half-life of a 4.36 nm (hydrodynamic diameter) QD was 48 minutes, compared to 20 hours for 8.65 nm (hydrodynamic diameter).

Nonspecific protein adsorption can increase the hydrodynamic diameter of small, easily cleared QDs to beyond 6 nm, preventing their renal clearance. By comparing cationic, zwitterionic, and neutral coatings, on nanoparticles before and after incubation with serum proteins, they found that cationic coatings promoted the adsorption of proteins, which increased their hydrodynamic diameter. They also point out that although PEG is neutral, it was not possible to make PEG coated QDs with hydrodynamic diameters less than 10 nm, as QDs with short PEG chains tended to flocculate. This has broad implications to the field because the use of PEG coatings for in vivo applications is nearly ubiquitous.

The fate of QDs with different PEG lengths was evaluated by Daou et al.62. Using in vivo fluorescence imaging, they detected accumulation in liver, spleen, stomach, skin and bone. Using ex vivo organ imaging to verify the results, liver and skin were shown to have greater signals than autofluorescence. Also, liver accumulation showed PEG length dependence, 4000 Da PEG–QDs showed the most accumulation, while 12 kDa, and 22 kDa PEG-QDs have near-baseline levels. Accumulation in the liver decreased after 3 hours.

Clearly, a comprehensive, well controlled, systematic study of in vivo toxicity is required to assess the impact of various factors such as size, shape, surface charge, surface coating chemistry, and surface coating chain length in a comparable way. Because of the multitude of factors and conditions that determine the toxicity of QDs both in vitro an in

- 257 -

vivo, it is difficult to make comparisons between various studies performed in different laboratory settings under different conditions.

8.5 Conclusions and future perspectives:

Reviewed, were the size- and shape-tunable optical properties of both semiconductor quantum dots and plasmonic metal nanoparticles, and provided examples of surface chemical modification routes as well as examples of both in vitro and in vivo applications and toxicity. These findings point not only to trends in research, but also to several critical questions.

There are predominantly two growing trends in QD diagnostics. The first is multiplexed detection and imaging in vivo. The second is the use of near IR fluorescence probes whose light emissions better penetrate tissue. Naturally, the use of multiple NIR probes would be most advantageous in vivo, allowing multiple targets in deep tissue to be imaged noninvasively. NIR emitting probes with broad absorption bands have been made with CdTe cores41. Because absorption is still strong at low wavelengths, multiphoton excitation can be employed, enabling smaller, more precise excitation volumes.

Similarly, multiplexed NIR probes are desirable in metal nanoparticle diagnostics for the same reasons. SERS probes, with their narrow spectral features, are apt for multiplexed detection. Combining multiple Raman probes with metal nanorod and nanoshell architectures that have NIR surface plasmon resonances can lead to strong multiplexed SERS detection of targets.

- 258 -

One of the challenges that multiplexed detection is meant to address is that of interference from nonspecific labeling. Many in vivo studies of cancer targeting strategies make use of mice bearing human tumor xenografts and probes targeted to the tumor using antibodies against human tumor proteins that are not found elsewhere in the mouse. As Diagaradjane et al. (2008)65 point out, this situation is not representative of clinical conditions, in which the surrounding tissue is likely expressing the same surface proteins as the tumor. Therefore, in the case of the xenograft-bearing mouse, identifying the tumor involves the detection of the presence of a marker, whereas, in the case of a human subject with a tumor, identifying the tumor involves the more difficult task of detecting differential expression of a marker. Although studies of the former are valuable as a proof of principle, the clinical relevance is uncertain. A multiplexed detection scheme would allay this problem by labeling multiple surface markers on the tumor, and reporting differential expression levels for many surface markers. Different tissues can be identified with more certainty by their unique signature of relative surface marker expression levels.

A related challenge is to develop a library of cell surface markers that can be used to identify diseased tissue, and a corresponding library of either antibodies or peptides that bind specifically to these markers. A peptide-based targeting moiety might be advantageous over antibodies because it may be possible to alter the binding affinity between the peptide and its target receptor by engineering the peptide. This is valuable because it has been demonstrated that the cellular uptake of a nanoparticle depends on how tightly it is bound to the surface receptors of the cell. A moderately bound peptide- nanoparticle conjugate can be more efficiently internalized than tightly bound or loosely bound peptide-NP conjugates.

With eventual human applications in mind, there is a need for more systematic studies of in vivo toxicity. Many of the in vivo toxicity studies reviewed were parts of larger disconnected studies where determination of toxicity was not the primary focus. Consequently, findings from different studies are difficult to compare because different

- 259 -

animal models, different disease states, and different methods of measuring and quantifying the accumulation of particles in various tissues exist. Clearly, the field would benefit from standardization of certain toxicity tests, as well as from a truly systematic study of various nanoparticles and surface coatings.

8.6 Acknowledgements

The author would like to thank Christina MacLaughlin for her valuable contributions to this work, and Warren C. W. Chan for pointing out useful references.

8.7 References

1. West, J.L. & Halas, N.J. Engineered Nanomaterials For Biophotonics Applications: Improving Sensing, Imaging, and Therapeutics. Annu. Rev. Biomed. Eng. 5, 285- 292 (2003).

2. Rosi, N.L. & Mirkin, C.A. Nanostructures in Biodiagnostics. Chemical Reviews 105, 1547-1562 (2005).

3. Penn, S.G., He, L. & Natan, M.J. Nanoparticles for bioanalysis. Current Opinion in Chemical Biology 7, 609-615 (2003).

