understanding how non-coalenvt interactions a ect

Understanding How Non-Covalent Interactions A�ect

Interfacial Biomolecular Dynamics

by

Aaron Christopher McUmber

B.S., University of Rochester, 2009

A thesis submitted to the

Faculty of the Graduate School of the

University of Colorado in partial ful�llment

of the requirements for the degree of

Doctor of Philosophy

Department of Chemical and Biological Engineering

2015

This thesis entitled:Understanding How Non-Covalent Interactions A�ect Interfacial Biomolecular Dynamics

written by Aaron Christopher McUmberhas been approved for the Department of Chemical and Biological Engineering

Alfred T. and Betty E. Look Professor Daniel K. Schwartz

Theodore W. Randolph

Date

The �nal copy of this thesis has been examined by the signatories, and we �nd that both the content andthe form meet acceptable presentation standards of scholarly work in the above mentioned discipline.

iii

McUmber, Aaron Christopher (Ph. D., Chemical and Biological Engineering)

Understanding How Non-Covalent Interactions A�ect Interfacial Biomolecular Dynamics

Thesis directed by Alfred T. and Betty E. Look Professor Daniel K. Schwartz

Biopolymers, such as proteins and nucleic acids, are omnipresent in modern applications. The need

to control interfacial molecular systems is becoming increasingly important in order to develop more sophis-

ticated biopolymer-based technologies. Non-covalent interactions such as electrostatic, van der Waals, and

hydrophobic interactions are integral in interfacial phenomena. These interactions dictate how molecules

adsorb, desorb, and di�use at interfaces. By understanding how these forces a�ect molecular dynamics,

we can better design biopolymer-based technologies. Interfacial adsorption and interaction mechanisms

are studied using polarized light microscopy and single-molecule total internal re�ection �uorescence mi-

croscopy (TIRFM). Polarized light microscopy allows for the detection of birefringence within liquid crystal

layers, corresponding to molecular orientation, while TIRFM allows detection of single-molecule adsorption,

desorption and di�usion events at an interface. Using these techniques, mechanisms for the formation of

surfactant-biopolymer complexes, electrostatically-driven protein adsorption, and protein layer formation

are identi�ed. From these results, single stranded DNA-surfactant complexes are found to increase the sur-

factant area per molecule leading to liquid crystal realignment. Electrostatic repulsion a�ected elementary

adsorption of protein to a charged interface without a�ecting either elementary desorption or interfacial dif-

fusion. Protein layer formation mechanisms were identi�ed by comparing dynamic signatures and applying

new analysis techniques to molecular trajectories. The development of surfactant-protein complexes creates

protective e�ects preventing interfacial protein gelation. The work done in this thesis led to higher-order

analysis of molecular trajectories. The new analysis techniques led to the development a new single-molecule

micro-rheological technique, providing an unprecedented level of mechanistic interpretation of developing

viscoelastic layers.

Dedication

To my sister for her patience and for keeping me focused on the important things.

To my father for his support and for pushing me the moment I needed it most.

To my mother for her love and for all the little things.

v

Acknowledgements

First I would like to my thesis advisors, Dan Schwartz and Ted Randolph, Dan for his guidance,

expert advice, and overwhelming patience and Ted for his unique perspective, sound advice, and constant

pool of ideas. Also thank you to the rest of my thesis committee: John Carpenter, Jen Cha, and Joel Kaar.

Their insight and suggestions were invaluable to this thesis.

I would also like to thank my funding, which helped me through this journey. The bulk of the work

was funded by the National Science Foundation (award No. CBET-1133871), the Liquid Crystal Materials

Research Center (NSF/MRSEC, Award No. DMR-820579) and the Colorado State Bioscience Proof-of-

Concept Grant (No. 09BGF13).

Thank you to the collaborators and colleagues whom I have had the pleasure to work with. I thank

Patrick Noonan for the work him and I completed together with DNA-surfactant complexes at the liquid

crystal interface presented in Chapter 2. He is a determined scientist and a good friend. I would like to

thank Alana with whom I worked with the 3M-Polysorbate study presented in Chapter 5. I would like to

thank Louise Stenstrup Holm for the collaboration we did on a side project for Pegylated -lysozyme. I would

also like to thank Nick Larson who painstakingly worked with me on purifying protein for a project that

never came to fruition. Nick also helped me �nish the protein layer formation work presented in Chapter 4.

His intense eagerness and incredible aptitude was infectious.

I would like to thank both the Schwartz lab and the Randolph lab past and present for their helpful

commentary. Thank you, Jon Monserud and Nathan Nelson, you two were both incredibly brilliant and

tremendously fun to spend my free time with. Thank you Kim Hasset and Carley Chisholm for the great

friendship and wonderful advice you have provided over the years.

vi

I would like to thank my friends who made life here in Colorado a blast. Thank you Alan Powers

and Jena Miller for the great runs we have had at Rocky Mountain National Park and almost every trail

in Boulder. Kevin McCabe and Emily Miller, thank you for their stout friendship and fantastic limoncello.

Jordan, Kelly, and Jer, thank you all for truly being some of best friends I have ever had.

Finally, I would like to thank my family. Without them I would not have seen this six year journey

to its end. They have supported me in my times of crisis and have encouraged me when the challenges were

greatest.

Contents

Chapter

1 Introduction 1

1.1 Non-Covalent Interactions Between Surfaces and Molecules . . . . . . . . . . . . . . . . . . . 2

1.1.1 Long-Range Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.2 Short-Range Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.1.3 DLVO Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.1.4 Hydrophobic E�ect and Surface Activity . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.1.5 Adsorption Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.2 Molecular Conformations in Solution vs Interface . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.3 Interfacial Molecule-Molecule Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.4 Single-Molecule and Ensemble-Averaging Techniques . . . . . . . . . . . . . . . . . . . . . . . 10

1.4.1 Single-Molecule Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.5 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.5.1 Liquid Crystal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.5.2 Silicone Oil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.6 Thesis Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.7 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Surfactant�DNA Interactions at the Liquid Crystal�Aqueous Interface 16

2.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

viii

2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.3.1 LC Film Preparation and Polyanion Addition . . . . . . . . . . . . . . . . . . . . . . . 20

2.3.2 OTES Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.3.3 Fluorescence Recovery after Photobleaching (FRAP) . . . . . . . . . . . . . . . . . . . 22

2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.4.1 ssDNA Adsorption and LC Reorientation . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.4.2 Cationic and Nonionic Surfactant Monolayers . . . . . . . . . . . . . . . . . . . . . . . 24

2.4.3 ssDNA Surface Coverage and Di�usion . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.4.4 PSS, PAA, and dsDNA Adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.4.5 Interfacial Hybridization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.7 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.8 Supplementary Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.8.1 dsDNA Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.8.2 Flow Cell Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Electrostatic Interactions In�uence Protein Adsorption � but not Desorption � at the Silica-Aqueous

Interface 35

3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42


3.6 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.7 Supporting Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

ix

3.7.1 Additional Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.7.2 Additional Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4 Molecular Trajectories Provide Signatures of Protein Clustering and Crowding at the Oil/Water

Interface 48

4.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.3 Materials & Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.4 Results & Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4.6 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

5 Surfactant E�ects on Particle Generation in Antibody Formulations in Pre-Filled Syringes 67

5.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68


5.3.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

5.3.2 Incubation of 3M Formulation with Polysorbate 20 (Above CMC) in PFS . . . . . . . 70

5.3.3 Agitation of 3M Formulations with Varying Surfactant:Protein Ratios in PFS . . . . . 71

5.3.4 Counting of Particles in Incubated 3M Formulations . . . . . . . . . . . . . . . . . . . 71

5.3.5 3M Labeling with Alexa Fluor 555 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

5.3.6 3M Molecule Tracking Using Total Internal Re�ection Fluorescence (TIRF) Microscopy 73

5.3.7 TIRF Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5.4.1 Particle Concentrations in 3M Formulations with 0.01% v/v Polysorbate 20 after In-

cubation in PFS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5.4.2 Particle Concentrations in 3M Formulations Containing Various Surfactant:Protein

Ratios After Agitation in PFS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

x

5.4.3 Interfacial Di�usion of Labeled 3M Molecules at the Silicone Oil-Water Interface in

Formulations with Varying Surfactant:Protein Molar Ratios . . . . . . . . . . . . . . . 78

5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

5.5.1 Particle Generation in Antibody Formulations in PFS . . . . . . . . . . . . . . . . . . 79

5.5.2 In�uence of Polysorbate 20 Concentration on Particle Generation in PFS . . . . . . . 80

5.5.3 In�uence of Polysorbate 20 Concentration on Gelation of 3M Molecules at the Silicone

Oil-Water Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

5.5.4 E�ects of Surfactants on Protein Gelation at Interfaces . . . . . . . . . . . . . . . . . 82

5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

6 Conclusions 85

Bibliography 87

List of Tables

Table

3.1 Population fractions and characteristic residence time �t from Figure 3.1 for each pH. . . . . 46

3.2 Population fractions and di�usion coe�cients �ts from Figure 3.2 for each pH. . . . . . . . . 46

5.1 3M concentrations and polysorbate 20 concentrations corresponding to the polysorbate 20:3M

molar ratios used in the formulations tested. The polysorbate 20 CMC is 0.007 % v/v (0.06

mM) [84]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

List of Figures

Figure

1.1 (a) LC molecule with ordinary and extraordinary axes, exhibiting refractive indices, no and ne,

respectively (b) light propagation through orthogonally crossed polarizers with a LC layer ex-

hibiting a tilted con�guration situated in between the polarizers (c) light propagation through

orthogonally crossed polarizers with a LC layer exhibiting a homeotropic con�guration situ-

ated in between the polarizers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.2 Molecular structure of PDMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.1 LC response to ssDNA adsorption and hybridization: (a) Polarized microscopy images of the

aqueous/LC interface laden with OTAB, (b) after subsequent adsorption of ssDNA, and (c)

after interfacial hybridization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2 Surfactants and polymers used in the experiments. From left to right: cationic surfactants

with decreasing surface activity (DODAB, OTAB, and DTAB), and a nonionic surfactant

(C12E4). The DNA analogs used included polyanions with or without hydrophobic sidegroup

moieties (PSS and PAA respectively). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.3 LC reorientation upon ssDNA adsorption at varying [OTAB] � Polarized light microscopy

images of the LC-aqueous interface at low [OTAB] (a-c) intermediate [OTAB] (d-f) and high

[OTAB] (g-i) before (a,d,g) and after ssDNA adsorption at 1min (b,e,h) and 15min (c,f,i). . . 24

2.4 ssDNA Di�usion: The di�usion coe�cient, measured via FRAP, of ssDNA at an OTAB laden

aqueous-LC interface with varying OTAB coverage. . . . . . . . . . . . . . . . . . . . . . . . . 25

xiii

2.5 ssDNA adsorption at an aqueous-LC interface: (a) Polarized light microscopy image of a

cationic surfactant laden aqueous-LC interface after ssDNA adsorption. (b) the same �eld of

view imaged with epi�uorescence microscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.6 dsDNA hybridization at an aqueous-LC interface: Polarized light microscopy (a,c) and epi�u-

orescence microscopy (b,d) images of an OTAB laden aqueous-LC interface after hybridization

using either �uorescently labeled target (a,b) or probe (c,d). . . . . . . . . . . . . . . . . . . 28

2.7 Schematic illustration of the mechanism for a LC reorientation upon ssDNA adsorption and

hybridization: ssDNA in bulk solution (a), adsorbed at the interface (zoomed in) (b), and

after interfacial hybridization (c). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.1 Adsorption rate coe�cients of BSA at FS surface as a function of pH in 10 mM CP (closed

circles) and 10 mM CP with 100 mM NaCl (open diamonds) obtained from single-molecule

adsorption observations made using TIRF. Error bars in plot represent the standard deviation

between three replicate experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.2 (a) Average surface residence times of BSA at the FS surface as a function of pH in 10

mM CP. (b) Characteristic residence times, τ , each of the three populations identi�ed by

analysis of cumulative surface residence time distributions (closed boxes, open diamonds, and

closed circles, respectively). (c) Population fractions associated with each of the characteristic

residence times shown in panel (b). The error bars represent the standard deviation between

three replicate experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.3 (a) Mean di�usion coe�cients, D, of BSA at FS a surface as a function of pH in 10 mM CP.

Error bar in both plots represent the standard deviation between three replicate experiments.

(b) Di�usion coe�cients, D, each of the four populations identi�ed by analysis of cumulative

squared displacement distributions (closed boxes, open diamonds, closed circles, and closed

diamonds, respectively). (c) Population fractions associated with each of the di�usion coef-

�cients shown in panel (b). The error bars represent the standard deviation between three

replicate experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

xiv

3.4 Cumulative residence time distributions of �uorescently labeled BSA on FS for 2.6 (red circles),

3.7 (yellow triangles), 4.7 (green squares), 5.7 (open diamonds), 7.4 (black triangles). The

curves represent the mean between three replicate experiments. Error bars represent 65%

con�dence expected from Poisson statistics. For clarity, the graphs have been o�set vertically. 45

3.5 Cumulative squared displacement distributions of �uorescently labeled BSA on FS for 2.6 (red

circles), 3.7 (yellow triangles), 4.7 (green squares), 5.7 (open diamonds), 7.4 (black triangles).

The curves represent the mean between three replicate experiments. Error bars represent 65%

con�dence expected from Poisson statistics. For clarity, the graphs have been o�set vertically. 47

3.6 Representative CD spectra of BSA in 10 mM CP normalized to CD spectra of 10 mM CP at

each respective pH. The black solid line represents BSA data captured at 2.6 pH, the dotted

line at 4.7 pH, and the dashed line at 7.4 pH. The data in black are measurements captured

at 20°C. The red solid line represents BSA data captured at 7.5 pH at 75°C. . . . . . . . . . . 47

4.1 Temporal evolution of the relative interfacial tension for BSA (�lled circles) and lysozyme

(open diamonds) at a bulk concentration of 70 nM in PBS where time is on a logarithmic

scale and the relative interfacial tension is de�ned as the instantaneous interfacial tension γ

divided by the initial value, γ0. Error bars indicate the standard deviation between three


4.2 Various properties of 70 nM BSA (�lled circles) and 70 nM lysozyme (open diamonds) plotted

versus the dimensionless time, Γ, a value calculated by adjusting for protein adsorption rate,

bulk concentration and molecular size. (a) Relative interfacial tension, (b) relative mean di�u-

sion coe�cient, (c) power law exponent, α, associated with the mean squared displacement vs.

time, and (d) the �rst non-trivial value, G(τ = 0.2s), of the velocity-velocity autocorrelation

function. Error bars from �gures (a) and (b) represent the standard deviation between three


xv

4.3 Population distribution of relative di�usion coe�cients observed for (a � c) lysozyme and (d �

f) BSA. (a) and (d) depict data obtained at D/ 〈Do〉 = 1.0; (b) and (e) depict D/ 〈Do〉 ' 0.4;

(c) and (f) depict late time distributions at D/ 〈Do〉 less than or equal to 0.2. . . . . . . . . . 60

4.4 Mean squared displacement plotted vs. time interval for (a) BSA and (b) lysozyme at selected

time bins as annotated in the legends. These time bins were chosen to represent aging times

(early, middle, and late) at which the respective protein layers exhibited similar behavior. . . 61

4.5 Velocity-velocity autocorrelation function vs. time interval for (a) BSA and (b) lysozyme

at selected time bins depicted in the legends. The time bins for (a) BSA are solid line

Γ = 8x10−4, dotted line Γ = 4x10−1, dashed line Γ = 7x10−1; and for (b) lysozyme are

solid line Γ = 4x10−4, dotted line Γ = 2, dashed line Γ = 5. . . . . . . . . . . . . . . . . . . . 64

5.1 Particle concentrations in 3M formulations with 0.01 % v/v polysorbate 20 and in bu�er

solutions with 0.01 % v/v polysorbate 20 agitated in PFS as a function of time. Open symbols

correspond to syringes incubated with no air bubble and closed symbols correspond to syringes

incubated with an air bubble. The particle concentrations in a bu�er solution (solid black

line) and in a 3M solution (dashed black line) with 0.01 % v/v polysorbate 20 that were not

incubated in syringes are also shown. The incubation conditions are as follows: (a) L-histidine

bu�er (no protein) in agitated, siliconized syringes, (b) 3M formulation in quiescent, siliconized

syringes, (c) 3M formulation in agitated, un-siliconized syringes, and (d) 3M formulation in

agitated, siliconized syringes. For comparison, the gray symbols in panel (d) correspond to a

3M formulation with no surfactant agitated in siliconized syringes with an air bubble [58]. . . 75

xvi

5.2 Particle concentrations measured in 3M formulations as a function of the polysorbate 20:3M

molar ratio in the formulation. Open symbols represent the particle concentrations in 3M

formulations that were not incubated. Closed symbols represent the particle concentrations

in 3M formulations that were agitated for 24 hours with an air bubble in siliconized syringes.

The color of each symbol corresponds to the polysorbate 20 concentration (% v/v) in the for-

mulation, as shown in the legend. At each polysorbate 20:3M ratio, the particle concentration

in a bu�er solution with the same polysorbate 20 concentration was subtracted from the par-

ticle concentration measured in the non-incubated 3M formulation and from that measured

in the agitated 3M formulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

5.3 Mean di�usion coe�cients (μm2/s) of trace amounts of labeled 3M at the silicone oil-water

interface as a function of the polysorbate 20:unlabeled 3M molar ratio. The open symbols

correspond to formulations with a bulk unlabeled 3M concentration of 1.0 mg/mL. The closed

symbols correspond to formulations with a bulk unlabeled 3M concentration of 7.6 mg/mL.

The solid line represents the mean di�usion coe�cient of interfacially-adsorbed labeled 3M

molecules in the absence of polysorbate 20 and without addition of any unlabeled 3M. . . . . 79

Chapter 1

Introduction

Biopolymers such as proteins and nucleic acids are omnipresent in modern applications. The need

to control molecular systems is becoming increasingly important in order to develop more sophisticated

biopolymer-based technologies.[147, 63] Protein molecules, for example, contain amino acid side-groups that

span a variety of hydrophobicities from large aromatic groups (e.g. tryptophan) to small hydrocarbon chains

(e.g. alanine) as well as a wide range of hydrophilicities containing polar uncharged (e.g. serine), acidic (e.g.

aspartic acid), or basic (e.g. lysine) side-groups. The complex chemistry of amino acid sequences lead to

the formation of secondary structures such as α-helices and β-sheets that, in turn, comprise the complex

protein structure, containing charged, polar, or hydrophobic surface patches.[106] These amphiphilic surface

patches, therefore inherently make protein molecules surface-active.[106, 117] Nucleic acid molecules, on the

other hand, contain both a polyanionic phosphate backbone and aromatic nucleobases capable of forming

hydrogen bonds. Though, this does not necessarily make nucleic acid molecules surface active, the di�erence

in chemical structure between the polyanionic backbone and the aromatic base pairs can lead situations

where nucleic acid molecules interact di�erently with hydrophobic or charged phases or molecules.[119, 132]

This chapter will discuss topics related to the types of non-covalent interactions molecules experience in bulk

solution and at interfaces, how these interactions a�ect interfaces, how interfacial chemistry a�ects molecular

conformation, how inter-molecular interactions compare at the interface and the bulk, and how interfaces

are studied using ensemble-averaging techniques and single-molecule techniques.