4. Biju, V., Itoh, T., Anas, A., Sujith, A. & Ishikawa, M. Semiconductor quantum dots and metal nanoparticles: syntheses, optical properties, and biological applications. Analytical and Bioanalytical Chemistry 391, 2469-2495 (2008).

5. Jain, K.K. Nanotechnology in clinical laboratory diagnostics. Clinica Chimica Acta 358, 37-54 (2005).

6. Murray, C.B., Norris, D.J. & Bawendi, M.G. Synthesis and characterization of nearly monodisperse CdE (E = sulfur, selenium, tellurium) semiconductor nanocrystallites. Journal of the American Chemical Society 115, 8706-8715 (1993).

- 260 -

7. Dabbousi, B.O. et al. (CdSe)ZnS Core−Shell Quantum Dots: Synthesis and Characterization of a Size Series of Highly Luminescent Nanocrystallites. The Journal of Physical Chemistry B 101, 9463-9475 (1997).

8. Trindade, T., O'Brien, P. & Pickett, N. Nanocrystalline semiconductors: Synthesis, properties, and perspectives. Chemistry of Materials 13, 3843-3858 (2001).

9. Hines, M.A. & Guyot-Sionnest, P. Synthesis and Characterization of Strongly Luminescing ZnS-Capped CdSe Nanocrystals. The Journal of Physical Chemistry 100, 468-471 (1996).

10. Rosenthal, S., McBride, J., Pennycook, S. & Feldman, L. Synthesis, surface studies, composition and structural characterization of CdSe, core/shell and biologically active nanocrystals. Surface Science Reports 62, 111-157 (2007).

11. Murphy, C.J. et al. Anisotropic Metal Nanoparticles: Synthesis, Assembly, and Optical Applications. J. Phys. Chem. B 109, 13857-13870 (2005).

12. Chen, S. & Kimura, K. Synthesis and Characterization of Carboxylate-Modified Gold Nanoparticle Powders Dispersible in Water. Langmuir 15, 1075-1082 (1999).

13. Chen, J. et al. Facile Synthesis of Gold−Silver Nanocages with Controllable Pores on the Surface. Journal of the American Chemical Society 128, 14776-14777 (2006).

14. Chan, W. & Nie, S. Quantum dot bioconjugates for ultrasensitive nonisotopic detection. SCIENCE 281, 2016-2018 (1998).

15. Kim, B.Y.S. et al. Biodegradable Quantum Dot Nanocomposites Enable Live Cell Labeling and Imaging of Cytoplasmic Targets. Nano Letters 8, 3887-3892 (2008).

16. Wu, X. et al. Immunofluorescent labeling of cancer marker Her2 and other cellular targets with semiconductor quantum dots. Nat Biotech 21, 41-46 (2002).

17. Gao, X., Cui, Y., Levenson, R.M., Chung, L.W.K. & Nie, S. In vivo cancer

- 261 -

targeting and imaging with semiconductor quantum dots. Nat Biotechnol 22, 969- 976 (2004).

18. Dubertret, B. et al. In Vivo Imaging of Quantum Dots Encapsulated in Phospholipid Micelles. Science 298, 1759-1762 (2002).

19. Tada, H., Higuchi, H., Wanatabe, T.M. & Ohuchi, N. In vivo Real-time Tracking of Single Quantum Dots Conjugated with Monoclonal Anti-HER2 Antibody in Tumors of Mice. Cancer Research 67, 1138-1144 (2007).

20. Gao, X. et al. In vivo molecular and cellular imaging with quantum dots. Current Opinion in Biotechnology 16, 63-72 (2005).

21. Akerman, M.E. Nanocrystal targeting in vivo. Proceedings of the National Academy of Sciences 99, 12617-12621 (2002).

22. Delehanty, J.B. et al. Self-Assembled Quantum Dot−Peptide Bioconjugates for Selective Intracellular Delivery. Bioconjugate Chem. 17, 920-927 (2006).

23. Jaiswal, J., Mattoussi, H., Mauro, J. & Simon, S. Long-term multiple color imaging of live cells using quantum dot bioconjugates. Nature Biotechnology 21, 47-51 (2003).

24. Prencipe, G. et al. PEG Branched Polymer for Functionalization of Nanomaterials with Ultralong Blood Circulation. Journal of the American Chemical Society 131, 4783-4787 (2009).

25. Roullier, V. et al. High-Affinity Labeling and Tracking of Individual Histidine- Tagged Proteins in Live Cells Using Ni2+ Tris-nitrilotriacetic Acid Quantum Dot Conjugates. Nano Letters 9, 1228-1234 (2009).

26. Mattoussi, H. et al. Self-assembly of CdSe-ZnS quantum dot bioconjugates using an engineered recombinant protein. Journal of the American Chemical Society 122, 12142-12150 (2000).

- 262 -

27. Bruchez, M., Moronne, M., Gin, P., Weiss, S. & Alivisatos, A. Semiconductor nanocrystals as fluorescent biological labels. Science 281, 2013-2016 (1998).

28. Dubois, F., Mahler, B., Dubertret, B., Doris, E. & Mioskowski, C. A versatile strategy for quantum dot ligand exchange. Journal of the American Chemical Society 129, 482-483 (2007).

29. Lin, C. et al. Design of an amphiphilic polymer for nanoparticle coating and functionalization. Small 4, 334-341 (2008).

30. Yu, W.W. et al. Forming Biocompatible and Nonaggregated Nanocrystals in Water Using Amphiphilic Polymers. Journal of the American Chemical Society 129, 2871- 2879 (2007).

31. Pellegrino, T. et al. Hydrophobic Nanocrystals Coated with an Amphiphilic Polymer Shell: A General Route to Water Soluble Nanocrystals. Nano Letters 4, 703-707 (2004).

32. Querner, C., Reiss, P., Bleuse, J. & Pron, A. Chelating Ligands for Nanocrystals' Surface Functionalization. Journal of the American Chemical Society 126, 11574- 11582 (2004).

33. Uyeda, H.T., Medintz, I.L., Jaiswal, J.K., Simon, S.M. & Mattoussi, H. Synthesis of Compact Multidentate Ligands to Prepare Stable Hydrophilic Quantum Dot Fluorophores. J. Am. Chem. Soc. 127, 3870-3878 (2005).

34. Link, S. & El-Sayed, M.A. Shape and size dependence of radiative, non-radiative and photothermal properties of gold nanocrystals. International Reviews in Physical Chemistry 19, 409 (2000).

35. Moskovits, M. Surface‐enhanced Raman spectroscopy: a brief retrospective. Journal of Raman Spectroscopy 36, 496 (2005).

36. Averitt, R.D., Westcott, S.L. & Halas, N.J. Linear optical properties of gold nanoshells. J. Opt. Soc. Am. B 16, 1824-1832 (1999).

- 263 -

37. Murray, C., Kagan, C. & Bawendi, M. Synthesis and characterization of monodisperse nanocrystals and close-packed nanocrystal assemblies. Annual Review of Materials Science 30, 545-610 (2000).

38. Qian, X. et al. In vivo tumor targeting and spectroscopic detection with surface- enhanced Raman nanoparticle tags. Nat Biotech 26, 83-90 (2008).

39. Gobin, A.M. et al. Near-Infrared Resonant Nanoshells for Combined Optical Imaging and Photothermal Cancer Therapy. Nano Letters 7, 1929-1934 (2007).

40. Sun, J., Zhu, M., Fu, K., Lewinski, N. & Drezek, R.A. Lead sulfide near-infrared quantum dot bioconjugates for targeted molecular imaging. Int J Nanomedicine 2, 235-240 (2007).

41. Kim, S. et al. Near-infrared fluorescent type II quantum dots for sentinel lymph node mapping. Nat. Biotechnol 22, 93-97 (2004).

42. Mahmood, U. & Weissleder, R. Near-Infrared Optical Imaging of Proteases in Cancer. Molecular Cancer Therapeutics 2, 489-496 (2003).

43. Stuart, D.A. et al. Glucose Sensing Using Near-Infrared Surface-Enhanced Raman Spectroscopy: Gold Surfaces, 10-Day Stability, and Improved Accuracy. Analytical Chemistry 77, 4013-4019 (2005).

44. Han, M., Gao, X., Su, J.Z. & Nie, S. Quantum-dot-tagged microbeads for multiplexed optical coding of biomolecules. Nat Biotech 19, 631-635 (2001).

45. DaCosta, R. et al. Multiplexed quantum dot fluorescence Imaging by in vivo targeting of epidermal growth factor receptor (EGFr) and villin expression in Barrett's Esophagus (BE) surgically induced by duodenoesophageal reflux in rats. Gastroenterology 128, A51-A51 (2005).

46. Lagerholm, B.C. et al. Multicolor Coding of Cells with Cationic Peptide Coated Quantum Dots. Nano Lett. 4, 2019-2022 (2004).

- 264 -

47. Stroh, M. et al. Quantum dots spectrally distinguish multiple species within the tumor milieu in vivo. Nat Med 11, 678-682 (2005).

48. Kobayashi, H. et al. Simultaneous Multicolor Imaging of Five Different Lymphatic Basins Using Quantum Dots. Nano Lett. 7, 1711-1716 (2007).

49. Hu, R. et al. Metallic Nanostructures as Localized Plasmon Resonance Enhanced Scattering Probes for Multiplex Dark-Field Targeted Imaging of Cancer Cells. The Journal of Physical Chemistry C 113, 2676-2684 (2009).

50. Yu, C., Nakshatri, H. & Irudayaraj, J. Identity Profiling of Cell Surface Markers by Multiplex Gold Nanorod Probes. Nano Letters 7, 2300-2306 (2007).

51. Keren, S. et al. Noninvasive molecular imaging of small living subjects using Raman spectroscopy. Proceedings of the National Academy of Sciences 105, 5844- 5849 (2008).

52. Lowery, A.R., Gobin, A.M., Day, E.S., Halas, N.J. & West, J.L. Immunonanoshells for targeted photothermal ablation of tumor cells. Int J Nanomedicine 1, 149-154 (2006).

53. Loo, C., Lowery, A., Halas, N., West, J. & Drezek, R. Immunotargeted Nanoshells for Integrated Cancer Imaging and Therapy. Nano Letters 5, 709-711 (2005).

54. Dickerson, E.B. et al. Gold nanorod assisted near-infrared plasmonic photothermal therapy (PPTT) of squamous cell carcinoma in mice. Cancer Lett 269, 57-66 (2008).

55. Elsayed, I., Huang, X. & Elsayed, M. Selective laser photo-thermal therapy of epithelial carcinoma using anti-EGFR antibody conjugated gold nanoparticles. Cancer Letters 239, 129-135 (2006).