2

1.1 Non-Covalent Interactions Between Surfaces and Molecules

1.1.1 Long-Range Interactions

Charged interfaces are di�cult to avoid in biomolecular systems; for example, both, fused silica[113]

and silicone oil[60] exhibit negative electrostatic charge in aqueous environments. Charged interfaces create

an electric �eld that penetrates into the aqueous solution and attracts counterions, partially neutralizing

the electric �eld. These attracted counterions are part of the electric double layer that was originally

proposed by Ludwig Helmholtz.[22] Later, Gouy and Chapman expanded Helmholtz's model by introducing

the concept of a continuum distribution of ions, permitting the application of Boltzmann statistics to the

electric double layer.[22] The Gouy-Chapman model applied both Boltzmann and Poisson statistics into

order to calculate the electrical potential, ψ, in solution in the presence of a charged interface, resulting in

the Poisson-Boltzmann (PB) equation:

∇2ψ =coe

ε εo·(eeψkBT − e−

eψkBT

)(1.1)

where co, ε εo, kB , and T is the bulk concentration of salt, the permittivity of the solution, the Boltzmann

constant, and temperature, respectively.[70] For the case of planar surfaces with low potentials (ψ . 25 mV)

this equation can be simpli�ed into the Debye-Hückel equation:

ψ(x) = ψoe− xλD (1.2)

where x is the distance from the interface, ψois the surface potential, and λD =(√

2coe2

εεokBT

)−1is the Debye

length.[22, 70]

With the electrical potential de�ned, the electrostatic Gibbs free energy can be calculated. If we

consider two charged objects interacting with one another, the Gibbs free energy is equal to the formation

energy of two electric double layers, plus the energy of two objects interacting with one another. If the two

objects are identical, the electrostatic energy per unit area, we, can be simpli�ed to:[22, 70]

we(x) = 64 kBT co λD · tanh2(eψo

4kBT

)· e−

xλD (1.3)

However, we must still be integrated over the area of the two objects in order to get the total electrostatic

interaction free energy, We. Since the decay length of surface forces are small compared of the curvature

3

surface in Equation 1.3, the Derjaguin approximation can be applied to account for object geometry with

respect to an end-to-end distance between the two objects, Dx. Using the Derjaguin approximation the total

interaction energy of the system can be calculated as:

W (Dx) =

∞

Dx

w(x) · dAdx· dx (1.4)

whereA is surface the area of the objects.[22] For example, integrating Equation 1.3 with Equation 1.4between

two charged plates simply results in:

We(Dx) = 64 kBT co λD · tanh2(eψo

4kBT

)· e−

DxλD (1.5)

while applying the same integration between two spheres of radius, R, yields:[70]

We(Dx) = 64π kBT R co λ2D · tanh2

(eψo

4kBT

)· e

DxλD (1.6)

Both solutions for We indicate exponential decay that is characterized by the Debye length. Fortunately,

the Debye length does not depend on geometry or surface potential of charged particles, but only the ionic

strength and the physical properties the solution. This allows easy approximations ofWe for complex systems

when precise geometries are uncertain.

The Gouy-Chapman model describes electric double layers remarkably well; however, the theory has

limitations describing interactions at the very near interface. Stern theory attempts to adjust this failing

by explaining near surface interactions. This is done by accounting for molecular size and hydration shells

associated with charged objects, including ions. The theory partitions the system into two layers, a Gouy

layer subject to the PB equation, and a Stern layer where charged objects interact with hydration shells

around ions and other objects.[22] Often, the Stern layer may be ignored, but Stern theory is important

to acknowledge when near-interface interactions become relevant and hydration shells must be taken into

account..

1.1.2 Short-Range Interactions

Primary attractive interactions objects in solution encounter consist of three primary forces comprised

of dipole-dipole, induced dipole, and dispersion forces. Dipole-dipole forces describe the attractive forces

4

molecules with static dipoles have on one another as the molecules approach. When static dipole molecules

approach one another the molecules will adjust themselves such that the dipoles align to attract one another.

Induced dipole forces describe the attractive forces a charge (or dipole) has on an molecule without a static

dipole by inducing a dipole within the molecule leading to similar attractive forces to dipole-dipole forces.

Dispersion forces describe how two non-polar molecules attract one another by the molecules' ability to

polarize itself from �uctuations in electron density creating momentary dipoles at some frequency. The sum

of these three forces make up the van der Waals forces.[22, 70]

Van der Waals forces can be calculated from the Helmholtz free energy per unit area of two molecules

separated at distance, x, as:

wvdW (x) = −Cdipole + Cinduced + Cdispersionx6

= −CvdWx6

(1.7)

where Cdipole, Cinduced, and Cdispersion are coe�cients that represent the dipole-dipole, induced dipole, and

dispersion forces, respectively.[22] The negative sign in Equation 1.7 indicates the attractive nature of the

forces. Similar to calculating We, the Derjaguin approximation (Equation 1.4) can be used in order to

calculate the total van der Waals interaction free energy, WvdW , between two objects. Integrating Equation

1.7 between of two plates results in an in�nite energy, but the energy per unit area results as:

wvdW (Dx) = − AH12πD2

x

(1.8)

while integrating Equation 1.7 between two spherical objects of radii R1and R2, yields:

WvdW (Dx) = − AH6Dx

· R1R2

(R1 +R2)(1.9)

where the AH = π2Ctotalρ1ρ2 is the Hamaker constant and ρ is the density of the object.[22, 70] Additionally,

Equation 1.9 can be expanded for a plate and a sphere of radius R by assuming one of the radius of the

second sphere is in�nitely large leaving:

WvdW (Dx) = −AHR6Dx

(1.10)

Equation 1.7 depends greatly on proximity, due to the 1/D6x scaling. This scaling appears to depend on the

geometry of the system, seen by the 1/D2x scaling between two plates (Equation 1.8) and the 1/Dx scaling

5

between two spheres (Equation 1.9). Fortunately, van der Waals energies scale similarly at geometries

relevant for interfacial studies (Equations 1.9 and 1.10).

Often the Hamaker constant can be obtained by previous work for common materials. If values are

not available, however, Liftshitz theory may be applied to approximate the Hamaker constant using static

permittivities, ε, and the refractive indices, n, of the two objects (denoted as '1' and '2') and the interstitial

media (denoted as '3'), shown as:

AH ≈3

4kBT

(ε1 − ε3ε1 + ε3

)·(ε2 − ε3ε2 + ε3

)+

3hνe

8√

2· (n21 − n23) · (n22 − n23)√

n21 + n23 ·√n22 + n23 ·

(√n21 + n23 +

√n22 + n23

) (1.11)

where h is Planck's constant, and νe is the mean ionization frequency of the system.[22]

1.1.3 DLVO Theory

Derjaguin, Landau, Vervey, and Overbeek developed a theory for colloidal stability that accounted for

attractive van der Waals forces countered by repulsive electrostatic forces. DLVO theory highlights how these

forces interact with one another and truly how the ionic strength of solution a�ects molecular interactions.[22]

For example, A negatively-charged molecule in a 10mM NaCl solution may experience electrostatically-driven

repulsive forces at intermediate distances. If the two molecules were able to overcome this repulsive force,

the molecules will eventually encounter an attractive interaction driven by van der Waals forces. If the same

two molecules were to encounter one another in a 1 M NaCl solution, the electrostatic screening provided

by the NaCl reduces the electrostatic repulsive forces so much that only attractive van der Waals forces are

encountered as the two molecules approach.[22] Conveniently, superposition of the two forces allows simply

adding the two forces together to obtain the full DLVO energy of molecules. The addition of Equations 1.6

and 1.9 result in the sum of the energies two spheres of size R1 and R2 encounter as they approach one

another:

W (Dx) = 64πRkBTRcoλ2D · tanh2

(eψo

4kBT

)· e

DxλD − AH

6Dx

R1R2

(R1 +R2)(1.12)

1.1.4 Hydrophobic E�ect and Surface Activity

Hydrophobic e�ects between materials and surface activity of molecules toward surfaces and interfaces

remains an important phenomenon that DLVO does not completely describe. First, image a system where

6

oil molecules are dispersed in an aqueous phase. Water molecules initially form a solvation shell around

the non-polar oil molecules. The water molecules in the solvation shell have a reduced number of hydrogen

bonds, restricting translational and rotational entropy of these molecules. The non-polar oil molecules will

be pushed together in order to minimize the solvation shell surface area, minimizing the entropy loss of the

water molecules. Eventually these partitioned oil molecules form macroscopic droplets, and will �oat above

the aqueous phase due to the di�erences in density between the two phases. This phenomenon of excluding

non-polar molecules in order to minimize entropic losses of water molecules is known as the hydrophobic

e�ect.[22, 70]

The result of this thought experiment leaves an interface of hydrophobic (the non-polar oil molecules)

molecules interacting with water molecules. One would imagine the interface between these two phases would

be �at in order to minimize the surface area of the solvation shell. If the interface were to be perturbed so

that the interface was no longer �at, the free energy of the system would by increased. This change in free

energy establishes a tension at the interface between the two phases dictating the energy of a system simply

due to interfacial geometry. This concept of a surface tension can be expressed as γ = dWdA .[22] Therefore, if

an interface is curved, there is a pressure di�erence between the two phases, ∆P , since there must be some

external force acting perturbing interface. The Young-Laplace equation describes this pressure di�erence in

relation to the surface tension:

∆P = γ ·(

1

R1+

1

R2

)(1.13)

where R1 and R2 are the principle radii of curvature.[22]

For the purposes of this thesis, two common methods are utilized to measure surface tension. The

�rst is the Wilhelmy-plate method, where the surface tension of a �at air-water interface is measured by

immersing a thin plate into the aqueous phase. The downward force of created by the surface tension acting

on the plate is F = 2γl, where l is the length of the plate which can easily be measured.[22] The second

method is a pendant bubble method, common for measuring surface tension between liquid-liquid interfaces.

A oil bubble is shaped at the end of a needle within aqueous solution and the bubble's curvature is measured.

The measurements obtain R1 and R2 as well as the mean radius of curvature of the bubble at its apex, b.

Knowing the di�erence in densities between the two �uids, ∆ρ, the Bashforth-Adams equation, an extension

7

of the Young-Laplace equation, is used to calculate the surface tension:

∆P = ∆ρgh+2γ

b= γ ·

(1

R1+

1

R2

)(1.14)

where g, and h are gravitational acceleration and height of the bubble, respectively.[8]

Measuring interfacial tensions between liquids is of particular use due to the presence of surface

active molecules, or surfactants. Surfactants consist of a hydrophilic section, known as the �head�, and the

hydrophobic section, known as the �tail.� The dual properties of the molecule allow the head of the molecule

to interact favorably within aqueous solutions and the tail to to be excluded from aqueous solutions. When

introduced into a solution these surfactants will partition to an oil-aqueous interface, establishing as a surface

excess of molecules, Γ.[22] The surface excess has been well described by the Gibbs equation describing Γ in

relation to the partial di�erential of the surface tension in relation to the surface activity of the molecule, a,

which can be approximated to the molecule concentration at low activities, c:

Γ = − a

RT· ∂γ∂a≈ − c

RT· ∆γ

∆c(1.15)

where R is the gas constant.[22]

Understanding surface activity becomes important for biopolymers due to the surface activity many

contain. Protein molecules contain many hydrophobic patches within their structure. Often these patches

cannot be completely protected by hydrophillic patches of the protein which inherently make protein surface

active.[106] Nucleic acids contain a charged phosphate backbone and mildly hydrophobic aromatic nucle-

obases. The nucleobases are capable of forming hydrogen bonds, but under the correct conditions, nucleic

acids may be encouraged to become surface active as well.[119, 132]

1.1.5 Adsorption Dynamics

The previous sections have detailed how molecules may encounter interfaces and the thermodynamic

advantages in doing so. Experimentally, capturing the level of detail that the theory provided is di�cult.

Thus, a thermodynamic events is the study of adsorption isotherms, where equilibrium states of adsorbed

molecules are measured at interfaces. Investigating adsorption isotherms as a function of co allow the

development of adsorption models that may provide mechanisms of how molecules adsorb and interact with

8

the interface. The most common adsorption model for biomolecules is the Langmuir adsorption isotherm

describes surface coverage relative to a maximum surface coverage, Γmono, as:

θ =KLco

1 +KLco(1.16)

where θ = Γ/Γmono is the relative surface coverage and KL is the Langmuir constant, which is the ratio

between the adsorption rate constant (kads) and desorption rate constant (kdes), KL = kads/kdes.[22] The

Langmuir isotherm assumes that there are a �nite number of adsorption sites on an interface. Once the

interface is saturated, Γ = Γmono, no net adsorption is observed. Adsorption isotherms represent equilibrium

states between adsorption events and desorption events.

Non-equilibrium adsorption events can be characterized by the chemical equation of a molecule, A,

adsorbing to the interface as:

Abulk + Skadskdes

Asurf (1.17)

where Abulk, Asurf , and S represent species A in solution, species A adsorbed to the interface and available

surface site. The rate of accumulation of species A is equal to:

d[Asurf ]

dt= kads [Abulk] [S]− kdes [Asurf ] (1.18)

This partial di�erential equation describes basic adsorption dynamics to a surface, where molecules are

adsorbing at rate kads and desorbing at rate kdes. At equilibrium, Equation 1.18 can be solved for the

Langmuir isotherm (Equation 1.16).

Experimentally, net adsorption rates can be calculated by measuring increases in mass at an interface[68,

71, 126] or by measuring interfacial tension and relating to the surface excess using the Gibbs equation (Equa-

tion 1.15).[54, 14, 110, 116] Unfortunately, both of these techniques require, assuming an adsorption model

like Equation 1.18 and convolutes kads with kdes due to the indirect measurements these techniques employ.

Unfortunately, such indirect measurements cannot calculate kads or kdes without making broad assumptions

of the system. Later we will discuss how single-molecule techniques allow the identi�cation of adsorption

and desorption events, enabling the direct measurement of kads and kdes.

9

1.2 Molecular Conformations in Solution vs Interface

Biopolymer structure depends on hydrophobic and attractive charge interactions within the molecule.

In solution, biopolymers experience strong forces from aqueous solution that a�ect molecular con�guration,

usually encouraging polar moieties and discouraging non-polar moieties to the surface of the molecule.

Though, completely partitioning non-polar moieties within the core of the molecule is often impossible,

resulting in exposed non-molar moieties leading to surface activity.[106, 119] Once biopolymers adsorb to the

interface, van der Waals and hydrophobic forces can potentially change molecular con�gurations drastically.

At a hydrophobic interface, the molecule may change conformation so that the hydrophobic moieties within

the molecule begin to interact with the hydrophobic interface and the polar moieties of the molecule may

interact to interact more with the aqueous phase.[106, 119] The resulting change in molecular con�guration

a�ects the moelcule's polarity as well as molecular shape.[106]

1.3 Interfacial Molecule-Molecule Interactions

Interfacial molecule-molecular interactions di�er from solution molecule-molecular interactions pri-

marily due change in environmental chemistry and molecular con�nement.As discussed above molecular

polarities and conformation often change at the interface, encouraging molecule-molecule interactions which

otherwise might have been unlikely in solution.[11, 17] Newly exposed hydrophobic moieties can interact

with molecules that develop strong molecule-molecule interactions such as DNA hybridization[132] or pro-

tein gelation.[7] Additionally, the chance of interfacial molecule-molecule interactions become more likely

due to molecules requiring to explore less of a given volume before encountering another molecule at the

interface compared to solution.[44] These interactions, however, are highly dependent on interface and molec-

ular chemistry and often must be experimentally investigated in order to determine if such inter-molecular

interactions are strong enough to encourage molecular associations or gelation. [7, 106, 117]

10

1.4 Single-Molecule and Ensemble-Averaging Techniques

Traditional technologies used to study interfacial phenomena are often top-down methods that make

ensemble-averaging techniques and assume a kinetic model to obtain desired kinetic parameters.[68, 54, 135,

77] Some take advantage of intrinsic molecular �uorescence (front face �uorescence),[178, 81] while others

measure changes in interfacial properties (interfacial tensiometry). [54, 14, 110, 116] Unfortunately, due to

the complex nature of biopolymers, many ensemble-averaging techniques leave many mechanistic questions.

Due to ensemble-averaging techniques' reliance on average system behavior, top-down meathods are unable to

separate heterogeneous systems or independently measure elementary molecular events.[77] For example, such

top-down methods are unable to identify distinct oligomer protein populations and must account for these

events within model development.[98, 176, 167] In order to understand such complex behaviors microscopic

measurements, like atomic force microscopy, are employed in order to gather information on interfacial

topology.[86] However, these techniques still face many challenges when measuring interfacial dynamics.

An alternative to ensemble-averaging techniques would be �uorescence correlation spectroscopy (FCS).

FCS observes �uorescent molecules entering and exiting a sample volume.[47, 48, 91] Through the use of

speci�c adsorption and di�usion models, molecular dynamics can be inferred from the data collected.[47,

48, 91] Such studies have investigated developing protein layers at liquid-liquid interfaces[47] as well as

correlating interfacial protein di�usion to Stokes-Einstein di�usion of a hard disk[137] as opposed to the

more intuitive model of a sphere di�using at the interface.[48] Despite the ability to temporally resolve

interfacial molecular events, FCS does not have the ability to spatially resolve such events. Therefore, FCS

must rely on theoretical models to analyze temporal autocorrelation functions FCS produces.

1.4.1 Single-Molecule Techniques

The application of total internal re�ection �orescence microscopy (TIRFM)[155, 56] has demonstrated

the ability to identify interfacial heterogeneity in the form of multiple molecular populations.[76, 66, 169,

152, 120] TIRFM utilizes �uorescence to excite molecules at near interfaces by internally re�ecting a light

source at an interface (e.g. oil-water interface), creating an evanescent wave that propagates into the re�ected

11

medium. This evanescent wave decays exponentially, e�ectively only illuminating molecules ∼ 100 nm slab

at the interface. This illumination reduces background �uorescence of low concentration solutions such that

single-molecules can be observed at the interface as a di�raction-limited spot, while illuminated molecules

not at the interface move too quickly to resolve, resulting in system noise. Because of this, TIRFM boasts

the ability to separately observe elementary adsorption events, desorption events, interfacial lateral position,

and object �uorescence intensity.[66, 169, 76] From these measurements, adsorption rates, desorption rates

(residence times), molecular di�usion coe�cients, and mean molecular intensities can be calculated inde-

pendently form one another.[66, 169, 76] The application of TIRFM has led to direct interfacial di�usion

measurements of surfactants,[66] apparent activation energies for surfactants[66] and proteins,[96] and DNA

association interactions[119] at the solid-liquid interface for multiple surface chemistries.[164, 76] Due to

the spatial �delity TIRFM can resolve, separate chemical environments may be explored in parallel using

the super-resolution technique MAPT (Mapping using Accumulated Probe Trajectories).[164] Additionally,

since TIRFM relies on basic �uorescent principles, similar �uorescent strategies such as resonance energy

transfer methods can be utilized to achieve a molecular sized ruler between two molecular probes.[79, 97]

The utilization of TIRFM has also been applied brie�y to the liquid-liquid interface.[167, 168, 152]

Initially, the application of the Stokes-Einstein equation[137] and the intensity distribution of molecules

observed at the silicone oil-water interface, allowed the distinct identi�cation of bovine serum albumin (BSA)

monomer, dimer and trimer oligomers in di�use concentration. These techniques were quickly followed by

an investigation of protein layer formation the the silicone oil-water interface, using �uorescent BSA as an

interfacial probe leading to the characterization of a heterogeneous model of protein layer formation.[168]

Finally, Sriram et al. investigated how di�usion coe�cients are a�ected by interfacial viscosity and that

di�usion coe�cients begin to deviate from the Stokes-Einstein equation at interfacial viscosities above ∼ 1000

cSt.[152]

1.5 Materials

This thesis focuses on aqueous phase interaction with various hydrophobic and hydrophillic interfaces.

Bu�ers selected in this thesis are used to encourage biopolymer interaction with the studied interface,

12

while attempting to prevent undesired interactions in solution like solution-based DNA hybridization or

protein aggregation. Various interfaces have been selected for this thesis, some for a reliable charge density

throughout the experiment (e.g. fused silica), others for their unique chemical properties (e.g. liquid cystals),

or their relevance in the pharmaceutical industry (e.g. silicone oil). This section will describe the importance

of some of the more exotic materials used within this thesis.

1.5.1 Liquid Crystal

Liquid crystals (LCs) contain birefringent properties like crystalline materials, while in a liquid phase.