56. Chen, J. et al. Immuno Gold Nanocages with Tailored Optical Properties for Targeted Photothermal Destruction of Cancer Cells. Nano Letters 7, 1318-1322 (2007).

- 265 -

57. Huang, X., El-Sayed, I.H., Qian, W. & El-Sayed, M.A. Cancer Cell Imaging and Photothermal Therapy in the Near-Infrared Region by Using Gold Nanorods. Journal of the American Chemical Society 128, 2115-2120 (2006).

58. Jayagopal, A., Russ, P. & Haselton, F. Surface engineering of quantum dots for in vivo vascular Imaging. Bioconjugate Chemistry 18, 1424-1433 (2007).

59. Loo, C. et al. Gold nanoshell bioconjugates for molecular imaging in living cells. Opt. Lett. 30, 1012-1014 (2005).

60. Lee, S. et al. Biological Imaging of HEK293 Cells Expressing PLCγ1 Using Surface-Enhanced Raman Microscopy. Analytical Chemistry 79, 916-922 (2007).

61. Bateman, R.M., Hodgson, K.C., Kohli, K., Knight, D. & Walley, K.R. Endotoxemia increases the clearance of mPEGylated 5000-MW quantum dots as revealed by multiphoton microvascular imaging. J. Biomed. Opt. 12, 064005-8 (2007).

62. Daou, T.J., Li, L., Reiss, P., Josserand, V. & Texier, I. Effect of poly(ethylene glycol) length on the in vivo behavior of coated quantum dots. Langmuir 25, 3040- 3044 (2009).

63. Cho, W. et al. Acute toxicity and pharmacokinetics of 13 nm-sized PEG-coated gold nanoparticles. Toxicology and Applied Pharmacology 236, 16-24 (2009).

64. Derfus, A., Chan, W. & Bhatia, S. Intracellular Delivery of Quantum Dots for Live Cell Labeling and Organelle Tracking. Adv. Mater. 16, 961-966 (2004).

65. Diagaradjane, P. et al. Imaging Epidermal Growth Factor Receptor Expression In vivo: Pharmacokinetic and Biodistribution Characterization of a Bioconjugated Quantum Dot Nanoprobe. Clinical Cancer Research 14, 731-741 (2008).

66. Aaron, J. et al. Plasmon resonance coupling of metal nanoparticles for molecular imaging of carcinogenesis in vivo. J. Biomed. Opt. 12, 034007-11 (2007).

67. Nie, S., Xing, Y., Kim, G.J. & Simons, J.W. Nanotechnology Applications in

- 266 -

Cancer. Annu. Rev. Biomed. Eng. 9, 257-288 (2007).

68. Al-Jamal, W.T. Tumor Targeting of Functionalized Quantum Dot−Liposome Hybrids by Intravenous Administration. Molecular Pharmaceutics 6, 530 (2009).

69. Hashizume, H. et al. Openings between Defective Endothelial Cells Explain Tumor Vessel Leakiness. Am J Pathol 156, 1363-1380 (2000).

70. Zhang, C., Yeh, H., Kuroki, M.T. & Wang, T. Single-quantum-dot-based DNA nanosensor. Nat Mater 4, 826-831 (2005).

71. Mitchell, G.P., Mirkin, C.A. & Letsinger, R.L. Programmed Assembly of DNA Functionalized Quantum Dots. Journal of the American Chemical Society 121, 8122-8123 (1999).

72. Skrabalak, S.E. et al. Gold Nanocages for Biomedical Applications. Adv. Mater. Weinheim 19, 3177-3184 (2007).

73. Huang, X., El-Sayed, I.H., Qian, W. & El-Sayed, M.A. Cancer Cells Assemble and Align Gold Nanorods Conjugated to Antibodies to Produce Highly Enhanced, Sharp, and Polarized Surface Raman Spectra: A Potential Cancer Diagnostic Marker. Nano Letters 7, 1591-1597 (2007).

74. Oyelere, A.K., Chen, P.C., Huang, X., El-Sayed, I.H. & El-Sayed, M.A. Peptide- Conjugated Gold Nanorods for Nuclear Targeting. Bioconjugate Chemistry 18, 1490-1497 (2007).

75. Lee, S. et al. Surface-enhanced Raman scattering imaging of HER2 cancer markers overexpressed in single MCF7 cells using antibody conjugated hollow gold nanospheres. Biosensors and Bioelectronics 24, 2260-2263 (2009).

76. Au, L. et al. A Quantitative Study on the Photothermal Effect of Immuno Gold Nanocages Targeted to Breast Cancer Cells. ACS Nano 2, 1645-1652 (2008).

77. Kumar, S., Aaron, J. & Sokolov, K. Directional conjugation of antibodies to

- 267 -

nanoparticles for synthesis of multiplexed optical contrast agents with both delivery and targeting moieties. Nat. Protocols 3, 314-320 (2008).

78. Nikoobakht, B. & El-Sayed, M.A. Preparation and Growth Mechanism of Gold Nanorods (NRs) Using Seed-Mediated Growth Method. Chemistry of Materials 15, 1957-1962 (2003).

79. Weissleder, R. A clearer vision for in vivo imaging. Nat Biotech 19, 316-317 (2001).

80. Lim, D., Kim, I. & Nam, J. DNA-embedded Au/Ag core-shell nanoparticles. Chem. Commun. 5312-5314 (2008).

81. Zhang, X., Young, M.A., Lyandres, O. & Van Duyne, R.P. Rapid Detection of an Anthrax Biomarker by Surface-Enhanced Raman Spectroscopy. Journal of the American Chemical Society 127, 4484-4489 (2005).