Birefringence is a property where polarized light perpendicular to the optical axis of a material (the ordinary

axis) experiences a refractive index, no, di�erent from polarized light parallel to the optical axis of a material

(the extraordinary axis), ne. The di�erence in these refractive indices is the birefringence:

∆n = ne − no (1.19)

For the case of thermotropic nematic LCs, molecules typically are cigar shaped where the optical axis lies

along the length of the molecule (Figure 1.1a).[33] Birefringence allows a material to change the polarization

of light incident on the material. Figure 1.1b demonstrates how light polarization is a�ected when incident

upon a birefringent medium of LC molecules. The incident light encounters both extraordinary and ordinary

axes of the molecules, changing the polarization of the light. Figure 1.1c demonstrates how light polariza-

tion is una�ected when incident upon LC molecules oriented along the axis of propagation (also known as

homeotropic alignment). The incident light only encounters the ordinary refractive index, thus leaving the

light polarization una�ected.

LC molecules also contain viscoelastic properties, which allow LC molecules to be aligned along a

director, −→n . The director does not necessarily have to be the same direction throughout the medium,

implying that one interface of a LC layer can be aligned in one direction while another interface can be

aligned completely di�erently (see Figure 1.1b). The physical properties of the LC and the temperature

of the system will dictate how well molecules align to the director, which is often referred to as the order

13

Figure 1.1: (a) LC molecule with ordinary and extraordinary axes, exhibiting refractive indices, no and ne,respectively (b) light propagation through orthogonally crossed polarizers with a LC layer exhibiting a tiltedcon�guration situated in between the polarizers (c) light propagation through orthogonally crossed polarizerswith a LC layer exhibiting a homeotropic con�guration situated in between the polarizers

parameter, S:

S =2 cos2θ − 1

2(1.20)

where θ is the angle between the LC molecular axis and the local director.[33] An order parameter of S = 1

would be a perfectly ordered system, where an order parameter of S = 0 would be a disordered system,

also known as an isotropic system. Typically LCs have an order parameter from S =0.3 to 0.8.[33] Liquid

phase thermotropic nematic LCs molecules only exhibit order parameters greater than 0 within a speci�c

temperature range, this is known as the nematic phase. When a LC exceed these temperatures, the molecules

will become disordered and become isotropic. These properties make LC phases interesting to study, since

interfacial interactions may also a�ect the LC director resulting in a visual response.[87, 132]

1.5.2 Silicone Oil

Silicone oil is ubiquitous within the biopharmaceutical industry due to the popular packaging strategy

of using pre�lled glass syringes as a delivery process, which require the use of silicone oil as a lubricant.

However, the presence of silicone oil interface has been of increasing scrutiny within the biopharmaceutical

industry due to increased observations protein aggregates within siliconized syringes.[153, 10, 27, 73, 109, 147]

The molecular structure of silicone oil is similar to polydimethylsiloxane (PDMS, Figure 1.2), except that

14

Figure 1.2: Molecular structure of PDMS

in silicone oil polymers have been cross-linked between one another in order to increase the oil's viscosity.

Silicone oil exhibits hydrophobic qualities and does not dissolve in water; however, silicone oil does exhibit

a negative charge when subject to zeta potential measurements.[60, 10] Silicone oil's complex charge and

hydrophobic properties make studying this material interesting s, due to the di�culty in predicting how

biopolymers such as protein will interact with the oil.

1.6 Thesis Goals

The objective of this thesis is to better understand mechanisms at which biopolymers interact at

liquid and solid interfaces. This is �rst accomplished by studying mechanisms of how biopolymer adsorption

a�ects surfactant organization and vise versa at a hydrophobic oil interface. Second, the e�ects of electrostatic

forces on interfacial molecular dynamics to a charged surface are studied. Finally, protein �lm layer formation

mechanisms are studied at a hydrophobic interface.

1.7 Thesis Organization

Chapter 2 of this thesis focuses on understanding the mechanism at which DNA interacts with cationic

surfactant at the aqueous-LC interface. Polarized light microscopy (PLM) and epi�uorescence are used to elu-

cidate the mechanism at which single stranded DNA (ssDNA) and hybridization events a�ect LC orientation

upon a cationic surfactant laden interface. The use of surfactants of di�erent chemistries and polyanion-

ions tested how ssDNA interacts with interfacially-laden surfactants that induce orientational changes in

the hydrophobic LC subphase. Thus providing mechanistic information on how nucleic acids interact with

surfactants at a hydrophobic interface.

Chapter 3 examines how elementary molecular events, such as adsorption rates, desorption rates, and

15

interfacial molecular di�usion, are a�ected by electrostatic forces, driven by charge repulsion and electrostatic

screening at the solid-liquid interface. TIRFM is used to investigate these molecular events of a protein

interacting with a negatively-charged interface.

Chapters 4 and 5 focus on the protein layer formation at the silicone oil-water interface. Chapter

4 utilizes TIRFM and dynamic interfacial tensiometry (D-IFT) to investigate molecular trajectories and

interfacial tension of a developing protein. The data collected allows the development of micro-rhological

techniques by analyzing interfacial molecular trajectories and the identi�cation of two protein layer formation

models. Chapter 5 utilizes TIRFM again to investigate how protein layer formation is a�ected by surfactant

concentration. Measurements were compared to macroscopic aggregation studies to propose a mechanism of

protection surfactants provide to proteins at interfaces.

Finally, Chapter 6 summarizes the work completed within the previous chapters and provides some

concluding remarks and �nal thoughts on the importance of understanding interfacial mechanisms and the

potential importance the development of micro-rheology analysis may aid in further mechanistic studies.

Chapter 2

Surfactant�DNA Interactions at the Liquid Crystal�Aqueous Interface

2.1 Abstract

The presence of single-stranded (ssDNA) vs. double-stranded (dsDNA) DNA at a surfactant-laden

aqueous-nematic liquid crystal (LC) interface results in distinctly di�erent orientations of the LC molecular

axis; this is of practical interest as a method to detect DNA hybridization. Results presented here provide

new insights into the molecular-level mechanisms of these phenomena. The adsorption of ssDNA to a cationic

surfactant-laden aqueous-LC interface caused LC reorientation, leading to coexistence between homeotropic

and planar (birefringent) oriented regions. Fluorescence microscopy revealed that ssDNA preferentially par-

titioned into the bire�ngent regions, presumably causing a decreased surface coverage of surfactant and

the resultant planar LC orientation. Both electrostatic and hydrophobic e�ects were found to be critical

to inducing LC reorientation. In particular, insu�cient ssDNA adsorption occurred in the absence of a

cationic surfactant (e.g. with no surfactant or with a non-ionic surfactant), demonstrating the importance

of electrostatic interactions with the polyanionic ssDNA. Even in the presence of a cationic surfactant, how-

ever, polyanions without hydrophobic side-group moieties (poly[acrylic acid] and dsDNA) caused no LC

reorientation, while polyanions with hydrophobic side-groups (polystyrene sulfonate and ssDNA) initiagfted

the desired LC reorientation. These observations are consistent with the fact that interfacial hybridiza-

tion of adsorbed probe ssDNA to complementary target ssDNA caused a reorientation from planar back to

homeotropic. We propose that ssDNA forms an electrostatic interfacial complex with cationic surfactant

where the hydrophobic nucleobases associate directly with the LC phase, e�ectively competing with surfac-

17

tant molecules for interfacial sites. Upon hybridization, the hydrophobic character of the ssDNA is lost and

the nucleobases no longer associate directly with the LC phase, allowing the surfactant molecules to pack

more closely at the interface.

KEYWORDS: hybridization, liquid crystal, hydrophobic interaction, interface, nucleic acid, biological

sensor

2.2 Introduction

The detection of DNA hybridization events has applications in a wide range of disciplines including the

agricultural[157] and food industries[131], infectious disease detection[122], and gene expression[89]. Exam-

ples of common technologies available to detect hybridization events include RT-PCR[127], colorimetric[175],

and �uorescence based assays[145]. While these strategies are well established and in some cases commercially-

available[102, 63], the associated technologies have limitations that a�ect their potential for point-of-care

(POC) applications, e.g. some approaches are su�ciently reliable for a lab setting but are not robust enough

for POC devices. Multiplexing has been demonstrated using certain well-established strategies[52]; however,

poorly-controlled nonspeci�c DNA binding to surfaces commonly used in microarray applications[25, 26]

reduces sensitivity, creates �false positive� responses, and limits the utility of these devices in quantitative

analysis. Thus, there is a need for a DNA hybridization detection strategy that eliminates these issues while

maintaining the high performance of these successful approaches.

Liquid crystal (LC) � based sensors have demonstrated potential as a platform for the detection of DNA

hybridization events[94, 111, 132], harmful gases[15, 1], protein reactions[17], and ssDNA quanti�cation[94,

28]. LC materials exhibit states of matter that are intermediate between a true crystal and an isotropic �uid.

Several LC phases exist with varying degrees of order, but the one most commonly-used in sensor applications

is the nematic phase of thermotropic LCs, which possesses orientational order and lacks translational order.

Orientational order leads to optical birefringence that is readily detected using polarized light; this provides a

convenient way to determine molecular orientation. LC sensors typically operate by detecting a perturbation

in the LC orientation caused by the presence of an analyte. This strategy o�ers several potential advantages in

microarray applications. LC sensors generally involve a label-free detection strategy that speci�cally detects

18

the presence of an analyte, thus, reducing the risk of potential false positive signals due to nonspeci�c binding

events. Furthermore, LC materials can be obtained at a low cost, provide a read-out signal (i.e. presence or

absence of optical birefringence) that can be observed by the naked eye, and can be readily developed into

a robust device making this strategy attractive for POC applications. Furthermore, LCs have an intrinsic

elastic energy that causes a small perturbation in the LC alignment to propagate over distances on the

order of 1 to 100 μm; a natural ampli�cation e�ect that provides a macroscopic response to a microscopic

perturbation.[111] While LC sensor applications are promising, the molecular mechanisms associated with

the underlying phenomena are generally poorly understood.

Figure 2.1: LC response to ssDNA adsorption and hybridization: (a) Polarized microscopy images of theaqueous/LC interface laden with OTAB, (b) after subsequent adsorption of ssDNA, and (c) after interfacialhybridization.

Several strategies for the detection of DNA hybridization using a LC platform have been reported. In

one approach, a solid substrate decorated with immobilized peptide-nucleic acids was shown to induce planar

LC anchoring; sequential hybridization and surfactant adsorption at the interface resulted in a transition

to homeotropic anchoring.[95] Another strategy detected di�erent LC orientations induced by either single-

stranded (ssDNA) or double-stranded DNA (dsDNA) combed onto a solid substrate.[111] In previous work by

our research group, it was observed that the adsorption of ssDNA to a cationic surfactant-laden aqueous-LC

19

interface induced a LC anchoring transition from homeotropic to planar (Figure 2.1a,b) and subsequent DNA

hybridization caused a transition back to homeotropic LC orientation (Figure 2.1b,c).[132] Our objective

here was to develop a mechanistic understanding of these particular phenomena, i.e. the LC reorientation

upon DNA adsorption and hybridization at an octadecyltrimethylammonium bromide (OTAB) aqueous-LC

interface.

Interactions between DNA and surfactants have previously been studied in several contexts, and

suggest complex e�ects due to both electrostatic and hydrophobic associations. Electrostatic interactions

contribute to the creation of DNA-surfactant complexes in bulk solution to form aggregates at surfactant

concentrations below the critical micelle concentration (cmc)[41], and dsDNA is also found to form complexes

with cationic surfactants at the air-water interface.[35, 50] The cationic surfactants in these DNA-surfactant

complexes create an environment that increases the melting temperature of dsDNA (i.e. they promote

hybridization).[144, 143] At the air water interface, cationic surfactants with an amine head group (ie.

hydrogen-bonding moiety) have been shown to denature DNA into its single-stranded form[50], also resulting

in a decreased interfacial concentration of the surfactant.[50, 144] With respect to hydrophobic e�ects, it has

been suggested that exposed hydrophobic bases of ssDNA result in a further reduction of the cmc in DNA-

surfactant complexes.[42] Furthermore, signi�cant adsorption of ssDNA to uncharged hydrophobic surfaces

has been demonstrated, and attributed to the hydrophobic nature of the exposed bases of ssDNA.[25]

These previous studies suggest that both electrostatic and hydrophobic interactions will likely play a

critical role in the system studied here, while the role of hydrogen bonding would appear to be less signi�-

cant since the surfactants under consideration (Figure 2.2) lack a hydrogen bonding moiety. Electrostatics

is omnipresent when considering the interaction between a polyanion and a cationic surfactant; however,

hydrophobic interactions, while of critical importance, will be particularly relevant to certain phenomena

(e.g. ssDNA conformation). The experiments described here were designed to elucidate the relative im-

portance of these various interactions with respect to the mechanisms associated with the LC orientational

response. In particular, the LC response to polyanion adsorption was measured as a function of polyanion

hydrophobicity, surfactant charge, and surfactant coverage. We also report measurements of DNA mobility

and DNA partitioning to either homeotropic or birefringent domains.

20

2.3 Materials and Methods

2.3.1 LC Film Preparation and Polyanion Addition

LC optical cells consisted of an aqueous/LC interface, where adsorbate molecules were introduced, and

an opposing solid/LC interface that induced homeotropic orientation. These cells were housed in wells that

were prepared by punching a ~5 mm diameter hole in 2.5 mm thick silicone rubber sheets (Sigma-Aldrich) us-

ing a standard hole punch. A sheet containing one or more wells was placed onto a soda lime glass microscope

slide (Corning Inc.) that had been modi�ed with an OTES SAM (see below), and an electron microscopy

grid (Electron Microscopy Sciences) was placed on the slide in the center of each well. In order to adsorb

surfactants at the aqueous/LC interface, surfactant molecules were dissolved into the LC phase. Figure 2.2

shows the di�erent surfactants used. We used three cationic surfactants: dioctadecyldimethylammonium

bromide (DODAB), octadecyltrimethylammonium bromide (OTAB), and dodecyltrimethylammonium bro-

mide (DTAB) (Sigma). We also used the nonionic surfactant tetraethylene glycol (C12E4) (Sigma). Stock

solutions of surfactant were prepared in chloroform (Fischer), known volumes of which were then added to

LC (either 5CB or E7, Merck KGaA, both LC materials are thermotropic and exhibit a nematic phase at

the operating conditions used; T=25°C) at the desired concentration. This mixture was then dried under a

stream of ultrapure N2 (Airgas), evaporating the chloroform from the LC and leaving a mixture of surfactant

dissolved in LC at a desired concentration. A micropipette was used to add 250nL of the surfactant/LC mix-

ture to the electron microscopy grid at a temperature above the nematic-to-isotropic transition temperature

of the particular LC (>35 ºC for 5CB and >60ºC for E7). Excess LC was then removed from the grid via

capillary action through contact with a clean micro capillary tube. The LC �lm was then allowed to slowly

cool to room temperature and the LC cell was placed onto the microscope stage. The silicone well was then

�lled (~50μL) with either an aqueous solution of 5mM NaCl (pH ~5.5-6) for most experiments or a 10 mM

Tris-Cl, 1mM EDTA bu�er (pH ~7.5) for pure dsDNA addition (see below) and was then imaged between

crossed polarizers using a a Nikon Eclipse Ti microscope equipped with a Nikon DS-Fi1 color C-MOS camera.

21

Figure 2.2: Surfactants and polymers used in the experiments. From left to right: cationic surfactants

with decreasing surface activity (DODAB, OTAB, and DTAB), and a nonionic surfactant (C12E4). The

DNA analogs used included polyanions with or without hydrophobic sidegroup moieties (PSS and PAA

respectively).

Polyanions were added through the aqueous phase in order to measure their e�ect on the LC orien-

tation upon adsorption, the mobility of certain polyanions at the interface (see below), and the partitioning

of DNA between LC domains. Prior to the addition of polyanion, the LC orientation due to the adsorption of

surfactant was imaged between crossed polarizers. Unlabeled ssDNA (probe) (5'AGAAAAAACTTCGTGC3';

Biosearch), �uorescently labeled ssDNA probe (3'-modi�ed with AlexaFluor 568; Biosearch), puri�ed dsDNA

(see 2.8), polystyrene sulfonate (PSS; Figure 2.2) (75,000 MW, Sigma), or poly (acrylic acid) (PAA; Figure

2) (1,800 MW, Sigma) was added to the aqueous phase at concentrations of 2.5 μM, 2.5 μM, 2.5 μM, 14

μM, and 14 μM respectively. The LC orientation was continuously monitored using polarized microscopy.

When applicable, the DNA partitioning was also monitored with epi�uorescence microscopy (Nikon Eclipse

Ti microscope out�tted with a second camera; Photometrics Cascade 512B). In experiments that involved

studying the LC response to interfacial hybridization, a solution containing complementary ssDNA (tar-

get) was subsequently added to the aqueous phase. If �uorescent probes were used in the �rst polyanion

addition a bu�er exchange was completed prior to target addition, to remove any background �uorescence

22

due to unabsorbed �uorescent probes. Either unlabeled target (5'GCACGAAGTTTTTTCT3', Biosearch)

or �uorescently labeled target (3'-modi�ed with �uorescein, Biosearch) was added to the aqueous phase

(~1�10pmol) and the LC orientation was continuously monitored using polarized microscopy; again the

DNA partitioning was monitored using epi�uorescence microscopy when applicable.

2.3.2 OTES Preparation

Self-assembled monolayers (SAMs) of octadecyltriethoxysilane (OTES) (Gelest Inc.) were prepared

according to published procedures.[163] Soda lime glass microscope slides (Corning Inc.) were cleaned

sequentially with 2% aqueous micro-90, deionized water (18.2 MΩ), and piranha solution (30% aqueous

H2O2 (Fisher Scienti�c) and concentrated H2SO4 (Fisher Scienti�c) 1:3, v/v) at 80°C for 1 hr. (Warning:

piranha solution reacts strongly with organic compounds and should be handled with extreme caution; do

not store in closed container). After piranha cleaning, the microscope slides were rinsed with deionized

water (18.2 MΩ) and dried under a stream of ultrapure N2. A deposition solution of n-butylamine (Fisher

Scienti�c) and OTES was prepared in toluene (Fisher Scienti�c) at 1:3:200 volumetric ratios, respectively.

The deposition solution was warmed to 60°C, clean and dry microscope slides were �rst rinsed with toluene,

and then submerged in the warm deposition solution. The slides were incubated in the deposition solution

for 1 hour at 60°C. Upon removal, the slides were rinsed with toluene, dried under a stream of ultrapure

N2, and stored at room temperature in a vacuum desiccator. A custom-built contact angle goniometer was

used to verify that the water contact angle (θC , measured via the static sessile drop method) of the prepared

SAMs was su�cient to achieve strong homeotropic anchoring (θC > 95◦).

2.3.3 Fluorescence Recovery after Photobleaching (FRAP)

LC �lms were prepared as described above with varying bulk concentrations of OTAB in LC (40,

80, 100, 140, or 200 μM) and contained within a �ow cell environment (see Supplementary Material) in-

stead of the micro well environment previously described. An aqueous solution of 5mM NaCl (pH=5.5-6)

was introduced to the �ow cell and incubated for 5-10min. A 2.5μM solution of �uorescently labeled ss-

DNA (5'AGAAAAAACTTCGTGC3' � Fluorescein; (Invitrogen)) was then introduced and incubated for

23

an additional 5-10 min. A bu�er exchange of the bulk aqueous phase was completed by �owing ~1mL of

aqueous 5mM NaCl (pH=5.5-6) through the �ow cell to remove background �uorescence due to the presence

of the �uorescent ssDNA. The LC orientation was measured using polarized light microscopy to evaluate

homeotropic or planar LC orientation. To photobleach a small circular region at the interface, the aperture

on the microscope was closed to its smallest size and the sample was illuminated for ~10 sec using a 120W

Mercury Vapor Short Arc lamp (model CHGFIE, Nikon). The �uorescence recovery was then measured by

opening the aperture and collecting a time series of images for 10-40 minutes. The �uorescence intensity was

analyzed using ImageJ (NIH Freeware) and the fractional �uorescence recovery curve (fK(t)) was calculated:

fK(t) =F (t)− FK(0)

FK(∞)− FK(0)

where FK(t) is the average intensity of the photobleached area at time t, and FK(∞) is the average intensity

of the region surrounding the photobleached area. A custom written Mathematica code was used to �t the

fractional recovery function to the equation[151]:

fK(t) = Ae−2τ/t[Io(2τ/t) + I1(2τ/t)]

where τ is the time constant used as the free parameter, A is a constant which represents the fraction of

mobile molecules (constrained to A = 1), t is time, Io is a zero-order modi�ed Bessel function of the �rst

kind and I1 is a �rst order modi�ed Bessel function of the �rst kind. The time constant, τ , was then used

to calculate the di�usion coe�cient (D):

D =r2

4τ

where r is the radius of the circular photobleached area.