82. Haes, A.J., Chang, L., Klein, W.L. & Van Duyne, R.P. Detection of a biomarker for Alzheimer's disease from synthetic and clinical samples using a nanoscale optical biosensor. J. Am. Chem. Soc 127, 2264-2271 (2005).

83. Shah, N.C., Lyandres, O., Walsh, Glucksberg, M.R. & Van Duyne, R.P. Lactate and Sequential Lactate−Glucose Sensing Using Surface-Enhanced Raman Spectroscopy. Analytical Chemistry 79, 6927-6932 (2007).

84. Yu, C., Nakshatri, H. & Irudayaraj, J. Identity Profiling of Cell Surface Markers by Multiplex Gold Nanorod Probes. Nano Letters 7, 2300-2306 (2007).

85. Kneipp, J., Kneipp, H., Rice, W.L. & Kneipp, K. Optical Probes for Biological Applications Based on Surface-Enhanced Raman Scattering from Indocyanine Green on Gold Nanoparticles. Analytical Chemistry 77, 2381-2385 (2005).

86. Kneipp, J., Kneipp, H., Rajadurai, A., Redmond, R.W. & Kneipp, K. Optical probing and imaging of live cells using SERS labels. Journal of Raman Spectroscopy 40, 1-5 (2009).

- 268 -

87. Cang, H. et al. Gold nanocages as contrast agents for spectroscopic optical coherence tomography. Optics Letters 30, 3048-3050 (2005).

88. Chen, J. et al. Gold Nanocages: Bioconjugation and Their Potential Use as Optical Imaging Contrast Agents. Nano Letters 5, 473-477 (2005).

89. Shah, J. et al. Photoacoustic imaging and temperature measurement for photothermal cancer therapy. J. Biomed. Opt. 13, 034024-9 (2008).

90. Yang, X., Skrabalak, S.E., Li, Z., Xia, Y. & Wang, L.V. Photoacoustic Tomography of a Rat Cerebral Cortex in vivo with Au Nanocages as an Optical Contrast Agent. Nano Letters 7, 3798-3802 (2007).

91. Mallidi, S. et al. Multiwavelength Photoacoustic Imaging and Plasmon Resonance Coupling of Gold Nanoparticles for Selective Detection of Cancer. Nano Lett. 9, 2825-2831 (2009).

92. Choi, M. et al. A Cellular Trojan Horse for Delivery of Therapeutic Nanoparticles into Tumors. Nano Letters 7, 3759-3765 (2007).

93. Lal, S., Clare, S.E. & Halas, N.J. Nanoshell-Enabled Photothermal Cancer Therapy: Impending Clinical Impact. Accounts of Chemical Research 41, 1842-1851 (2008).

94. Huang, X., Jain, P., El-Sayed, I. & El-Sayed, M. Plasmonic photothermal therapy (PPTT) using gold nanoparticles. Lasers in Medical Science 23, 217-228 (2008).

95. Barhoumi, A., Huschka, R., Bardhan, R., Knight, M.W. & Halas, N.J. Light-induced release of DNA from plasmon-resonant nanoparticles: towards light-controlled gene therapy. Chem Phys Lett 482, 171-179 (2009).

96. Kelly, K.L., Coronado, E., Zhao, L.L. & Schatz, G.C. The Optical Properties of Metal Nanoparticles: The Influence of Size, Shape, and Dielectric Environment. The Journal of Physical Chemistry B 107, 668-677 (2003).

97. Link, S. & El-Sayed, M.A. Optical Properties and Ultrafast Dynamics of Metallic

- 269 -

Nanocrystals. Annu. Rev. Phys. Chem. 54, 331-366 (2003).

98. Raether, H. Surface plasmons on smooth and rough surfaces and on gratings. 111, (Springer-Verlag: 1988).

99. Jain, P.K., Huang, X., El-Sayed, I.H. & El-Sayed, M.A. Noble Metals on the Nanoscale: Optical and Photothermal Properties and Some Applications in Imaging, Sensing, Biology, and Medicine. Accounts of Chemical Research 41, 1578-1586 (2008).

100. Jain, P.K., Lee, K.S., El-Sayed, I.H. & El-Sayed, M.A. Calculated Absorption and Scattering Properties of Gold Nanoparticles of Different Size, Shape, and Composition: Applications in Biological Imaging and Biomedicine. The Journal of Physical Chemistry B 110, 7238-7248 (2006).

101. Schultz, S., Smith, D.R., Mock, J.J. & Schultz, D.A. Single-target molecule detection with nonbleaching multicolor optical immunolabels. Proceedings of the National Academy of Sciences of the United States of America 97, 996-1001 (2000).

102. Oldenburg, S.J., Averitt, R.D., Westcott, S.L. & Halas, N.J. Nanoengineering of optical resonances. Chemical Physics Letters 288, 243-247 (1998).

103. Lu, X. et al. Mechanistic Studies on the Galvanic Replacement Reaction between Multiply Twinned Particles of Ag and HAuCl4 in an Organic Medium. Journal of the American Chemical Society 129, 1733-1742 (2007).

104. Kneipp, K., Kneipp, H., Itzkan, I., Dasari, R.R. & Feld, M.S. Ultrasensitive Chemical Analysis by Raman Spectroscopy. Chemical Reviews 99, 2957-2976 (1999).

105. Campion, A. Surface-enhanced Raman scattering. Chemical Society Reviews 27, 250 (1998).

106. Nie, S. & Emory, S.R. Probing Single Molecules and Single Nanoparticles by Surface-Enhanced Raman Scattering. Science 275, 1102-1106 (1997).

- 270 -

107. Moskovits, M. Surface roughness and the enhanced intensity of Raman scattering by molecules adsorbed on metals. J. Chem. Phys. 69, 4159-4161 (1978).

108. Zhu, Z., Zhu, T. & Liu, Z. Raman scattering enhancement contributed from individual gold nanoparticles and interparticle coupling. Nanotechnology 15, 357- 364 (2004).

109. Xu, H., Aizpurua, J., Kall, M. & Apell, P. Electromagnetic contributions to single- molecule sensitivity in surface-enhanced Raman scattering. Phys. Rev. E 62, 4318 (2000).

110. Lutz, B. et al. Raman Nanoparticle Probes for Antibody-based Protein Detection in Tissues. J. Histochem. Cytochem. 56, 371-379 (2008).

111. Lee, P.C. & Meisel, D. Adsorption and surface-enhanced Raman of dyes on silver and gold sols. The Journal of Physical Chemistry 86, 3391-3395 (1982).

112. Nguyen, C.T. Towards surface enhanced raman scattering (SERS) nanoparticles for detection of chronic lymphocytic leukemia cells. (2009).

113. Chithrani, B.D., Ghazani, A.A. & Chan, W.C.W. Determining the Size and Shape Dependence of Gold Nanoparticle Uptake into Mammalian Cells. Nano Letters 6, 662-668 (2006).

114. Balogh, L. et al. Significant effect of size on the in vivo biodistribution of gold composite nanodevices in mouse tumor models. Nanomedicine: Nanotechnology, Biology and Medicine 3, 281-296 (2007).

115. Fischer, H.C. & Chan, W.C. Nanotoxicity: the growing need for in vivo study. Current Opinion in Biotechnology 18, 565-571 (2007).

116. Connor, E.E., Mwamuka, J., Gole, A., Murphy, C.J. & Wyatt, M.D. Gold nanoparticles are taken up by human cells but do not cause acute cytotoxicity. Small 1, 325-327 (2005).

- 271 -

117. Dobrovolskaia, M.A. & McNeil, S.E. Immunological properties of engineered nanomaterials. Nat Nano 2, 469-478 (2007).

118. Huff, T.B., Hansen, M.N., Zhao, Y., Cheng, J. & Wei, A. Controlling the Cellular Uptake of Gold Nanorods. Langmuir 23, 1596-1599 (2007).

119. Hauck, T.S., Ghazani, A.A. & Chan, W.C.W. Assessing the effect of surface chemistry on gold nanorod uptake, toxicity, and gene expression in mammalian cells. Small 4, 153-159 (2008).

120. Goldenberg, N.M. & Steinberg, B.E. Surface Charge: A Key Determinant of Protein Localization and Function. Cancer Research 70, 1277-1280 (2010).

121. Bar-Ilan, O., Albrecht, R., Fako, V. & Furgeson, D. Toxicity Assessments of Multisized Gold and Silver Nanoparticles in Zebrafish Embryos. Small 5, 1897- 1910 (2009).

122. Brus, L. Electronic wave functions in semiconductor clusters: experiment and theory. The Journal of Physical Chemistry 90, 2555-2560 (1986).

123. Brus, L.E. Electron--electron and electron-hole interactions in small semiconductor crystallites: The size dependence of the lowest excited electronic state. J. Chem. Phys. 80, 4403-4409 (1984).

124. Alivisatos, A.P. Semiconductor Clusters, Nanocrystals, and Quantum Dots. Science 271, 933-937 (1996).

125. Wu, B.J. et al. Growth and characterization of II--VI blue light-emitting diodes using short period superlattices. Appl. Phys. Lett. 68, 379-381 (1996).

126. Wolkin, M.V., Jorne, J., Fauchet, P.M., Allan, G. & Delerue, C. Electronic States and Luminescence in Porous Silicon Quantum Dots: The Role of Oxygen. Phys. Rev. Lett. 82, 197 (1999).

127. Takagahara, T. & Takeda, K. Theory of the quantum confinement effect on excitons

- 272 -

in quantum dots of indirect-gap materials. Phys. Rev. B 46, 15578 (1992).

128. Medintz, I., Uyeda, H., Goldman, E. & Mattoussi, H. Quantum dot bioconjugates for imaging, labelling and sensing. Nature Materials 4, 435-446 (2005).

129. Parak, W., Pellegrino, T. & Plank, C. Labelling of cells with quantum dots. Nanotechnology 16, R9-R25 (2005).

130. Xing, Y., Xia, Z. & Rao, J. Semiconductor Quantum Dots for Biosensing and In Vivo Imaging. IEEE TRANSACTIONS ON NANOBIOSCIENCE 8, 4-12 (2009).

131. Portney, N. & Ozkan, M. Nano-oncology: drug delivery, imaging, and sensing. Analytical and Bioanalytical Chemistry 384, 620-630 (2006).

132. Bakalova, R. et al. Quantum dot anti-CD conjugates: Are they potential photosensitizers or potentiators of classical photosensitizing agents in photodynamic therapy of cancer? Nano Letters 4, 1567-1573 (2004).