2.4 Results

2.4.1 ssDNA Adsorption and LC Reorientation

In previous work, we found that under appropriate conditions, adsorption of ssDNA to an OTAB laden

aqueous-LC interface induced a reorientation of the LC director from homeotropic to planar.[132] As shown

24

in Figure 2.3, systematic studies found that this reorientation occurred only when operating at an OTAB

surface coverage near the minimum required for homeotropic orientation, which corresponded to a bulk

[OTAB] ∼= 100 µM in our experimental system. At a bulk [OTAB] well below 100μM (Figure 2.3a � 2.3c),

the LC orientation was una�ected by the adsorption of ssDNA (i.e. the LC remained birefringent), while at

a bulk [OTAB] su�ciently greater than 100 μM (Figure 2.3g-2.3i), homeotropic LC orientation was observed

regardless of the presence of ssDNA at the interface (although a transient nucleation of small birefringent

domains was sometimes seen; Figure 2.3h). At an [OTAB] ∼= 100 µM, where the surface coverage was

near the minimum required for homeotropic orientation (Figure 2.3d-2.3f) a LC reorientation was observed

upon ssDNA adsorption. In particular, small birefringent domains nucleated (Figure 2.3e) and eventually

coalesced to form large domains of stable planar LC orientation (Figure 2.3f).

Figure 2.3: LC reorientation upon ssDNA adsorption at varying [OTAB] � Polarized light microscopy imagesof the LC-aqueous interface at low [OTAB] (a-c) intermediate [OTAB] (d-f) and high [OTAB] (g-i) before(a,d,g) and after ssDNA adsorption at 1min (b,e,h) and 15min (c,f,i).

2.4.2 Cationic and Nonionic Surfactant Monolayers

The LC response to ssDNA adsorption and hybridization was studied using surfactants with varying

surface activity and head group charge. Three cationic surfactants with decreasing surface activity were

employed (DODAB, OTAB, DTAB, respectively) as well as a nonionic surfactant (C12E4). In all cases, a

25

Figure 2.4: ssDNA Di�usion: The di�usion coe�cient, measured via FRAP, of ssDNA at an OTAB ladenaqueous-LC interface with varying OTAB coverage.

homeotropic LC orientation was induced by the presence of su�cient surfactant. The same qualitative LC

response was observed with all three cationic surfactants upon addition of probe ssDNA; i.e. a LC reorienta-

tion was observed from homeotropic to planar upon ssDNA adsorption (see Figure 2.2a,b). Furthermore, the

characteristic response to hybridization upon addition of complementary target ssDNA (see Figure 2.2b,c)

was also observed for all three cationic surfactants. However, with nonionic surfactant no LC reorientation

was observed upon ssDNA adsorption; in fact, little to no �uorescently-labeled ssDNA remained adsorbed

to the interface following an exchange of the bulk aqueous phase. The same lack of ssDNA adsorption was

observed in the absence of surfactant. These results suggest that electrostatic interactions are critical to

achieve su�cient ssDNA adsorption to realize a LC response, and that the surfactant surface activity has

little to no a�ect on the ability to achieve a LC reorientation upon interfacial hybridization.

2.4.3 ssDNA Surface Coverage and Di�usion

Fluorescence microscopy was used to measure relative surface coverage and mobility of ssDNA at

the aqueous-LC interface as a function of OTAB surface coverage. At [OTAB] = 100 μM, the coexistence

of homeotropic and birefringent domains observed following ssDNA adsorption provided an opportunity to

measure the ssDNA coverage and mobility simultaneously in both planar and homeotropic regions. The

di�usion coe�cients obtained were on the order of 0.1-1µm2/sec as shown in Figure 2.4.

26

According to the Einstein relation, the di�usion coe�cient, D, is related to the mobility, µ, through the

expressionD = µkBT , where kB is the Boltzmann constant and T is the absolute temperature. Depending on

the experimental conditions, the interfacial mobility, µ, may be dominated by viscous drag from the adjacent

bulk phases (the small Boussinesq number limit)[69] or from drag associated with molecular crowding within

the interfacial layer (the large Boussinesq number limit)[141], which considers surface crowding (i.e. surface

viscosity) to be the dominant determinant of surface di�usion. In the former case, the minimum di�usion

coe�cient one could observe is 2.78 µm2/s, assuming an extreme radius of gyration, RH = 47.2 Å, for ssDNA

lying �at at the interface. This suggests that the drag on ssDNA during interfacial di�usion is dominated

by surface viscosity (i.e. lateral interactions between molecules within the interfacial layer) rather than drag

from the LC and aqueous phases; the measured interfacial di�usion coe�cients are reasonable for this regime.

For example a value of D = 0.6 µm2/s is consistent with a molecule with RH = 10 Å di�using within a layer

with a surface viscosity of 1.79 x 10-9 P-m. Therefore, the interfacial mobility provides information about

the magnitude of interfacial crowding.

Notably, the di�usion coe�cient was consistently smaller in regions of planar LC orientation than

in regions of homeotropic LC orientation. As discussed above, smaller di�usion coe�cients are indicative

of increased surface crowding. Fluorescence images illustrating the partitioning of ssDNA at the interface

validate this claim. Figure 2.5a shows a polarized light microscopy image of a cationic surfactant laden

aqueous-LC interface after the adsorption of �uorescently labeled ssDNA, while Figure 2.5b shows its corre-

sponding epi�uorescence image. These images illustrate that, in general, the ssDNA surface concentration

is higher within the regions of planar LC orientation than in the regions with homeotropic LC orientation,

consistent with the slower di�usion in these regions. We note that some anomalous weakly-birefringent

sub-domains in Figure 2.5 exhibit a lower relative ssDNA coverage. Following these anomalous regions with

time, we found that they eventually transitioned to a homeotropic orientation.

2.4.4 PSS, PAA, and dsDNA Adsorption

We studied the LC response upon adsorption of PSS and PAA to isolate the role of electrostatic and

hydrophobic interactions. The rationale for these experiments involved the notion that the structure of PSS

27

Figure 2.5: ssDNA adsorption at an aqueous-LC interface: (a) Polarized light microscopy image of a cationicsurfactant laden aqueous-LC interface after ssDNA adsorption. (b) the same �eld of view imaged withepi�uorescence microscopy.

is analogous to that of ssDNA in that it is polyanionic with hydrophobic side group moieties, while PAA is

purely polyanionic and lacks hydrophobic moieties. Excess PSS and PAA were added to the aqueous phase

in contact with LC �lms containing [OTAB] ∼= 100 µM. In the case of PSS, we observed reorientation of the

LC director from homeotropic to planar/tilted, consistent with the LC reorientation observed upon ssDNA

adsorption. Interestingly, upon the addition of PAA to the OTAB-laden interface there was a negligible

e�ect on LC alignment (i.e. the LC orientation remained homeotropic). Similarly, the addition of puri�ed

dsDNA also did not induce a LC reorientation. These results suggest that a polyanion lacking hydrophobic

moieties does not have the ability to induce a change in the LC orientation at a cationic surfactant laden

aqueous/LC interface.

2.4.5 Interfacial Hybridization

As observed previously[132], upon hybridization, the �uorescently labeled target DNA was concen-

trated in regions of homeotropic LC orientation (Figure 2.6a,b), proving that the dsDNA is preferentially

found within homeotropic regions. Interestingly, we also found that when unlabeled target was hybridized

to previously-adsorbed �uorescently-labeled probe (Figures 2.6c,d), the �uorescently labeled probe was also

found at higher concentrations in the homeotropic regions. Combined, these two observations suggest that

the interfacial packing density of dsDNA (in homeotropic regions) after hybridization was higher than the

28

Figure 2.6: dsDNA hybridization at an aqueous-LC interface: Polarized light microscopy (a,c) and epi�uo-rescence microscopy (b,d) images of an OTAB laden aqueous-LC interface after hybridization using either�uorescently labeled target (a,b) or probe (c,d).

packing density of ssDNA (in birefringent regions) before hybridization.

2.5 Discussion

The results reported above suggest that attractive electrostatic interactions between the interfacially-

bound surfactant and the polyanion introduced into the aqueous phase are necessary for signi�cant polyanion

adsorption to occur, but not su�cient to induce an LC reorientation from homeotropic to planar. In par-

ticular, polyanions with hydrophobic side-chains (ssDNA and PSS) were able to induce LC reorientation

from homeotropic to planar following adsorption at appropriated surfactant concentrations, while polyan-

ions without hydrophobic moieties (dsDNA and PAA) did not a�ect the LC orientation. This suggests that,

in conjunction with electrostatic interactions between polyanions and a cationic interface, hydrophobic in-

teractions between hydrophobic side-groups and the LC material are necessary to induce a LC reorientation.

Our results also suggest that a critical surface coverage of long-chain surfactant is necessary to induce

homeotropic anchoring. Similar conclusions have been drawn from a number of previous studies where planar

anchoring was observed at the LC-aqueous interface in the presence of low surfactant concentrations, and a

transition to homeotropic anchoring was observed at a critical surfactant concentration.[132, 16, 18] This is

also consistent with LC anchoring on solid surfaces, where a low concentration of long alkyl chains was shown

29

to induce planar anchoring, but a su�ciently high surface concentration of long chains caused homeotropic

anchoring.[123] Electrostatic e�ects have also been shown to a�ect the orientation of polar liquid crystals

(like the cyanobiphenyls used here); however, the expected orientational e�ects due to electrostatics would

be exactly opposite to what we observe. For example, the adsorption of anionic ssDNA to the cationic

surface layer has the potential to produce a polar double-layer; however, it has been shown that this sort of

double-layer causes homeotropic anchoring[179, 154], while we observe the exact opposite trend. Therefore,

it is reasonable to conclude that the orientational transitions observed upon surfactant adsorption, probe

ssDNA adsorption, and target DNA hybridization are dominated by �steric� LC anchoring e�ects associated

with changes in the local interfacial concentration of the long-chain surfactant.

We hypothesize that a reduction in the interfacial surfactant concentration can be caused by the

intercalation of hydrophobic side chains (e.g. nucleobases) upon ssDNA adsorption. At su�ciently low

OTAB coverage the DNA molecules are capable of intercalating between the surfactant molecules, but at

higher surfactant coverage the surfactant molecules are too densely-packed for this to occur. In particular, at

an [OTAB] = 100 μM (the minimum concentration required for homeotropic anchoring), a lower limit for the

surfactant molecular area (assuming all surfactant molecules adsorb at the interface) is calculated to be 0.92

nm2/molecule. This is ~5 times the close-packed molecular area (~0.18 Å), suggesting that a substantial

amount of open space is available at the interface, roughly 0.74 A2 per surfactant molecule. The notion that

such a relatively low surface concentration of long chains is able to induce hometropic anchoring is consistent

with recent results at a solid surface where a long-chain surface coverage of only ~11% caused homeotropic

anchoring.[123] Given an approximate cross-sectional area of ~0.4 A2/nucleotide in ssDNA[36], there is space

for 1�2 nucleotides per surfactant molecule to reside within the interfacial layer and interact strongly with

the LC subphase. Furthermore, ssDNA is highly �exible (persistence length ~5.9Å)[36] allowing for the DNA

molecule to bend in a way that allows these hydrophobic interactions to occur. We hypothesize that at high

bulk OTAB concentrations, the interface is su�ciently crowded with surfactant molecules, that individual

nucleotides can no longer easily intercalate into the layer and associate with the LC phase. In this case, the

ssDNA may still bind electrostatically to the surface, but no LC reorientation is observed.

These hypotheses are supported by our observations of ssDNA interfacial di�usion as a function of

30

OTAB concentration and LC orientation. Speci�cally, when the ssDNA intercalates between the surfactant

molecules one would expect increased surface crowding and a commensurate decrease in di�usion. This is

in agreement with our observations, where we observe higher ssDNA concentrations and slower di�usion in

birefringent regions (corresponding to low OTAB concentration) and lower ssDNA concentrations and faster

di�usion in homeotropic regions (corresponding to high OTAB concentration).

The results of experiments using ssDNA analogs provide further support for the idea that hydropho-

bic side-chain moieties intercalate between surfactant molecules to cause LC reorientation. In particular,

the hydrophobic aromatic moieties on PSS are capable of interacting with the LC phase via hydrophobic

interactions in analogy with the nucleobase/LC interactions described above. In the absence of hydrophobic

moieties, adsorption of a polyanion (PAA) caused no LC response. In related work, Kinsinger et al.[88]

found that the addition of a PSS salt, sodium polystyrene, to an amphiphilic polymer adsorbed to the

LC/aqueous interface induced a LC reorientation from homeotropic to planar, providing further evidence

that the addition of a hydrophobic polyanion to an adsorbed cationic surfactant, either free or polymeric

may induce a LC reorientation. The prominent role of hydrophobic interactions as evidenced from our ex-

perimental observations and the literature further supports the theory that the LC realignment is related

to the intercalation of the exposed hydrophobic bases of ssDNA intercalating between surfactant molecules

and altering the coverage of surfactant.

We propose that the strongly-associated interfacial DNA-surfactant complex with intercalated nu-

cleobases results in ssDNA con�gurations that are predominantly two-dimensional within the plane of the

interface. While this two-dimensional con�nement involves an entropic penalty, this is o�set by the fa-

vorable interactions between the nucleobases and the hydrophobic LC phase. A reasonable estimate for

the strength of this hydrophobic interaction is 30 cal mol−1 A−2.[146] If we assume conservatively that

half of the nucleotides take part in these interactions, the total binding energy is ~11 kcal/mol. The ap-

proximate loss in entropy due to two-dimensional con�nement of the polymer strand can be calculated as

∆S = kB [ln(W3D)�ln(W2D)], where W3D and W2D are the multiplicities of a random walk in three and

two dimensions respectively. This calculation yields the value T∆S ∼= 5kcal/mol at room temperature, sug-

gesting that the favorable hydrophobic interactions may plausibly compensate for the loss of con�gurational

31

entropy resulting from 2D con�nement. As discussed below, this hypothetical 2D con�nement results in an

e�ective entropic repulsion between surfactant molecules associated with the ssDNA chains.

Figure 2.7: Schematic illustration of the mechanism for a LC reorientation upon ssDNA adsorption andhybridization: ssDNA in bulk solution (a), adsorbed at the interface (zoomed in) (b), and after interfacialhybridization (c).

Figure 2.7 illustrates a schematic representation of the proposed mechanism. ssDNA in bulk solution

(Figure 2.7a) adopts a relatively globular con�guration with the hydrophobic bases partially shielded within

the molecular core. The negatively-charged ssDNA is attracted to the cationic surfactant at the aqueous/LC

interface via electrostatic interactions. As the ssDNA adsorbs to the interface (Figure 2.7b), short range

hydrophobic interactions in�uence the DNA to uncoil, allowing the hydrophobic bases to intercalate between

the surfactant molecules and interact with the hydrophobic LC subphase. The combination of electrostatic

and hydrophobic interactions induces a two-dimensional con�nement of ssDNA resulting in the increased

molecular area of the cationic surfactant proposed above. The radius of gyration of a polymer scales as Nν ,

whereN is degree of polymerization and ν is a scaling exponent with values of ν3D = 0.6 in 3D and ν2D = 0.75

in 2D for self-avoiding walks.[161] Therefore, a polymer strand is more compact in bulk solution than when

con�ned to an interface. A ssDNA molecule will initially adsorb to the interface in its more compact form,

but as it forms a complex with cationic surfactants it will tend to spread, increasing its interfacial area by a

factor of ~1.5. Due to the strong association of the ssDNA with the surfactants, this increase in molecular

area of the ssDNA translates to a similar increase in the molecular area of the surfactant. Therefore, when

32

the surfactant concentration is only slightly higher than the critical concentration required for homeotropic

orientation, this increase in surfactant molecular area serves to decrease the surface coverage below that

required for homeotropic orientation, resulting in the observed anchoring transition.

Upon interfacial DNA hybridization (Figure 2.7c), several phenomena occur that may contribute to

the observed LC reorientation. The persistence length of the DNA increases by a factor of ~140,[112, 36] the

hydrophobic bases of the DNA are no longer exposed, and the linear charge density of the DNA approximately

doubles. Each nucleobase of ssDNA prefers to bind to its complementary base than to remain intercalated

between the surfactant molecules, since hybridization involves not only hydrophobic interactions among the

base pairs but also hydrogen bonding and aromatic stacking. Therefore, the ssDNA strands reorganize at

the interface to hybridize with their complement; and once hybridized can be considered as rigid rods with

twice the linear charge density. This rigidity of the dsDNA strand allows for more e�cient packing of dsDNA

at the interface and its increased charge density may promote an increase in the local surface coverage of

cationic surfactants. Furthermore the hydrophobic bases that were exposed on the ssDNA strand are now

hidden within the core of the dsDNA helix, preventing intercalation of the bases between the surfactant

molecules via hydrophobic intercations. The combination of these phenomena allows for a reorganization

of the surfactants at the interface to the original surface coverage prior to ssDNA adsorption (or greater),

resulting in the observed anchoring transition from planar to homeotropic LC orientation.

2.6 Conclusions

Electrostatic and hydrophobic interactions were both found to be of critical importance to promote

the formation of a strongly associated ssDNA-surfactant complex. This ssDNA-surfactant complex is likely

con�ned to two dimensions resulting in an increase in the surfactant area per molecule. The increased

surfactant molecular area induces a LC anchoring transition from homeotropic to planar, as the LC orien-

tation is strongly a�ected by the surfactant coverage. A subsequent LC anchoring transition from planar

to homeotropic upon interfacial hybridization is explained by a reorganization of the surfactant at the in-

terface due to the dramatic changes in the DNA structure associated with hybridization. This proposed

mechanism for ssDNA adsorption and hybridization not only provides insight into the driving forces behind

33

the LC sensor described here but also has implications to other disciplines involving polymer � surfactant

interactions.

2.7 Acknowledgments

This work was supported by the Liquid Crystal Materials Research Center (NSF/MRSEC, Award No.

DMR-820579) and the Colorado State Bioscience Proof-of-Concept Grant (09BGF13). The authors would

also like to thank Dr. Kevin McCabe for help with dsDNA puri�cation.

2.8 Supplementary Information

2.8.1 dsDNA Preparation

Two reverse complementary ssDNA strands (5'TATTAGGGGATGAAGGGCACGAAGTTTTTTCT3';

5'AGAAAAAACTTCGTGCCCTTCATCCCCTAATA3'; Integrated DNA Technologies) were annealed by

adding equal parts into a microcentrifuge tube, heating to 95 ºC for 5 min and then slowly cooling to

room temperature for 2 hours. The annealed dsDNA (TM = 61.1 ºC) was then treated with exonuclease I

(New England Biolabs) suspended in 67 mM Glycine-KOH, 6.7 mM MgCl2, and 10 mM 2-Mercatoethanol

at pH~9.5 (New England Biolabs) for 1 hour to digest any residual ssDNA in solution into individual nu-

cleotides. The dsDNA solution was then puri�ed with a QIAquick nucleotide removal kit (Qiagen) to separate

the residual nucleotides and enzyme from the annealed dsDNA. The puri�ed dsDNA pellet was resuspended

in 10 mM Tris-Cl and 1 mM EDTA bu�er (Qiagen). An aliquot of puri�ed dsDNA was stained with ethid-

ium bromide, xylene cyanol, and bronopheonol blue then electrophoresis was run in a 15% acrylamide gel

for 1hr with bu�er conditions of 89 mM Tris-Cl, 89 mM Borate, 2mM EDTA bu�er (Sigma) to verify that

the DNA was in fact dsDNA. Furthermore, UV-vis analysis indicated that the puri�ed dsDNA was free of

any impurities. The remaining puri�ed dsDNA was then precipitated with ethanol and sodium acetate and

resuspended to 39 μM.

34

2.8.2 Flow Cell Assembly

In certain experiments a LC �lm was prepared in a �ow cell environment to facilitate an e�cient

bu�er exchange of the aqueous phase. Mixtures of surfactant in LC were prepared as described in the

main text. An electron microscopy grid was placed onto a 1� diameter circular glass cover slip (Electron

Microscopy Sciences). The desired mixture of surfactant in LC was then housed within the pores of the

electron microscopy grid as described in the main text and the entire glass cover slip was placed onto the

base of the �ow cell assembly. The base of the �ow cell assembly contains a circular opening, slightly less

than 1� in diameter, in which an o-ring (Kalrez, McMaster-Carr) of the proper diameter was placed. The

cover slip was then placed on top of this o-ring. Custom fabricated Te�on spacers (~240uM thick) designed

with a hole in the center was placed directly on top of the glass cover slip to create a gap between the cover

slip and the top glass (fused silica, Mark optics). The base of the �ow cell has channels built into it that allow

for one to introduce �ow through this gap created by the Te�on spacers. The top glass slide was fastened

into place by the top piece of the �ow cell assembly through which screws were used to clamp the entire �ow

cell assembly together. Bu�er was introduced through the inlet port using care to �ush out any air bubbles

that may be present in the cavity created by the Te�on spacers. Once assembly is complete, ssDNA was

introduced through the inlet port at the desired concentration and at a total load of ~250μL to ensure that

the entire volume in the �ow cell cavity was �ushed out. When a bu�er exchange of the aqueous phase was

conducted, excess volumes on the order of ~0.5-1mL of the bu�er were introduced through the inlet port to

ensure an e�cient bu�er exchange.

Chapter 3

Electrostatic Interactions In�uence Protein Adsorption � but not Desorption �

at the Silica-Aqueous Interface

3.1 Abstract

High-throughput single-molecule total internal re�ection �uorescent microscopy was used to inves-

tigate the e�ects of pH and ionic strength on bovine serum albumin (BSA) adsorption, desorption, and

interfacial di�usion at the aqueous � fused silica interface. At high pH and low ionic strength, negatively-

charged BSA adsorbed slowly to the negatively-charged fused silica surface. At low pH and low ionic strength,

where BSA was positively charged, or in solutions at higher ionic strength, adsorption was approximately

1,000 times faster. Interestingly, neither surface residence times nor the interfacial di�usion coe�cients

of BSA were in�uenced by pH or ionic strength. These �ndings suggested that adsorption kinetics were

dominated by energy barriers associated with electrostatic interactions, but once adsorbed, protein-surface

interactions were dominated by short-range non-electrostatic interactions. These results highlight the ability

of single-molecule techniques to isolate elementary processes (e.g. adsorption and desorption) under steady

state conditions, which would be impossible to measure using ensemble-averaging methods.

Keywords: Electrostatic, hydrophobic, van der Waals interactions, DLVO, TIRF, single molecule

3.2 Introduction

The dynamic behavior of proteins at solid-liquid interfaces is critically important to various areas

of biotechnology and biomaterials,[77] including biopharmaceuticals,[12] vaccines,[159] bioseparations[130]

36

and biosensing.[174] Silica (i.e., glass) surfaces are particularly relevant to pharmaceutical, chromatographic

and sensing technologies. Due to their complex molecular structure, proteins interact with surfaces via a

combination of long-range (e.g. electrostatic) and short-range non-covalent (van der Waals, hydrophobic,

hydrogen bonding) interactions.[162] As the only interaction that can be repulsive, electrostatic e�ects have

received particular attention, since they can be controlled (using pH and ionic strength) and exploited to

stabilize protein solutions against aggregation[139, 29] or to reduce the rate of interfacial adsorption.[162, 121]

While many models have been developed to describe the combined e�ects of surface-protein interactions,[77,

134, 177] the typical experiments (surface plasmon resonance, quartz crystal microbalance, etc.) that are

used to study dynamic behavior of proteins at interfaces rely on ensemble-averaging, and thus measure

the net e�ect of many elementary processes (adsorption, desorption, surface-mediated protein associations,

conformational changes, etc.), making it di�cult to isolate the in�uence of speci�c environmental conditions

on each elementary process. Even the apparently simple process of adsorption is di�cult to measure directly,

since rates determined using transient ensemble-average measurements actually re�ect a net adsorption rate

involving both elementary adsorption and desorption rates (the latter of which depends on surface coverage).

While these elementary rates can be extracted by assuming speci�c theoretical kinetic models,[126] it is

desirable to measure the elementary rates directly, preferably under steady-state conditions. Single-molecule

total internal re�ection �uorescence microscopy (SM-TIRFM) provides this capability, permitting direct

observations of individual molecular processes such as adsorption, interfacial di�usion, and desorption.[77, 66]

In particular, by separating net adsorption into its elementary components � adsorption and desorption � the

e�ects of electrostatic interactions (a long-range interaction) can be disentangled from those of short-range

interactions.

Proteins are formed from diverse amino acids. Their topologically complex surfaces are composed of

amino acid residues that may be charged (positively or negatively), uncharged but polar, or hydrophobic.[82,

142] Proteins often display multiple positively and negatively charged surface patches. Since these charged

groups are acidic or basic, they may be neutralized depending on the local environment and the pH of the

environment. To �rst order, the overall electrostatic character of a protein can be inferred by measuring the

isoelectric point (pI), which is the pH at which the protein's net charge is equal to zero.[142] In solutions with

37

pH that is above the pI, protein molecules exhibit a net negative charge, while the net charge is positive below

the pI. However, bu�er choice, salt concentration, and protein composition all a�ect the actual magnitude

of the charge at a given pH.[142]

Using SM-TIRFM, we investigated the interfacial dynamics of bovine serum albumin (BSA) at the

aqueous � fused silica (FS) interface as a function of pH and ionic strength. Under these conditions we mea-

sured adsorption rates, residence times, and interfacial di�usivity of large numbers of individual molecules

for each experimental condition. Under the conditions of these experiments, the protein and surface had

opposite charges at low pH and like charges at high pH, so we expected to measure the e�ects of electrostatic

repulsion and attraction on molecular adsorption, which should be sensitive to long-range protein-surface in-

teractions. By comparing the dynamics as a function of ionic strength, the in�uence of charge-screening could

be characterized. Moreover, these experiments enabled us to explore the in�uence of electrostatic interac-

tions on desorption and interfacial di�usion, phenomena that are dominated by the short-range interactions

of proteins and surfaces in intimate contact.

3.3 Results and Discussion

Steady-state adsorption rate coe�cients, kads, for �uorescently-labeled BSA were calculated from

single-molecule TIRFM. These adsorption rate constants represented the absolute adsorption rate, as opposed

to the net adsorption rate, which is uniquely measured using single-molecule observations. At pH values

below,[174] when the protein and surface had opposite net charges, fast adsorption was observed, with kads

> 1000 nm/s (Figure 3.1). Intuitively, an adsorption rate coe�cient of 1000 nm/s implies that the number

of solute molecules in a 1000 nm thick slab of solution adsorb the surface each second. Similar values of kads

were observed previously for a �uorescently labeled fatty acid on a hydrophobic surface at both low and high

ionic strength,[121] suggesting that this order of magnitude of kads re�ects conditions where strong repulsion

is absent. At pH values above 4.7, where both the surface and protein were expected to exhibit negative

net charges, kads decreased sharply by several orders of magnitude, reaching a value of 2.9±0.3 nm/s at

pH=7.4 (Figure 3.1). Under conditions of high ionic strength (100 mM NaCl was added to the solution),

as indicated by the open symbols in Figure 3.1, kads at pH=2.6 remained similar to the values measured

38

under low ionic strength conditions. However, in solutions with pH=7.4, kads increased almost 1000 fold, to

2000±200 nm/s, compared to the kads measured at low ionic strength.

Figure 3.1: Adsorption rate coe�cients of BSA at FS surface as a function of pH in 10 mM CP (closed circles)and 10 mM CP with 100 mM NaCl (open diamonds) obtained from single-molecule adsorption observationsmade using TIRF. Error bars in plot represent the standard deviation between three replicate experiments.

The isoelectric point (pI) of BSA is at 4.7.[142] In solutions with pH < pI, the net charge of BSA is

expected to be positive, whereas the net charge is expected to be negative for pH > pI.[142] The surface

charge of FS is expected to be negatively charged above pH≈ 2,[113] i.e. for all the conditions explored here.

Under the low ionic strength conditions associated with 10 mM citric acid � sodium phosphate bu�er (CP),

the e�ective Debye length is ~2.5 nm, and strong electrostatic interactions are expected between BSA and

FS. In particular, for pH>pI, DLVO theory suggests that electrostatic repulsion provides a strong energetic

barrier to adsorption, overwhelming the short-range van der Waals attraction that becomes dominant at

close contact, resulting in a barrier to adsorption of roughly 10�30 kT.[22] This is a reasonable explanation

for the dramatic decrease in the adsorption rate at high pH and have been observed in net adsorption rate

measurements.[30] However, under the higher ionic strength conditions associated with the addition of 100

mM NaCl, the Debye length for the system decreased to 0.65 nm, and electrostatic interactions between

BSA and FS are therefore greatly weakened relative to van der Waals attraction, resulting in a much lower

barrier to adsorption on the order of kT. As a result, under high ionic strength conditions, the adsorption

rate remained high even for pH>pI, when surface and protein had like charges. These observations were

consistent with an adsorption rate that was dominated by electrostatic interactions within the context of

electrical double-layer (DLVO) theory.[64]

39

Interestingly, for pH<pI, when protein and surface exhibit opposite charges, it was evident that

electrostatic attraction did not signi�cantly enhance adsorption to the interface, as demonstrated by the

similarity of kads between low and high ionic strength conditions at 2.6 pH (Figure 3.1). At �rst glance, this

may appear counter-intuitive, since one might expect an increase in bulk transport towards the surface due

to electrostatic attraction. However, this is likely a small e�ect compared to di�usive transport, especially

given the fact that the Debye length is only 2.5 nm, and the adsorption kinetics under these conditions were

presumably limited by other energetic barriers to adsorption. For example, it has been postulated that an

intrinsic barrier to adsorption involves the energy required for a solute to displace the hydration shell formed

at the near surface,[162, 121, 113] and we have previously measured an activation barrier to adsorption that

is associated with the energy required to �unbind� solvent molecules.[67, 65]

While the adsorption rate was expected to be sensitive to long-range interactions, we hypothesized that

the desorption rate, and the interfacial di�usion coe�cient, would be dominated by short-range interactions,

since these behaviors occur under conditions when the protein and surface are in close contact. As described

above, cumulative surface residence time distributions (see Figure 3.4 in the Supporting Information) were

analyzed using a three population exponential mixture model to extract characteristic residence times, τi,

of each population, and their respective fractions. These data are summarized in Figure 3.2 and all �tting

parameters are given in Table 3.1 in the Supporting Information. In principle, it might be possible to assign

a population associated with a given term in a mixture model to a speci�c physical population, which could

be associated with a molecular conformation, an aggregation state, or even a particular type of surface site.

However, the actual data obtained in these experiments do not enable us to make these speci�c assignments.

The mixture model did, however, provide a quantitative description of the data and permit the accurate

calculation of mean values.

As shown in Figure 3.2a, the average surface residence time exhibited only a very subtle dependence

on pH, and only between pH values of 5.7 and 7.4. Moreover, looking more closely at parameters associated

with the individual populations, no systematic trend was observed with pH for any of the characteristic

residence times or fractions. These results support the notion that under conditions of close protein-surface

contact (i.e. when a protein is adsorbed), interactions are presumably dominated by short-range non-covalent

40

interactions, as opposed to electrostatic forces. The e�ect of surface hydrophobicity on protein desorption,

for example, was reported in previous work.[76] This suggests that the kinetics of desorption should be

relatively insensitive to environmental conditions that primarily a�ect electrostatic interactions.

Figure 3.2: (a) Average surface residence times of BSA at the FS surface as a function of pH in 10 mM

CP. (b) Characteristic residence times, τ , each of the three populations identi�ed by analysis of cumulative

surface residence time distributions (closed boxes, open diamonds, and closed circles, respectively). (c)

Population fractions associated with each of the characteristic residence times shown in panel (b). The error

bars represent the standard deviation between three replicate experiments.

BSA interfacial di�usion was characterized by analyzing the cumulative squared displacement distri-

butions (see Figure 3.5 in the Supporting Information) using a Gaussian mixture model to represent the

presence of multiple di�usive modes. These data are summarized in Figure 3.3 and all �tting parameters

are given in Table 3.2 in the Supporting Information. The presence of multiple di�usive modes does not

41

necessarily suggest that individual molecular populations exhibit distinct di�usive modes. In fact, we have

generally observed that individual molecular trajectories may switch between di�usive modes. This is likely

due to conformational changes, changes in molecular orientation with respect to the surface, and/or a change

between �crawling� and ��ying� modes of di�usion.[76, 148] The mean apparent di�usion coe�cients, D, were

calculated using a weighted average of the various di�usive modes. As shown in Figure 3.3, the mean di�usion

coe�cients show no systematic trend, consistent with the behavior of the surface residence times shown in

Figure 3.2. Again, this supports the notion that non-electrostatic short-range forces dominate the behavior

of BSA proteins once adsorbed to the FS surface.

Previous work by our group has connected interfacial di�usion to desorption and adsorption processes

through intermittent desorption-mediated ��ights�.[148] In these experiments, as described above, the rate of

desorption was only weakly dependent on pH; however, the adsorption rate was sensitive to these parameters.

While in principle the interfacial di�usion should be related to both processes, we have often observed that

the overall magnitude of the interfacial di�usion coe�cient is primarily determined by the distribution of

�waiting times� between �ights (i.e. the desorption rate). This is clearly the case here, as the fractions

associated with the extremely slow modes 1 and 2 dominated the di�usive motion.

42

Figure 3.3: (a) Mean di�usion coe�cients, D, of BSA at FS a surface as a function of pH in 10 mM CP.

Error bar in both plots represent the standard deviation between three replicate experiments. (b) Di�usion

coe�cients, D, each of the four populations identi�ed by analysis of cumulative squared displacement dis-

tributions (closed boxes, open diamonds, closed circles, and closed diamonds, respectively). (c) Population

fractions associated with each of the di�usion coe�cients shown in panel (b). The error bars represent the

standard deviation between three replicate experiments.

3.4 Conclusions

Using high-throughput single-molecule tracking methods, we isolated elementary processes associated

with BSA adsorption to and desorption from fused silica surfaces under steady-state conditions. BSA

adsorption rates on fused silica were dramatically in�uenced by electrostatic forces, slowing by three orders

of magnitude under pH conditions where protein and surface had like charges. This e�ect was greatly

diminished at high ionic strength, supporting the interpretation in terms of electrostatic e�ects. Adsorption

rates were not apparently increased by like charge attraction, suggesting that other short-range barriers to

43

adsorption are the limiting factor under these conditions. Interestingly, desorption and interfacial di�usion,

dynamic molecular behaviors that occur under conditions of close protein-surface proximity, were una�ected

by pH or ionic strength, suggesting that these phenomena are primarily in�uenced by non-electrostatic

short-range interactions. While consistent with the predictions of conventional DLVO theory, which describe

colloidal interactions, these �ndings contradict the intuitive notion that surfaces with opposite charge from

an adsorbing protein will retain proteins more strongly than neutral or like-charged surfaces.


BSA conjugated with Alexa Fluor 555 (5 labels/molecule, as reported by the vendor) was purchased

from Invitrogen (CAS A34786). Universal bu�er solutions were composed of 10 mM CP. Five separate

bu�ers for 2.6, 3.7, 4.7, 5.7, and 7.4 pH were prepared by adjusting the concentrations of citric acid and

sodium phosphate for the desired pH as described by Dawson and coworkers.[37]

For TIRFM experiments, the pH of the bu�er was adjusted by hydrochloric acid and sodium hydroxide

to the desired pH. Fluorescently-labeled BSA was diluted with CP to a concentration range of 10−4 � 10−6

mg/ml. Fused silica (FS) coverslips were cleaned with cationic detergent (micro-90, International Product

Corp.) and rinsed with 18.2 MΩ-cm water . Coverslips were then incubated in warm piranha solution for

one hour and subsequently treated in a UV-ozone chamber for one hour. Once cleaned, a custom-made

Te�on® ring (0.5 cm inner diameter, 1.5 cm outer diameter, 0.8 cm height) was placed in contact with the

coverslip, creating a well to contain a small volume of bu�er. 100 μl of bu�er �uorescently labeled BSA at

a concentration of 10−6 mg/mL in solutions with a pH of 2.6, 3.7, or 4.7 pH, and a concentration of 10−4

mg/ml in solutions with a pH of 5.7 and 7.4, were then added to the well and allowed to relax. Once stable,

a sequence of images was captured using TIRFM.

TIRFM experiments were performed using a Nikon Eclipse TI-93 out�tted with a custom through-

the-objective TIRF illuminator used in conjunction with a 100x oil immersion objective. A cooled CCD

camera (Photometrics Cascade 512B) operating at �80°C was used to capture a sequences of images with

an acquisition time of 200 ms. A Cobolt Samba laser emitting at 532 nm was used as an excitation source;

movies were continuously captured for 3 minute durations. Movies were analyzed using a custom-written

44

molecule identi�cation and tracking algorithm;[80, 165] approximately 103 to 5x104 identi�ed objects were

tracked per experiment. Objects that were present for one or two frames were ignored because of the

high error these short trajectories exhibit due to noise within the image. For each movie, the adsorption

rate coe�cient, kads, was calculated using the equation kads = n/Δt a cbulk where n was the number of

molecules that adsorbed during the course of the movie, ∆t was the time duration of the movie, a was

the area of the �eld of view, and cbulk was the bulk concentration. The mean adsorption rate coe�cient

was calculated from three separate movies and the reported error was the standard deviation between these

measurements. Characteristic surface residence times, τ , were calculated by �tting the cumulative probability

distribution C(t) of molecular residence times to an exponential mixture model comprised of three exponential

functions, C(t) =∑3i=1 aie

−t/τi , where ai was the population fraction for population i and τi was the

characteristic surface residence time for population i.[80, 165] The mean surface residence time was calculated

as the weighted average of these populations, using the expression τ =∑i ai τi. Di�usion coe�cients were

calculated by �tting the cumulative squared displacement distribution, C(r2, t), to a Gaussian mixture model,

comprised of four Gaussian functions, C(r2, t) =∑4i=1 bie

−x/Di , where bi was the fraction associated with

di�usive mode i, Di was the e�ective di�usion coe�cient of di�usive mode i, and x = r2/4∆t was the time

step (∆t) corrected squared displacement (r2).[80, 165] The mean di�usion coe�cient was calculated as the

weighted average of these modes, using the expression D =∑i biDi.

3.6 Acknowledgments

The authors gratefully acknowledge support from the National Science Foundation award #CBET-

1133871. Aaron McUmber would also like to thank Dr. Blake Langdon for many helpful conversations about

experimental design and analysis.

45

3.7 Supporting Information

3.7.1 Additional Materials and Methods

For circular dichroism (CD) measurements, lyophilized BSA (Aldrich CAS A2153) was reconstituted to

a concentration of 10 uM in 10mM CP. CD spectra were measured using a Chirascan�-plus CD spectrometer

(AppliedPhotophysics). The instrument and sample chamber were purged under a nitrogen stream for several

hours before experiments were performed. Quartz cuvettes with a volume of ~170 μl and 1 mm path length

were used to measure the CD signal from 190�260 nm with 0.5 nm steps, for 0.5 seconds per step at 20° C.