133. Rzigalinski, B. & Strobl, J. Cadmium-containing nanoparticles: Perspectives on pharmacology and toxicology of quantum dots. Toxicology and Applied Pharmacology 238, 280-288 (2009).

134. Lovric, J., Cho, S.J., Winnik, F.M. & Maysinger, D. Unmodified Cadmium Telluride Quantum Dots Induce Reactive Oxygen Species Formation Leading to Multiple Organelle Damage and Cell Death. Chemistry & Biology 12, 1227-1234 (2005).

135. Chen, Z. Bio-distribution and metabolic paths of silica coated CdSeS quantum dots. Toxicology and Applied Pharmacology 230, 371 (2008).

136. Stone, V., Johnston, H. & Schins, R. Development of in vitro systems for nanotoxicology: methodological considerations. Critical Reviews in Toxicology 39, 613-626 (2009).

137. Chang, E., Thekkek, N., Yu, W.W., Colvin, V.L. & Drezek, R. Evaluation of

- 273 -

quantum dot cytotoxicity based on intracellular uptake. Small 2, 1412-1417 (2006).

138. Frohlich, E. et al. Cytotoxicity of nanoparticles independent from oxidative stress. Journal of Toxicological Sciences 34, 363-375 (2009).

139. Linse, S. et al. Nucleation of protein fibrillation by nanoparticles. Proceedings of the National Academy of Sciences 104, 8691-8696 (2007).

140. Gejyo, F. A new form of amyloid protein associated with chronic hemodialysis was identified as β2-microglobulin. Biochemical and Biophysical Research Communications 129, 706 (1985).

141. Hardy, J. & Selkoe, D.J. The Amyloid Hypothesis of Alzheimer's Disease: Progress and Problems on the Road to Therapeutics. Science 297, 353-356 (2002).

142. Budka, H. et al. Neuropathological Diagnostic Criteria for Creutzfeldt-Jakob Disease (CJD) and Other Human Spongiform Encephalopathies (Prion Diseases). Brain Pathology 5, 459-466 (1995).

143. Lorenzo, A., Razzaboni, B., Weir, G.C. & Yankner, B.A. Pancreatic islet cell toxicity of amylin associated with type-2 diabetes mellitus. Nature 368, 756-760 (1994).

144. Koeneman, B. et al. Experimental approach for an in vitro toxicity assay with nonaggregated quantum dots. Toxicology in Vitro 23, 955-962 (2009).

145. Soo Choi, H. et al. Renal clearance of quantum dots. Nat Biotechnol 25, 1165-1170 (2007).

- 274 -

Chapter 9: Summary of Part II and Future Perspectives

9.1 Liposomes and nanoparticle functionalization

The application of various nanoparticles in medical and biodiagnostics has become the focus of intense research efforts. Often competing directly with trusted and mature molecular diagnostic techniques ensures the field is deep with challenges. Perhaps one of the most important advantages that semiconductor and metal nanoparticles have over molecular probes in optical or photonic-based biodiagnostics is that their optical properties are more strongly determined by geometric parameters such as size and shape, than by specific chemical properties or functional groups, allowing the optical and chemical properties of the particles to be engineered more independently.

This means that innovations in surface chemistry such as targeting to specific cells or proteins of interest, or adaptations to endure hostile conditions, can be applied modularly to particle based probes with various optical properties. This ultimately has many advantages for multiplexed detection, for example: allowing quantum dots of various colours to be targeted to different tissues using fundamentally the same method. In contrast, the chemical modification of molecular dyes (i.e. their covalent attachment to a targeting species) are often limited to specific families of dyes that have the same functional group, and because the functional group likely affects the optical properties of the dye, there may only be a limited selection of dyes that can be modified by a given method.

This independence of optical and chemical properties works in two dimensions: while probes with different optical properties can be modified with the same surface chemistry,

- 275 -

likewise a probe of fixed optical properties can be modified with various surface chemistries. Arguably, modularity of surface functionalizations can extend in a third dimension, which crosses the barriers between platforms, for example: enabling surface modifications developed for quantum dots to be used interchangeably between gold, polymer, and magnetic nanoparticles. An example might be the amphiphilic copolymer surface modification discussed in the introduction/literature review.

In the case of plasmonic metal nanoparticles, this modularity allows metal or bimetallic nanospheres, nanorods, nanoshells, and nanocages to be modified in a variety of ways while maintaining their plasmonic properties. Likewise, nanospheres or nanorods of various geometric parameters and therefore various LSPR resonances can also be functionalized in a variety of ways. However, for SERS particles specifically, some degree of modularity is lost by the reliance of dye molecules to provide spectrally distinct probes. This can be compensated for by having a modular surface functionalization platform that allows for various dyes species, targeting moieties, and surface chemical species to be incorporated or interchanged, while leaving the synthetic route mostly unchanged.

Indeed, the demands on surface modification platforms are many, and varied, Surface functionalizations are expected to:

• protect the particle from the widely varying and sometimes hostile environments found in biology, while preserving the optical properties of the probe.

• Control binding to specific targets, and resist binding to nonspecific targets.

- 276 -

• Control the fate of the particle by influencing circulation time, secretion, extravasation, and immune recognition and response (and consequently toxicity).

For SERS particles, whose main advantage over other nanoparticle technologies is the promise of simultaneous multiplexed detection of several surface markers, there is the additional requirement of incorporating a variety of dye species. Clearly a highly modular functionalization platform is required to address these requirements.