Each sample was measured 10 times and the resulting CD spectra were averaged and smoothed between each

two adjacent points. BSA CD spectra were then normalized by subtracting the CD spectra measured with

10mM CP at the respective pH values. Each experimental condition was measured three separate times.

3.7.2 Additional Results

Cumulative probabilities of surface residence times for �uorescently labeled BSA on FS are shown

in Figure 3.4. Characteristic surface residence times were calculated by �tting the cumulative probabilities

to a three population exponential mixture model discussed in the text are shown in Table 3.1. The data

presented in Table 3.1 are presented in the text as Figure 3.2.

Figure 3.4: Cumulative residence time distributions of �uorescently labeled BSA on FS for 2.6 (red circles),3.7 (yellow triangles), 4.7 (green squares), 5.7 (open diamonds), 7.4 (black triangles). The curves representthe mean between three replicate experiments. Error bars represent 65% con�dence expected from Poissonstatistics. For clarity, the graphs have been o�set vertically.

46

Table 3.1: Population fractions and characteristic residence time �t from Figure 3.1 for each pH.

Population 1 2 3pH a1 τ1(s) a2 τ2(s) a3 τ3(s)2.6 0.53(8 0.39(7) 0.35(4) 1.9(7) 0.10(3) 9(4)3.7 0.5(1) 0.4(1) 0.36(8) 1.8(6) 0.10(3) 7(3)4.7 0.62(2) 0.47(3) 0.33(1) 2.3(2) 0.05(1) 9.0(3)5.7 0.46(5) 0.39(4) 0.39(3) 1.6(2) 0.15(2) 4.9(5)7.4 0.64(8) 0.36(7) 0.40(7) 1.9(3) 0.06(4) 7(4)

Cumulative squared displacement distributions for �uorescently labeled BSA on FS are shown in

Figure 3.5. Di�usion coe�cients were calculated by �tting the cumulative probabilities to four population

Gaussian mixture model discussed in the text are shown in Table 3.2. The data presented in Table 3.2 are

presented in the text as Figure 3.3.

Table 3.2: Population fractions and di�usion coe�cients �ts from Figure 3.2 for each pH.

Population 1 2 3 4pH b1 D1(µm2/s) b2 D2(µm2/s) b3 D3(µm2/s) b4 D4(µm2/s)2.6 0.2(1) 0.0011(3) 0.61(1) 0.0058(5) 0.21(7) 0.015(3) 0.03(2) 0.38(9)3.7 0.10(4) 0.0009(1) 0.49(3) 0.0051(6) 0.38(4) 0.014(3) 0.03(2) 11(2)4.7 0.15(3) 0.00095(6) 0.60(4) 0.0052(3) 0.17(2) 0.025(5) 0.07(2) 17(1)5.7 0.10(1) 0.00105(3) 0.63(2) 0.0052(3) 0.20(2) 0.020(1) 0.07(1) 149(6)7.4 0.10(8) 0.0010(2) 0.64(3) 0.0054(2) 0.23(4) 0.0162(4) 0.04(1) 31(3)

Circular dichroism (CD) spectroscopy measurements were conducted using unlabeled BSA at 2.6, 4.7,

and 7.4 pH, shown in Figure 3.6. BSA structure appears to remain mostly unchanged between 7.4 and 4.7

pH conditions, but partial loss of secondary structure is apparent at 2.6 pH. This is indicated by the decrease

in CD signal at 191 and 210 nm indicating a loss of α-helical structure, though not as dramatic as a fully

denatured BSA indicated by the red line. Measuring the change in peak intensity at 191 nm, BSA retains

63±1% of its structure, assuming that BSA loses all of its α-helical structure at 75°C. Despite the partially

unfolded state observed at 2.6 pH, no dramatic changes were observed between 4.7 and 2.6 pH for adsorption

rate, residence time, or di�usion, suggesting that BSA interfacial dynamics are largely una�ected by these

changes in secondary structure.

47

Figure 3.5: Cumulative squared displacement distributions of �uorescently labeled BSA on FS for 2.6 (redcircles), 3.7 (yellow triangles), 4.7 (green squares), 5.7 (open diamonds), 7.4 (black triangles). The curvesrepresent the mean between three replicate experiments. Error bars represent 65% con�dence expected fromPoisson statistics. For clarity, the graphs have been o�set vertically.

Figure 3.6: Representative CD spectra of BSA in 10 mM CP normalized to CD spectra of 10 mM CP ateach respective pH. The black solid line represents BSA data captured at 2.6 pH, the dotted line at 4.7 pH,and the dashed line at 7.4 pH. The data in black are measurements captured at 20°C. The red solid linerepresents BSA data captured at 7.5 pH at 75°C.

Chapter 4

Molecular Trajectories Provide Signatures of Protein Clustering and Crowding

at the Oil/Water Interface

4.1 Abstract

Using high throughput single-molecule total internal re�ection �uorescence microscopy (TIRFM), we

have acquired molecular trajectories of bovine serum albumin (BSA) and hen egg-white lysozyme during

protein layer formation at the silicone oil-water interface. These trajectories were analyzed to determine the

distribution of molecular di�usion coe�cients, and for signatures of molecular crowding/caging, including

subdi�usive motion and temporal anticorrelation of the instantaneous velocity vector. The evolution of these

properties with aging time of the interface was compared with dynamic interfacial tension measurements.

For both lysozyme and BSA, we observed an overall slowing of protein objects, the onset of both subdi�usive

and anticorrelated motion (associated with crowding), and a decrease in the interfacial tension with aging

time. For lysozyme, all of these phenomena occurred virtually simultaneously, consistent with a homogeneous

model of layer formation that involves gradual crowding of weakly interacting proteins. For BSA, however,

the slowing occurred �rst, followed by the signatures of crowding/caging, followed by a decrease in interfacial

tension, consistent with a heterogeneous model of layer formation involving the formation of protein clusters.

The application of micro-rheological methods to single molecule trajectories described here provides an

unprecedented level of mechanistic interpretation of interfacial events that occurred over a wide range of

interfacial protein coverage.

Keywords: Aggregation, interface, subdi�usion, caging, molecular con�nement, 2D anomalous di�u-

49

sion

4.2 Introduction

Protein aggregation at interfaces and the formation of interfacial protein layers represent impor-

tant phenomena for several areas of biotechnology and biomaterials.[77] For example, fouling of biosensing

devices[174] or �ltration membranes,[130] aggregation of protein molecules in therapeutic formulations,[147]

and undesirable immune responses to implanted devices[59] have all been related to the formation of ad-

sorbed protein layers. Aggregation of protein molecules at silicone oil-water interfaces speci�cally, has

been the subject of increasing scrutiny within the biopharmaceutical industry.[51, 12, 109, 153, 39] Silicone

oil-water interfaces are now ubiquitous due to the popular packaging strategy of using pre�lled glass sy-

ringes as storage and delivery devices, which require the use of silicone oil as a lubricant. Studies have

indicated that an increased incidence of protein aggregation may be related to the presence of silicone oil-

water interfaces.[109, 153, 10, 27, 73] From a fundamental perspective, liquid-liquid interfaces, such as the

silicone oil-water interface, represent an attractive system to study the mechanisms that lead to protein

layer formation, in part because they exhibit excellent spatial homogeneity, which virtually eliminates con-

founding e�ects due to isolated surface defects (a.k.a. strong binding sites) inevitably found on solid-liquid

interfaces.[66, 149, 76]

The formation of protein layers at �uid interfaces is readily characterized by macroscopic phenom-

ena that involve thermodynamic (e.g. interfacial tension) or dynamic/rheological (e.g. changes in inter-

facial mobility) properties. Historically, dynamic interfacial tensiometry (D-IFT)[54, 14, 110, 116] has

been the standard approach for studying interfacial adsorption at liquid-liquid interfaces, while interfa-

cial rheology[53, 4, 23] has been used to characterize well-developed protein layers that display measurable

viscoelastic properties. Additional techniques such as �uorescence-activated cell sorting[105] and front-face

�uorescence spectroscopy have been applied to analysis of proteins adsorbed to liquid-liquid interfaces in high

surface-area systems (e.g. emulsions).[178, 81] Interestingly, the correlation between the temporal evolution

of thermodynamic and hydrodynamic phenomena is highly variable and poorly understood. For example,

one can de�ne �lag times� to provide an approximate description of the interfacial aging time required to

50

measure a signi�cant change in the interfacial tension (tIFT ) or interfacial di�usivity (tdif ). Here we de-

scribe the behavior of lysozyme at silicone oil-water interfaces, where a reduction of interfacial tension follows

quickly behind the reduction in interfacial di�usivity (tIFT ≈ 2tdif ), and the contrasting behavior of bovine

serum albumin (BSA) at silicone oil-water interfaces, where interfacial mobility decreases long before the

interfacial tension begins to drop (tIFT > 10tdif ).

While the physical interpretation of interfacial tension is complex for interfaces laden with proteins or

colloidal particles,[54, 53, 129, 160] nevertheless, a signi�cant reduction in interfacial tension is general related

to large coverage of interfaces with surface-active molecules. For example, for Langmuir monolayers of poorly

soluble fatty acids or phospholipids, which have been thoroughly characterized, a signi�cant reduction of

interfacial tension (e.g. by ~10%) generally requires a fractional surface coverage of 0.3-0.5.[62, 2, 118, 90, 125]

So intuitively, one can expect that a signi�cant reduction of interfacial tension will be observed for an interface

that is approaching its upper limit of surface coverage. In this context, the two lag time regimes described

above correspond to situations where the mobility begins to decrease when the surface is already highly

crowded (lysozyme) or when the interfacial protein coverage is still very sparse (BSA).

It has been proposed that interfacial �slowing down� may be caused by two distinct mechanisms, which

can potentially act in concert.[168] The so-called homogeneous (or two-dimensional) mechanism involves a

reduction in molecular mobility that is primarily due to drag experienced by individual proteins due to

collisions with other adsorbed proteins, i.e. interfacial crowding. On the other hand, a heterogeneous

mechanism for interfacial slowing involves a reduction in mobility caused by increased viscous drag from

the adjacent bulk phases; this is hypothetically due to the formation of interfacial clusters/aggregates that

exhibit large hydrodynamic radii. It is reasonable to hypothesize that since homogeneous slowing can occur

only under crowded interfacial conditions where the surface coverage is relatively high, homogeneous slowing

should be correlated with the reduction of interfacial tension. However, heterogeneous slowing, which relies

upon the formation of molecular clusters/aggregates, can potentially occur even under conditions of relatively

low surface coverage.

While these are reasonable hypotheses, it is clear that macroscopic, ensemble-averaging measurements

cannot distinguish directly between these mechanisms for interfacial protein layer formation. However, the

51

recent application of single-molecule methods such as �uorescence correlation spectroscopy (FCS)[47, 48, 91]

and TIRF[155, 55] has directly demonstrated the ability to identify interfacial heterogeneities in the form of

multiple molecular populations (e.g., based on oligomerization or conformational state).[66, 76, 167, 152, 120]

TIRF, in particular, explicitly separates distinct dynamic mechanisms and provides detailed information

about interfacial mobility and adsorption kinetics, including the distribution of interfacial di�usion coe�-

cients. In addition, TIRF measurements may be used to decouple net adsorption into individual adsorption

and desorption events and to probe step-size distributions of molecular trajectories.[66, 149, 76, 152, 166]

Donsmark and coworkers used FCS to show that the dynamic behavior of insulin[47] and beta-

lactoglobulin[48] slowed signi�cantly with aging time at the oil-water interface, and that the details of

this slowing were highly dependent on the identity of the protein. Walder et al.[167] applied a molecular

tracking approach to study the slowing-down behavior of BSA at an oil-water interface. They found that

the distribution of individual di�usion coe�cients broadened signi�cantly as the interface aged, consistent

with a heterogeneous mechanism for layer formation. They supported this interpretation by comparing the

experimental observation with the results of a population-balance model for protein cluster formation. These

results provided support for the importance of protein-protein interactions, and cluster formation, as part

of interfacial protein layer formation. However, the analysis of di�usion coe�cient distributions represented

a limited approach for several reasons. For example, these distributions are sensitive primarily to the very

early stages of layer formation, precluding direct comparisons with the kinetics associated with interfacial

tension reduction. Moreover, the distribution of individual di�usion coe�cients need not necessarily broaden

signi�cantly, depending on the detailed mechanism of cluster formation, and importantly, these distributions

fail to provide direct information about interfacial crowding.

To address these limitations, here we apply more sophisticated analysis tools, commonly applied in

micro-rheological studies of colloidal systems,[133] to the interfacial trajectories of individual protein objects.

These molecular trajectories provide detailed information about the environment experienced by the molec-

ular objects, including the presence of crowding or �caging� e�ects.[133] Regular unconstrained Brownian

motion is characterized by the expression⟨r2(τ)

⟩= 4Dτ , where the mean square displacement (MSD) ex-

hibits a linear dependence on the time interval, τ, over which it is measured. In general, anomalous di�usion

52

can be described using an ad hoc power law expression law such that:⟨r2⟩

= 2dKτα, where⟨r2⟩is the MSD,

d is the dimensionality of the system (d = 2 in the case of interfacial di�usion), K is the mobility coe�cient,

and α is the power law exponent. This equation corresponds to simple Brownian motion in the limit that

α = 1, in which case K corresponds to the Brownian di�usion coe�cient D.[91] Values of α < 1 are char-

acteristic of trajectories that are called subdi�usive; this behavior is commonly observed for motion within

crowded environments or in the presence of obstacles.[133] Values of α > 1 are less commonly observed, for

trajectories that are called superdi�usive. It is widely observed that di�usion in crowded environments devi-

ates from traditional Brownian motion, becoming subdi�usive; this can be observed through careful analysis

of molecular mean squared displacement versus time interval.[91] Moreover, di�usion in structured and/or

crowded environment can become history-dependent, or non-Markovian, which may result in temporal anti-

correlation of the instantaneous velocity vector.[91] For example, previous experimental and computational

studies have connected temporal velocity anticorrelation to �caging� e�ects in glassy systems,[172] motion in

viscoelastic media[40] (e.g. for the fractional Brownian motion model[21]) or dynamic heterogeneity.[61] In

each of these systems, anticorrelated velocity is directly related to highly crowded conditions, typically with

volume fractions >0.2. Compared with colloidal systems which can be imaged with extraordinarily high

time resolution and contrast, this type of analysis is challenging in the context of single-molecule tracking

which intrinsically provides low signal/background data. However, by performing careful statistical analyses,

we are able to directly identify clear signatures of the onset molecular crowding in the form of correlated

subdi�usive trajectories and temporally anti-correlated velocities for lysozyme and BSA layer formation at

the oil-water interface. Lysozyme and BSA have previously been shown to exhibit distinctly di�erent rates

of interfacial layer formation.[4] For lysozyme, these signatures of crowding appear at aging times similar

to those where both the interfacial mobility and interfacial tension begin to decrease, consistent with a ho-

mogeneous mechanism for layer formation. In contrast, for BSA the signatures of crowding/caging appear

long after the interfacial mobility decreases, but before the reduction of interfacial tension, consistent with

a heterogeneous mechanism (involving protein cluster formation).

53

4.3 Materials & Methods

Hen egg white lysozyme (Aldrich CAS 12650-88-3) was labeled with Alexa Fluor 555 succinimidyl

ester using Molecular Probes microscale protein labeling kit (Invitrogen CAS A30007). The average labeling

e�ciency was found to be 3±1 �uorophores per protein using UV-visible spectroscopy at 280 and 555 nm

(data not shown). Bovine serum albumin (BSA) conjugated with Alexa Fluor 555 (5 labels/molecule, as

reported by the vendor) was purchased from Invitrogen (CAS A34786). All lysozyme and BSA (Aldrich CAS

A2153) solutions were made using 1x PBS bu�er (Gibco CAS 10010-23), pH 7.4 with 0.1 mg/ml sodium

azide (Aldrich CAS 438456).

TIRFM experiments were performed using a Nikon Eclipse TI-93 out�tted with a custom through-

the-objective TIRF illuminator used in conjunction with a 100x oil immersion objective. A cooled CCD

camera (Photometrics Cascade 512B) operating at -80°C was used to capture a sequences of images with

a typical acquisition time of 200 ms. A Cobolt Samba laser emitting at 532 nm was used as an excitation

source; movies were continuously captured for short-time experiments and movies of 3 minute duration were

captured at various time intervals for longer time-scale experiments.

For TIRFM experiments, glass coverslips (Fisher CAS 12-542-C) were cleaned using a cationic surfac-

tant (Micro-90 International Products M9031-12), Millipore �ltered water with a resistance of 18.2 MΩ-cm,

and isopropanol (Aldrich W292907), and were then dried under a nitrogen stream. Silicone oil droplets with

a viscosity of 350 cSt (Dow Corning 360 CAS 01472666) were added to the clean coverslips and stabilized

using a nickel TEM grid, ensuring a stable planar interface between the silicone oil and bu�er.[167, 115, 17]

A Te�on® ring was then placed in contact with the coverslip, surrounding the silicone oil-�lled TEM grid,

creating a well to contain a small volume of bu�er. 100 μl of bu�er containing either 7.0x 10−2 nM of

Alexa555-labeled lysozyme or 1.5x 10−3 nM of Alexa555-labeled BSA were then added to the well and al-

lowed to reach steady state. Once stable, a sequence of images was captured, and then 100 μl of 140 nM

of unlabeled protein, doped with either 7.0x 10−2 nM of Alexa555-labeled lysozyme or 1.5x 10−3 nM of

Alexa555-labeled BSA, were added to the well, leaving a �nal concentration of 70 nM of unlabeled pro-

tein. Following addition of unlabeled protein, a series of 3-minute, time-lapse TIRF movies was captured at

54

periodic time intervals during the experiment. TIRF movies were analyzed using a custom-made molecule

identi�cation and tracking algorithm.[80, 165] Data from experiments involving only a very low concentration

of labeled protein in the absence of unlabeled protein (i.e. low surface-density experiments) were analyzed to

identify the surface residence times and di�usion coe�cients for individual molecules and small, well-de�ned

oligomers, as reported previously for BSA.40 Time-lapse data from experiments that included various concen-

trations of unlabeled protein (i.e. high surface-density experiments) were obtained in triplicate and analyzed

by binning the movies into 60 s and 20 s time segments for lysozyme and BSA, respectively, and considering

each time segment as one time point. Within each time bin approximately 500 � 1,000 identi�ed objects were

tracked and an e�ective di�usion coe�cient was calculated for each individual object. The mean squared

displacement versus time interval,⟨r2(τ)

⟩, was calculated for each object, and �t using the expression⟨

r2(τ)⟩

= 4Dτ . This allowed us to determine the distribution of di�usion coe�cients and the mean di�usion

coe�cient, as a function of aging time. The mean e�ective di�usion coe�cient was calculated by taking the

arithmetic mean of di�usion coe�cients for each molecule within binned time segments. Hydrodynamic radii

were calculated from the di�usion coe�cients found by applying the Stokes-Einstein equation for di�usion

of a disc at the interface between two viscous liquids using the equation: D = 3kbT/16R(ηoil+ηwater) where

D, T , kb, R, and η represent the di�usion coe�cient, temperature, Boltzmann constant, hydrodynamic

radius, and viscosity, respectively.[137] Relative mean di�usion coe�cients were �rst calculated by dividing

the mean di�usion coe�cient of objects within a time bin by the di�usion coe�cients for objects in the �rst

time bin (t = 0). Then the arithmetic mean was calculated for relative di�usion coe�cients calculated from

each triplicate measurement. Reported error bars were calculated as the standard deviation from the mean

of the experimental triplicates.