Lipid coatings may represent such a functionalization platform. It has been demonstrated in the liposome drug delivery field that liposomes can be modified to protect and deliver cargo for in vivo use. These innovations can potentially be applied to liposome-coated SERS particles provided that multiple dyes can be incorporated, and that these coatings can be applied to nanoparticles. The work presented in this thesis demonstrates the incorporation of multiple dyes by multiple routes that all rely on the same basic process of sonicating the particles with lipids. The main results are summarized in the next section.

9.2 Summary of results

The results presented in Chapter 8 demonstrate not only the encapsulation of gold nanoparticles with a lipid bilayer, but also the incorporation of three distinct dye species, as well as the flexibility of having three avenues by which to incorporate dyes. TEM was used to directly visualize the lipid bilayer surrounding the gold nanoparticle. The thickness ranged from 4-6nm, which is within the acceptable range for the thickness of a bilayer. Furthermore, DLS and UV-Vis absorbance were used to characterize the lipid layer.

- 277 -

The three dyes, malachite green, rhodamine-lissamine, and tryptophan were each conjugated to the particles by a different method. The methods involved associating the dye with the particle prior to encapsulation, using lipid-bound dye, and associating the dye with the lipid prior to encapsulation. The association of each of the dyes with the particles was verified by collecting the SERS spectrum of each dye/particle conjugate.

Interestingly, the charge on the rhodamine-lipid species likely caused it to partition into the outerleaflet where SERS enhancement is poor. The addition of Ca2+ was employed to reverse the repulsion between the citrate on the surface and the charged dye. This improved the SERS spectrum dramatically, and resulted in spectra that resembled that of Rhodamine. These particles also exhibited a second distinct peak in the UV-Vis absorption spectrum. This second peak may signifying the formation of a distinct population of fairly homogeneous aggregates. Alternatively the second peak could also have resulted from the absorption of light by the rhodamine dye.

Longevity of the SERS signal was also demonstrated, with little to no change in the MGITC SERS over the course of 25 days. Surprisingly, the SERS spectrum of the Rhodamine aggregates also remained unchanged over the course of the 7 days tested. The lipid-coated particles were found to resist flocculation in the presence of acid and calcium, conditions under which the citrate coated particles precipitated. The lipid coated particles, including the aggregates were stable over several centrifugation steps, and were readily resuspended.

9.3 Future work

An important topic of future investigation would be the interplay between particles with different surface charges and dye or lipids with various charges. It is hypothesized that the negatively charged rhodamine-labeled lipids will associate readily with the particles

- 278 -

that have positive capping groups such as CTAB. Meanwhile lipids conjugated to positively charged dyes would form readily around the citrate coated particles.

Another clear route for future investigations is to incorporate targeting moieties and demonstrate their preferential binding to specific targets such as cancer cells. There exists many means by which to incorporate targeting peptides, or antibodies, since there are many commercially available head-group modified lipids with reactive linkers. To borrow a lesson from liposomal drug delivery, it has been reported that antibody labeled liposomes are rapidly sequestered in vivo by the reticuloendothelial system, which has prompted the development of so-called “stealth liposomes” that evade immune recognition. These stealth liposomes feature long PEG chains anchored to lipids in the bilayers, and antibodies anchored to the ends of a fraction of these PEG chains. This construct has been reported to have improved circulation time compared to antibody-only labeled liposomes. For some in vitro applications, it may suffice to make traditional antibody conjuagated lipid encapsulated SERS particles. However, it appears inevitable that some “stealth” liposome-coated-SERS particles will be required for in vivo experiment.

Additionally, since drug delivery liposomes are designed to deliver cargo into the cell, these lipid-coated particles may likewise be able to deliver nanoparticles into the cytoplasm. There are several membrane fusion proteins that are employed by envelope viruses to enter the cytoplasm: by fusing their lipid envelope to the cytoplasm, they can enter the cytoplasm and bypass the endocytic pathway where they might be subject to harsh conditions in late endosomes. Likewise, employing the same strategy, nanoparticles can conceivably be delivered to cells without being sequestered into the endocytic pathway. Targeting ligands bound directly to the particle could be used to direct them to subcellular targets.

- 279 -

It may also be possible to incorporate some natural cell membranes and membrane proteins, such as those employed in Part I. This could possibly play a role in an in vitro colorimetric assay where cell membrane extracts are used to coat particles, and the presence of diseased cells in the original population of cells is detected by introduction of antibodies against specific surface markers for the disease. If the antigen is present in the membrane extract, the presence of antibodies should cause the aggregation of the lipid- coated particles, which will cause a detectable shift in the UV-Vis spectrum.

Another avenue of exploration might be the application of these techniques to functionalizing other nanoparticles from plasmonic particles of various shape, to quantum dots and polymer particles as well. Further more, the flexibility of this lipid-coating procedure towards the use of nearly any lipids affords many possibilities in future investigations of lipid-coated nanoparticles.

9.4 Final remarks

The variety of environmental conditions present in biological systems, and the complexity of regulatory systems and defenses in cells and higher organisms ensures that many challenges will continue to exist for in vivo and in vitro particle diagnostics. The modular nature of the lipid bilayer, and the ability to employ a gamut of commercially available modified lipids with minimal modification of the encapsulation procedure makes lipid coatings a promising and versatile surface functionalization platform to enable nanoparticles to overcome the challenges of biodiagnostics.

- 280 -

- 281 -