To determine the anomalous power law exponent associated with di�usive motion, trajectories were

accumulated as a function of aging time, and the mean squared displacement,⟨r2(τ)

⟩, versus time interval

was calculated for both BSA and lysozyme. In order to minimize e�ects from possible immobilized objects

and short trajectories, MSD values were calculated for molecules that were present on the interface for

greater than 10 frames (2 s) and that displayed di�usion coe�cients within two standard deviations from

the mean di�usion coe�cient or that had di�usion coe�cients greater than 0.001 µm2/s. For each time

55

bin, all squared displacements associated with a given time interval were obtained, and averaged using equal

weights. These data were �tted to the anomalous di�usion expression,⟨r2⟩

= 2dKτα, where⟨r2⟩is the

MSD, d is the dimensionality of the system (d = 2 in the case of interfacial di�usion), K is the mobility

coe�cient, and α is the power law exponent that was used for subsequent analysis.

The velocity-velocity autocorrelation function, G(τ) = 〈v(t)�v(t+ τ)〉 /⟨v(t)2

⟩, was also calculated,

where v(t) represents the instantaneous velocity vector at time t. Similar to the MSD calculations, e�ects

from possible immobilized objects and short trajectories were minimized by calculating values based on

molecules that existed on the interface for greater than 10 frames (2 s) and displayed di�usion coe�cients

within two standard deviations from the mean di�usion coe�cient or that had di�usion coe�cients greater

than 0.001 µm2/s. For each time bin, all velocity-velocity scalar products, v(t)�v(t + τ), associated with a

given time interval were obtained, and averaged using equal weights.

D-IFT measurements were performed using a custom-made pendant bubble �ow cell apparatus. A

droplet of silicone oil was initially injected from a stainless steel needle into bu�er, remaining in contact with

the needle. After the oil droplet relaxed, 10 ml of bu�er containing either lysozyme or BSA were injected

into the cell and a time-lapse movie of the pendant drop was acquired immediately after injection. The

shape of the oil droplet was then analyzed with software developed by First Ten Angstroms® wherein the

curvature of the bubble was used to calculate the surface tension using a modi�ed Young-Laplace equation

of an axisymmetric bubble known as the Bashforth-Adams equation.[8] This analysis permitted the dynamic

determination of interfacial tension as previously described.[170] Relative interfacial tension was calculated

by dividing the interfacial tension by the initial interfacial tension. Measurements were grouped into time

bins, and each condition was repeated three times. Error bars indicate the standard deviation between

replicates.

4.4 Results & Discussion

Adsorption rates for each protein were calculated from single-molecule measurements captured using

TIRFM by examining a low concentration of �uorescently-labeled protein in PBS bu�er. Under these

conditions, steady state conditions were reached quickly and the adsorption rate constant was calculated

56

by counting the number of adsorbed molecules with respect to the observation duration and area observed

and normalizing by solution concentration. We note that this represents the absolute adsorption rate, as

opposed to the net adsorption rate, and is uniquely measured using single molecule observations. The

adsorption rate constants for BSA and lysozyme were 600±300 nm/s and 3000±1500 nm/s, respectively,

where experiments were repeated three times, and uncertainties represent the standard deviation between

replicates. Intuitively, an adsorption rate constant of 600 nm/s suggests that the number of protein molecules

equivalent to those contained in a 600 nm thick slab of solution impinges on the interface each second. Under

the pH conditions employed here, previous work has suggested that the silicone-oil interface exhibits an

e�ective negative charge,[60] whereas BSA and lysozyme are expected to exhibit net negative and positive

charges, respectively, due to their respective isoelectric points.[142, 99] Therefore, BSA molecules in the bulk

solution are expected to experience repulsive electrostatic interactions with the interface, whereas those of

lysozyme molecules are attractive. The modest di�erence in the observed adsorption rates is likely due to

these factors, despite the relatively high ionic strength of the PBS solution used. However, other e�ects,

including hydrophobic exposure, can signi�cantly in�uence the measured adsorption rate. Most importantly,

these measured adsorption rates provide an important scaling parameter that will be used below to calculate a

�dimensionless time� variable, permitting direct comparison of layer formation kinetics for BSA and lysozyme.

Dynamic interfacial tension measurements were obtained as a function of interfacial aging time for

BSA and lysozyme. As described above, interfacial tension provides approximate information about the

extent to which the interface is covered by a protein layer. For example, a 10% reduction in the relative

interfacial tension (γ/γ0 = 0.9 in Figure 4.1) is likely to correspond to a situation where the surface coverage

has reached 30-50% of its maximal value.[14, 110, 116] As seen in Figure 4.1, the dynamic interfacial tension

for both proteins exhibits a characteristic initial lag time before dropping sharply. The behavior is slightly

di�erent for the two proteins, for example BSA exhibits a more signi�cant initial downward drift during the

induction phase and begins to saturate after the interfacial tension decreases to ~70% of its initial value.

However, the salient features in Figure 4.1 involve the distinctive sharp decreases in interfacial tension,

occurring at ~400s for BSA and ~1500s for lysozyme. The decrease in interfacial tension occurs distinctly

earlier for BSA than for lysozyme.[14, 110, 116]

57

Figure 4.1: Temporal evolution of the relative interfacial tension for BSA (�lled circles) and lysozyme (open

diamonds) at a bulk concentration of 70 nM in PBS where time is on a logarithmic scale and the relative

interfacial tension is de�ned as the instantaneous interfacial tension γ divided by the initial value, γ0. Error

bars indicate the standard deviation between three replicate experiments.

To account for di�erences in adsorption rate and molecular size in comparing these dynamic interfacial

tension measurements, it is convenient to calculate a dimensionless time, Γ. Assuming that each protein

molecule occupies an area corresponding to amol = πR2, where R is the solution-phase hydrodynamic radius

(3.3 and 1.8 nm for BSA and lysozyme, respectively),[156, 128] this dimensionless time is de�ned using the

equation

Γ = kads c amol t

where kads is the adsorption rate coe�cient described above, c is the bulk protein concentration, and t is time.

Intuitively, Γ represents the approximate fraction of the surface that would be covered by protein assuming

that all absorbing molecules remain on the surface permanently in their three-dimensional conformation. In

practice, Γ represents an upper limit on the surface coverage, since molecules often desorb from the interface.

In the surface science community, this quantity would correspond to the dose of adsorbate molecules, and

would be expressed in units of Langmuirs. Figure 4.2a shows the relative interfacial tension as a function of

Γ. While the shapes of the data for BSA and lysozyme are somewhat di�erent, it is interesting that, after

adjusting for protein size and adsorption rate, the aging times at which the interfacial tensions drop sharply

fall within a fairly narrow range (Γ = 0.5 for BSA and Γ = 0.8 for lysozyme). Importantly, the quantities

measured using single-molecule tracking allow us to account for di�erences in the rate at which proteins

58

appear at the interface and to express the kinetics of aging in terms of time intervals that can be directly

compared, despite the signi�cant di�erences in protein, size, charge, hydrophobicity, etc.

Figure 4.2: Various properties of 70 nM BSA (�lled circles) and 70 nM lysozyme (open diamonds) plotted

versus the dimensionless time, Γ, a value calculated by adjusting for protein adsorption rate, bulk concen-

tration and molecular size. (a) Relative interfacial tension, (b) relative mean di�usion coe�cient, (c) power

law exponent, α, associated with the mean squared displacement vs. time, and (d) the �rst non-trivial value,

G(τ = 0.2s), of the velocity-velocity autocorrelation function. Error bars from �gures (a) and (b) represent

the standard deviation between three replicate experiments.

59

The most direct way to characterize the molecular motion is to measure the apparent di�usion coe�-

cients, D, of the objects observed as a function of aging time. Figure 4.2b shows the relative mean di�usion

coe�cient as a function of aging time (normalized by the initial mean di�usion coe�cient). For each replicate

experiment, the values of D for individual trajectories were averaged (using equal weights) within each time

bin, these averages were normalized by the initial mean di�usion coe�cient for that replicate, and the overall

average was obtained by averaging the relative mean di�usion coe�cients calculated individually from three

replicate experiments (using equal weights). For BSA, it is interesting to note that the mean di�usion coef-

�cient decreases signi�cantly by a dimensionless time Γ ~0.01, whereas the interfacial tension only begins to

drop sharply at a dimensionless time Γ ~0.5. Thus, the lag times for mobility and interfacial tension di�er

by approximately two orders of magnitude. In fact, at a dimensionless time of Γ ~0.5, the average di�usion

coe�cient of the BSA layer decreases by an order of magnitude, while the interfacial tension decreases by

only 10%. This is in contrast to the behavior of lysozyme, where the dimensionless lag times for mobility (Γ

~0.4) and interfacial tension (Γ ~0.8) di�er by only a factor of two. As mentioned previously, this behavior

is consistent with a scenario where slowing down is dominated by a heterogeneous mechanism of protein

cluster formation for BSA (which can occur at low overall surface coverage) and by a homogeneous crowding

mechanism for lysozyme (which occurs only at high surface coverage).

60

Figure 4.3: Population distribution of relative di�usion coe�cients observed for (a � c) lysozyme and (d �

f) BSA. (a) and (d) depict data obtained at D/ 〈Do〉 = 1.0; (b) and (e) depict D/ 〈Do〉 ' 0.4; (c) and (f)

depict late time distributions at D/ 〈Do〉 less than or equal to 0.2.

In order to characterize the dynamic heterogeneity of the di�using objects during interfacial aging,

we examined the distributions of apparent relative di�usion coe�cients (normalized by the initial mean

di�usion coe�cient) as a function of aging time. Figure 4.3 shows representative distributions, in the form of

histograms, of the relative di�usion coe�cients at various stages of interfacial aging. For short aging times,

both lysozyme (Figure 4.3a) and BSA (Figure 4.3d) exhibit relatively narrow distributions of di�usion

coe�cients, with the BSA distribution being somewhat narrower. As the aging time increases, and the mean

di�usion coe�cient decreases, the peaks of the distributions for both proteins gradually shift to smaller

values. For BSA (Figures 4.3e and 4.3f), the peak of the distribution clearly broadens, and appears to

approach a �at distribution for the longest aging time. For lysozyme (Figures 4.3b and 4.3c), the main peak

of the distribution maintains an approximately constant width, and a subtle tail is also observed to develop

with aging time. Thus there is subtle evidence for the development of greater heterogeneity in the case of

BSA than for lysozyme, which would be expected for the layer formation mechanisms hypothesized above.

However, on the basis of these data alone, there is no question it would be di�cult to make very strong

mechanistic conclusions. In particular, our ability to observe the expected very broad distribution associated

with heterogeneous layer formation is hampered by the fact that we cannot resolve very small values of the

di�usion coe�cient (these small values cluster in a resolution-limited peak). For lysozyme, where one might

61

expect to see a well-de�ned peak whose peak moves to smaller values with aging time, it is inevitable that we

will see a small population of fast-moving objects, due in part to short-lived tracking artifacts, and in part to

molecules that may adsorb into a second layer on top of the main protein layer. Moreover, in general, it is not

clear that cluster formation should always lead to a broad distribution of apparent di�usion coe�cients, since

some clustering mechanisms (which involve dynamic association and dissociation) can result in a distribution

of clusters with a characteristic size that increases with time. Therefore, it became useful to perform more

detailed types of trajectory analysis that provided direct signatures of molecular crowding.

Figure 4.4: Mean squared displacement plotted vs. time interval for (a) BSA and (b) lysozyme at selected

time bins as annotated in the legends. These time bins were chosen to represent aging times (early, middle,

and late) at which the respective protein layers exhibited similar behavior.

The MSD versus time interval for both BSA and lysozyme, calculated from a typical experimental

run, is shown in Figure 4.4. On the logarithmic axes of Figure 4.4, the evolution of the power law exponent

is clearly visible as a change in slope with aging time. These data were �tted to the anomalous di�usion

expression described above, and the power law exponents, α, from these �ts were determined, and are shown

62

in Figure 4.2c versus dimensionless time for both BSA and lysozyme. As expected for isolated proteins

di�using at a liquid-liquid interface, both BSA and lysozyme initially exhibit Brownian motion (α = 1).

As aging time and surface coverage increase, anomalous subdi�usive motion is observed for both protein

systems, suggesting that obstacles within a crowded interface hinder molecules (or di�using clusters) from

exploring their environment to the extent that would occur in the absence of obstacles/crowding. Possible

explanations for interfacial subdi�usion include surface crowding by an excess of molecules or con�nement

of the di�using molecules by larger objects that create con�ned environments.[133]

The evolution of the subdi�usive behavior di�ers between the two protein systems in important and

interesting ways. The decrease in α for BSA (indicating the onset of subdi�usion) appears at Γ~0.2, long after

the interfacial mobility has slowed but slightly before interfacial tension begins to decrease (Figures 4.2a�c).

This observation further supports the hypothesized heterogeneous mechanism for the BSA slowing-down

phenomenon. A heterogeneous model implies that initial slowing is due to the creation of protein clusters

that are initially isolated (hence exhibiting Brownian di�usion). However, as the clusters become crowded

enough to serve as obstacles, their trajectories would become subdi�usive. Because the clusters include some

large aggregates that do not pack e�ciently, the subdi�usive motion would be expected to occur at relatively

low values of total surface coverage before the interfacial tension begins to decrease signi�cantly. This e�ect

would be even more pronounced for aggregates that are dendritic or fractal, as opposed to compact.[24] For

lysozyme, on the other hand, the power law exponent α begins to decrease at values of surface coverage (Γ~1)

that are similar to those at which the mobility and interfacial tension also begin to decrease (Figures 4.2a-

c). This is consistent with a homogenous model of layer formation since interfacial crowding of individual

proteins simultaneously slows and obstructs di�usion, and decreases the interfacial tension.

Crowding or �caging� e�ects have been widely observed to in�uence the correlation of subsequent

steps within di�usive trajectories. For example, in principle an uncon�ned Brownian trajectory exhibits no

correlation whatsoever between subsequent steps. However, in real systems, displacements that are measured

at very short time intervals generally exhibit some positive correlation, which decays over a characteristic

relaxation times. Interestingly, in situations where trajectories are obstructed or con�ned (or in viscoelastic

media), displacements measured over an appropriate range of time scales may exhibit anticorrelated behavior,

63

i.e. a step is more likely to be in the opposite direction than the previous step.[133, 172, 40, 21, 61]

Although the low signal-to-background ratio associated with single-molecule measurements greatly limits

the time resolution of our observations, we nevertheless decided to calculate velocity-velocity autocorrelation

functions to see if crowding e�ects resulted in any measurable anticorrelation. In particular, we calculated the

autocorrelation function, G(τ) = 〈v(t)�v(t+ τ)〉 /⟨v(t)2

⟩, where v(t) represents the instantaneous velocity

vector at time t.

Figure 4.5 plots G(τ) as a function of time interval for various aging times, for both BSA and lysozyme

for a representative experimental run. The time resolution of the experiments is necessarily coarse, due to

the need to acquire su�cient photon counts from individual �uorescent molecules. However, it is interesting

to note that a clear empirical trend is observed in the overall shapes of G(τ) with aging time. As expected,

for short aging times where di�using objects are isolated and un-obstructed, both protein systems initially

exhibit either an absence of autocorrelation or slight positive autocorrelation for τ > 0. However, as the

interface ages, anticorrelation is observed for G(τ = 0.2s), i.e. the shortest measureable time interval.

Figure 4.2d plots the evolution of G(τ = 0.2s) as a function of surface coverage, Γ. For time intervals of

0.4s or longer, no signi�cant correlation, positive or negative, is observed. Similar behavior has recently

been observed by Granick and coworkers for crowded colloidal systems,[61] where anticorrelated motion was

observed for time intervals of τ = 0.1s at su�ciently high volume fractions.

64

Figure 4.5: Velocity-velocity autocorrelation function vs. time interval for (a) BSA and (b) lysozyme at

selected time bins depicted in the legends. The time bins for (a) BSA are solid line Γ = 8x10−4, dotted line

Γ = 4x10−1, dashed line Γ = 7x10−1; and for (b) lysozyme are solid line Γ = 4x10−4, dotted line Γ = 2,

dashed line Γ = 5.

Remarkably, there is essentially a perfect correlation, with respect to aging time, between the appear-

ance of anticorrelated motion (Figure 4.2d) and the onset of subdi�usive motion (Figure 4.2c), despite the

fact that these analyses are completely independent. As with the anomalous power law exponent described

above, the appearance of anticorrelated motion coincided with the decrease in mobility and interfacial tension

for lysozyme, again consistent with a homogeneous mechanism for surface layer formation. For BSA, like

the evolution of the anomalous power law exponent, the onset of temporally anticorrelated motion occurred

long after the decrease in average mobility, but slightly before the decrease in interfacial tension. By the

same reasoning described previously, this is also consistent with a heterogeneous model of layer formation,

assuming that the anticorrelated motion is a consequence of the caging of di�using protein clusters by other

vicinal clusters.

It is interesting to consider the molecular-level mechanisms that could potentially lead to the distinctly

65

di�erent layer-formation mechanisms observed for BSA and lysozyme. The most likely considerations are

electrostatic interactions, which could serve to stabilize proteins against interfacial aggregation, and hy-

drophobic interactions, which would tend to enhance clustering and aggregation as seen for BSA. Under the

high ionic strength conditions associated with PBS bu�er in the experiments performed here, the e�ective

Debye length was only 0.56 nm, and electrostatic interactions were signi�cantly reduced. Nevertheless, the

absolute value of the zeta potential was greater for lysozyme than for BSA, which may have provided a

modest stabilizing e�ect, reducing the propensity for lysozyme cluster formation compared to BSA.[142, 99]

In addition, BSA exhibits greater hydrophobic exposure in its native state compared to lysozyme,[49, 20]

and BSA has also been shown to restructure on hydrophobic surfaces, while lysozyme demonstrates mini-

mal restructuring.[171, 138] Thus, it is reasonable to hypothesize that the greater exposure of hydrophobic

residues for BSA, combined with the potential for electrostatic colloidal stabilization of lysozyme are respon-

sible for the distinct mechanisms of interfacial layer formation.

4.5 Conclusions

Empirical observations suggest that di�erent protein systems exhibit distinctive mechanisms of inter-

facial layer formation. While they can be modeled using undetermined parameters, it has been di�cult to

obtain direct information connecting interfacial structure to dynamic phenomena, particularly in the context

of spatial and population heterogeneity. Previous observations, employing single molecule/particle tracking

methods, suggested the importance of a heterogeneous mechanism for layer formation that involved the cre-

ation of interfacial protein clusters. However, that approach, based on distributions of individual di�usion

coe�cients, did not generally enable direct connections between behavior at short and long aging times.

By employing trajectory analysis approaches traditionally applied to microrheological studies of colloidal

systems, we showed that dynamic single molecule/particle analysis provided direct signatures of interfacial

crowding, in addition to measuring the evolution of mobility. By comparing these dynamic signatures to

macroscopic measurements of dynamic surface tension, we readily distinguished between a system where layer

formation (and slowing down) was dominated by interfacial cluster formation (BSA) and one where layer

formation was dominated by crowding of individual protein molecules (lysozyme). These mechanisms are,

66

of course, not mutually exclusive. We believe that the methods can be applied generally, e.g. to understand

how changing experimental conditions may alter mechanisms of interfacial layer formation.

4.6 Acknowledgments

The authors gratefully acknowledge support from the National Science Foundation award #CBET-

1133871.

Chapter 5

Surfactant E�ects on Particle Generation in Antibody Formulations in

Pre-Filled Syringes

5.1 Abstract

Protein aggregation and particle formation have been observed when protein solutions contact hy-

drophobic interfaces, and it has been suggested that this undesirable phenomenon may be initiated by

interfacial adsorption and subsequent gelation of the protein. The addition of surfactants, such as polysor-

bate 20, to protein formulations has been proposed as a way to reduce protein adsorption at silicone oil

water interfaces and mitigate the production of aggregates and particles. In an accelerated stability study,

monoclonal antibody formulations containing varying concentrations of polysorbate 20 were incubated and

agitated in pre-�lled glass syringes (PFS), exposing the protein to silicone oil-water interfaces at the sili-

conized syringe walls, air-water interfaces, and agitation stress. Following agitation in siliconized syringes

that contained an air bubble, lower particle concentrations were measured in the surfactant-containing an-

tibody formulations than in surfactant-free formulations. Polysorbate 20 reduced particle formation when

added at concentrations above or below the critical micelle concentration. The ability of polysorbate 20 to

decrease particle generation in PFS corresponded with its ability to inhibit gelation of the adsorbed protein

layer, which was assessed by measuring the interfacial di�usion of individual antibody molecules at the sili-

cone oil-water interface using total internal re�ectance �uorescence (TIRF) microscopy with single-molecule

tracking.

Keywords: PFS, silicone oil, microparticles, protein formulation, protein aggregation, surfactant,

68

adsorption, monoclonal antibody, TIRFM, protein gelation, interfacial di�usion

5.2 Introduction

Therapeutic protein molecules may encounter a variety of interfaces (air-liquid, solid-liquid, and liquid-

liquid) during their manufacturing, transportation, and storage. Proteins are generally surface active and

readily adsorb to many interfaces[106]. In some formulations, adsorbed proteins may undergo conformational

changes at interfaces[43, 158, 150, 138, 173, 19, 45, 57] and they also may form viscoelastic interfacial

protein gels[129, 7, 103, 114]. In turn, formation of interfacial gels may be associated with agitation-induced

formation of protein aggregates [103, 114].

Interfaces are a particular concern for protein therapeutics formulated in glass pre-�lled syringes

(PFS). In PFS, protein molecules may be exposed to air-water interfaces due to air bubbles that typically

remain after syringe �lling and stoppering. In addition, because silicone oil is often used as a lubricant

on the syringe wall to provide low, smooth glide forces during injection, protein molecules may encounter

silicone oil-water interfaces in PFS. Adsorption to air-water interfaces and silicone oil-water interfaces has

been shown to foster protein aggregation and particle formation [3, 11, 51, 57, 101, 109, 140].

A common strategy used by the biopharmaceutical industry to decrease the negative e�ects associated

with protein adsorption to interfaces is to add nonionic surfactants such as polysorbate 20 (Tween 20®)

or polysorbate 80 (Tween 80®) to protein formulations[136, 100]. The addition of nonionic surfactants

has been shown to decrease protein aggregation[85, 153, 6, 107, 31, 38] and inhibit the formation of visi-

ble and sub-visible particles[107, 108] in a number of protein formulations subjected to a variety of stress

conditions. The protective e�ects of surfactants are commonly attributed to competitive adsorption of the

surfactant to interfaces[93, 83, 104, 103, 153]. Because of their strong a�nity for interfaces, it has been

proposed that surfactants may out-compete proteins for adsorption to interfaces, an e�ect that should cor-

relate with the critical micelle concentration (CMC) of the surfactant[93]. Polysorbate 80 has been shown

to decrease the amount of lysozyme and Factor VIII that adsorb on hydrophobic silica surfaces [75, 74], and

addition of polysorbate 20 decreased the adsorption of four di�erent model proteins at the silicone oil-water

interface[104]. Polysorbate 20 is also e�ective at displacing β-lactoglobulin from the n-hexadecane-water

69

interface[34]. Some proteins also form surfactant-protein complexes which inhibit aggregation[5]. Polysor-

bate 20 binds to hydrophobic patches on the surface of recombinant human growth hormone and decreases

aggregation at surfactant:protein molar ratios above 2[5]. Furthermore, at concentrations below their re-

spective CMC's, polysorbate 20 and polysorbate 80 inhibit agitation-induced aggregation of Albutropin and

darbepoetin alfa due to the formation of surfactant-protein complexes[31, 38].

An additional e�ect of surfactants on proteins adsorbed to interfaces is the ability of surfactants to

inhibit gelation of adsorbed protein layers. Polysorbate 20 prevented gelation of β-lactoglobulin at the air-

water interface[129] and at the n-hexadecane-water interface[34]. Addition of polysorbate 20 to formulations

of keratinocyte growth factor 2 (KGF-2) also prevented gelation at the air-water interface, and the addition

of polysorbate 20 to a pre-formed KGF-2 gel caused the gel to break down[103]. Reversal of the gelation

process was also observed when sodium dodecyl sulfate (SDS) was added to a pre-formed β-casein gel[7].

Recently, several studies attributed agitation-induced aggregation and particle formation in protein

formulations to mechanical rupture of the adsorbed protein gel layer at air-water interfaces and at oil-

water interfaces[140, 13, 57, 114, 9] . Previously, we studied protein aggregation and particle formation

in surfactant-free protein formulations in siliconized PFS. We observed that, especially in the presence of

air-water interfaces, agitation induced extensive particle formation. We attributed this particle generation

to agitation-induced rupture of a gelled protein layer at the silicone oil-water interface[58]. In the current

study, we hypothesize that addition of a nonionic surfactant to a protein formulation will inhibit interfacial

gel formation at the silicone oil-water interface and thus reduce the number of particles generated in similar

agitated PFS.

To test our hypothesis, we added the nonionic surfactant polysorbate 20 at concentrations that spanned

a range above and below the critical micelle concentration (CMC) to formulations of a model monoclonal

antibody. These formulations were �lled into glass syringes which were subsequently agitated by end-over-

end rotation. After agitation, the concentrations of particles in the formulations were measured. In addition,

particle generation was monitored in formulations wherein the polysorbate 20:monoclonal antibody molar

ratio was varied in order to probe whether protective e�ects were related to the CMC of polysorbate 20 or

to speci�c binding of polysorbate 20 to the monoclonal antibody. Finally, to assess the ability of polysorbate

70

20 to inhibit formation of interfacial protein gels, we used total internal re�ectance �uorescence (TIRF)

microscopy with single-molecule tracking to measure the e�ect of various bulk concentrations of polysorbate

20 on the interfacial di�usion of single �uorescently-labeled monoclonal antibody molecules adsorbed to

silicone oil-water interfaces.


5.3.1 Materials

Humanized IgG1 monoclonal antibody (molecular weight 146 kDa), here denoted as �3M�, was pro-

vided by MedImmune (Gaithersburg, MD).[124] The antibody was obtained at a stock concentration of 150

mg/mL in 10 mM L-histidine at pH 6. 3M was chosen because of previous studies[58] that showed it to be

prone to aggregation when exposed to silicone oil-water interfaces. Polysorbate 20 (>97 % purity, Fisher

BioReagents) was obtained from Fisher Scienti�c (Pittsburgh, PA). All bu�er salts were of reagent grade

or higher, and all solutions were prepared in de-ionized water �ltered through a 0.22 μm Millipore �lter

(Billerica, MA). Silicone oil (Dow Corning 360, 100 cSt) was of medical grade and purchased from Nexeo

Solutions (Denver, CO). The syringes used in the incubation studies were BD Hypak SCF 1mL long 27G1/2

(BD Medical-Pharmaceutical Systems, Franklin Lakes, NJ). Glass coverslips, Micro-90, and isopropyl alcohol

were obtained from Fisher Scienti�c (Waltham, MA). Nickel TEM grids (EMS G100-Ni) were obtained from

Electron Microscopy Sciences (Hat�eld, PA). Te�on® rings were fabricated in-house at the University of

Colorado Boulder.

5.3.2 Incubation of 3M Formulation with Polysorbate 20 (Above CMC) in PFS

A formulation containing 1mg/mL 3M with 0.01 % v/v polysorbate 20 in 10 mM L-histidine pH 5

was prepared using the 3M stock (described above) and a 1 % v/v stock solution of polysorbate 20 in 10 mM

L-histidine pH 5. This 3M formulation was used to �ll glass syringes. Prior to �lling, the silicone oil coating

on some of the syringes was removed, as previously described[58]. To prepare syringes containing an air

bubble, 1.26 mL of the formulation was pipetted into the syringe, and the syringe was stoppered, creating a

headspace containing 30 μL of air. For incubation conditions without headspace, the syringes were stoppered

71

such that no air bubbles remained. Triplicate syringes were prepared for each incubation condition at each

time-point. For incubation conditions with agitation, the syringes were rotated end-over-end at 1.5 rpm

at room temperature. For quiescent incubation conditions, the syringes were incubated horizontally on the

bench top at room temperature. In addition, solutions containing 10 mM L-histidine bu�er only (no protein)

were incubated in siliconized syringes either with or without headspace.

5.3.3 Agitation of 3M Formulations with Varying Surfactant:Protein Ratios in PFS

To evaluate how the surfactant:protein molar ratio in the formulation a�ects the number of parti-

cles generated by agitation in pre-�lled syringes, protein formulations containing polysorbate 20 at surfac-

tant:protein molar ratios ranging from 0 to13.1 were prepared by varying the polysorbate 20 concentration

and the 3M concentration (Table 5.1). A volume of 1.26 mL of each formulation was pipetted into siliconized

syringes, and the syringes were stoppered such that a headspace containing 30 μL of air remained in the

syringe. Triplicate syringes were prepared for each surfactant:protein molar ratio, and the syringes were

rotated end-over-end at 1.5 rpm for 24 hours at room temperature.

Table 5.1: 3M concentrations and polysorbate 20 concentrations corresponding to the polysorbate 20:3Mmolar ratios used in the formulations tested. The polysorbate 20 CMC is 0.007 % v/v (0.06 mM) [84].

3M Concentration (mg/mL) Polysorbate 20 Concentration (% v/v) Polysorbate 20:3M Ratio1.0 0.0000 0.07.6 0.0005 0.12.2 0.0005 0.31.0 0.0005 0.71.0 0.0010 1.31.0 0.0020 2.61.0 0.0050 6.51.0 0.0100 13.17.6 0.0800 13.1

5.3.4 Counting of Particles in Incubated 3M Formulations

Using the same protocol described previously[58], at each time-point during the incubation, syringes

were un-stoppered, and the formulation was removed from the �anged end of the syringe using a transfer

pipet. The protein formulation was not ejected using the syringe needle to avoid the generation of particles

72

due to plunger movement along the syringe barrel. For each sample, particles between 2 μm to 2 mm

(equivalent spherical diameter) were counted using a Fluid Imaging Technologies Benchtop FlowCAM®

(Scarborough, ME). The FlowCAM was �tted with a FC100 �ow cell, a 10X objective and collimator, and

a 0.5 mL syringe. The gain and �ash duration were set such that the average intensity mean of the image

was consistently between 180 and 200. A sample volume of 0.2 mL was analyzed for each sample at a �ow

rate of 0.145 mL/min. Particle counts were normalized by dividing the number of particles per sample

by the total volume imaged per sample to obtain the particle concentration (#/mL). In addition to the

samples incubated in syringes, triplicate samples of a bu�er solution and a protein solution (both containing

0.01 % v/v Tween 20) that had not been incubated in syringes were analyzed by FlowCAM. Furthermore,

triplicate samples of bu�er solutions and protein solutions containing the Tween 20 concentration and the

3M concentration corresponding to each surfactant:protein molar ratio tested were analyzed by FlowCAM

without prior incubation.

5.3.5 3M Labeling with Alexa Fluor 555

For TIRF microscopy experiments, 3M was dialyzed into 10mM sodium acetate pH 5 and was labeled

with Alexa Fluor® 555 succinimidyl ester (Molecular Probes, Eugene, OR) following the manufacturer's

protocol (MP 30007). The average labeling e�ciency was 9±1 �uorophores per protein molecule and was

measured using UV-visible spectroscopy at 280 nm and 555 nm, following the manufacturer's protocol. After

labeling, the protein was dialyzed back into 10 mM L-histidine pH 5 using a Slide-A-Lyzer 10,000 MWCO

dialysis cassette (Thermo Scienti�c, Rockford, IL) before TIRF microscopy experiments were conducted.

After labeling and dialysis, labeled 3M was analyzed by size exclusion chromatography (SEC) using

a TSK-GEL G3000SWXL column with guard column (TOSOH Biosciences, Montgomeryville, PA). The

�owrate was 0.6 mL/min, and the mobile phase was 0.2 M potassium phosphate monobasic, 0.2 M potassium

chloride, and 0.1 g/L sodium azide at pH 7. The absorbance was monitored at 280 nm using a Beckman

Coulter (Fullerton, CA) System Gold 166 UV detector. SEC analysis con�rmed that labeled 3M was in a

monomeric state.

73

5.3.6 3M Molecule Tracking Using Total Internal Re�ection Fluorescence (TIRF) Mi-

croscopy

TIRF microscopy experiments to examine the e�ects of added surfactant on the ability of 3M to

form interfacial gels were performed using a Nikon Eclipse TI-93 out�tted with a custom illuminator used

in conjunction with a 100x oil immersion objective. A cooled CCD camera (Photometrics Cascade 512B)

operating at -80°C was used to capture a sequence of images with a typical acquisition time of 200 ms per

image. A Cobolt Samba laser emitting at 532 nm was used as an excitation source; 900 frame movies were

captured, corresponding to ca. 3 minutes in duration.

For TIRF microscopy experiments, glass coverslips were cleaned using a cationic surfactant (Micro-

90®), Millipore-�ltered water to a resistance of 18.2 M-ohm-cm, and isopropyl alcohol and were then dried

under a nitrogen stream. Silicone oil droplets with a viscosity of 100 cSt were added to the clean coverslips

and stabilized using a nickel transmission electron microscopy (TEM) grid, ensuring a stable planar interface

between the silicone oil and the bu�er[115, 167, 17]. A Te�on® ring was then placed in contact with the

coverslip, surrounding the silicone oil-�lled TEM grid and creating a well to contain a small volume of bu�er.

A volume of 100 μl of bu�er containing 10-6 mg/mL Alexa555-labeled 3M was added to the well, and 15

minutes was allowed for the system to equilibrate, after which time a movie was taken. Then, 100 μl of a

solution containing polysorbate 20 and unlabeled 3M at various surfactant:protein molar ratios (Table 5.1)

and doped with 10�6 mg/mL of labeled 3M were added to the well, and again 15 minutes was allowed for

the system to reach equilibrium. Upon equilibrium, three movies were captured in various locations on the

silicone oil-water the interface.

5.3.7 TIRF Data Analysis

TIRF movies were analyzed using custom-designed molecule identi�cation and tracking algorithms

wherein molecules were identi�ed by a �uorescence intensity threshold. After molecules were identi�ed for

each frame, molecular trajectories were linked between frames such that an identi�ed object observed within

a 0.8 μm radius of an identi�ed object in the previous frame was classi�ed as the same object[78, 165].

The di�usion coe�cient of each molecule was calculated assuming �rst-order Fickian di�usion kinetics based

74

on the molecule's mean squared displacement and the frame rate. From the di�usion coe�cient of each

individual trajectory, the arithmetic mean was calculated, yielding the mean interfacial di�usion coe�cient

for 3M molecules adsorbed at the silicone oil-water interface under each condition tested.

5.4 Results

5.4.1 Particle Concentrations in 3M Formulations with 0.01% v/v Polysorbate 20 after

Incubation in PFS

After incubation in PFS, particles of a size greater than 2 μm were detected in all of the 3M and

the protein-free formulations that contained 0.01% v/v polysorbate 20. In all cases, even after incubation

periods of up to two weeks, the particle concentrations did not exceed 100,000 particles/mL (Figure 5.1).

The presence of an air bubble within PFS caused a small increase in the particle concentrations

observed during agitation of 3M formulations in both siliconized and un-siliconized syringes (Figure 5.1c and

5.1d). In contrast, in the absence of air-water interfaces, particle counts were roughly constant (Figure 5.1c

and 5.1d, open symbols), and there was not a noticeable di�erence in the particle concentrations between

syringes that were siliconized and syringes that were un-siliconized. For comparison, Figure 5.1d also shows

particle concentration data collected in a previous study[58] for a 3M formulation without polysorbate 20

agitated in siliconized syringes with an air bubble. In the absence of polysorbate 20, the particle counts

increased above 100,000 particles/mL after only 1 day of agitation[58].

75

Figure 5.1: Particle concentrations in 3M formulations with 0.01 % v/v polysorbate 20 and in bu�er solutions

with 0.01 % v/v polysorbate 20 agitated in PFS as a function of time. Open symbols correspond to syringes

incubated with no air bubble and closed symbols correspond to syringes incubated with an air bubble. The

particle concentrations in a bu�er solution (solid black line) and in a 3M solution (dashed black line) with 0.01

% v/v polysorbate 20 that were not incubated in syringes are also shown. The incubation conditions are as

follows: (a) L-histidine bu�er (no protein) in agitated, siliconized syringes, (b) 3M formulation in quiescent,

siliconized syringes, (c) 3M formulation in agitated, un-siliconized syringes, and (d) 3M formulation in

agitated, siliconized syringes. For comparison, the gray symbols in panel (d) correspond to a 3M formulation

with no surfactant agitated in siliconized syringes with an air bubble [58].

5.4.2 Particle Concentrations in 3M Formulations Containing Various Surfactant:Protein

Ratios After Agitation in PFS

Particle generation in 3M formulations containing polysorbate 20 at various surfactant:protein molar

ratios was monitored before and after 24 hours of agitation in siliconized PFS containing a headspace

volume of 30 µL (Figure 5.2). The resulting change in particle concentrations depended on the polysorbate

20:3M molar ratio. The greatest increase in particle concentrations was measured in polysorbate-free 3M

formulations. In this case, there was a two order of magnitude increase in the particle concentrations after

76

the 24 hour agitation period (Figure 5.2). Addition of even very small amounts of polysorbate 20 (e.g,

at a polysorbate 20:3M molar ratio of 0.1) reduced the number of particles formed. The e�ect of added

surfactants appeared to reach a saturation plateau at polysorbate 20:3M molar ratios greater than ca. 1,

where only minimal increases in particle concentrations could be observed after agitation.

77

Figure 5.2: Particle concentrations measured in 3M formulations as a function of the polysorbate 20:3M

molar ratio in the formulation. Open symbols represent the particle concentrations in 3M formulations that

were not incubated. Closed symbols represent the particle concentrations in 3M formulations that were

agitated for 24 hours with an air bubble in siliconized syringes. The color of each symbol corresponds to

the polysorbate 20 concentration (% v/v) in the formulation, as shown in the legend. At each polysorbate

20:3M ratio, the particle concentration in a bu�er solution with the same polysorbate 20 concentration was

subtracted from the particle concentration measured in the non-incubated 3M formulation and from that

measured in the agitated 3M formulation.

78

5.4.3 Interfacial Di�usion of Labeled 3M Molecules at the Silicone Oil-Water Interface in

Formulations with Varying Surfactant:Protein Molar Ratios

To assess the e�ects of polysorbate 20 on the formation of 3M gels at the silicone oil-water interface,

trace amounts (10-6 mg/mL, ca. 10-11 M) of Alexa Fluor®-labeled 3M were added as reporter molecules

to solutions of 3M at various bulk concentrations, and the mean di�usion coe�cient of the labeled 3M at

the silicone oil-water interface was measured as a function of the polysorbate 20:unlabeled 3M bulk molar

ratio (Figure 5.3) using TIRF microscopy.

In the absence of unlabeled 3M and polysorbate 20, the mean di�usion coe�cient of labeled 3M

molecules at the silicone oil-water interface was assumed to re�ect only protein-interface interactions (Figure

5.3, solid line). This assumption was made because the bulk concentration (10-6 mg/mL) of labeled 3M was

quite low, and the labeled molecules that could be individually observed on the interface appeared to be at

low surface density.

In samples that contained 3M at bulk concentrations ranging from1.0-7.6 mg/mL, the interfacial mean

di�usion coe�cients of the labeled 3M molecules were signi�cantly reduced compared to the sample that

contained only labeled 3M at 10-6 mg/mL (Figure 5.3). Furthermore, in samples with added unlabeled 3M

that did not contain polysorbate 20, gels formed at the silicone oil-water interface, as evidenced by mean

di�usion coe�cients for the labeled 3M molecules that were almost zero (Figure 5.3).

Addition of polysorbate 20 inhibited gel formation in samples containing added, unlabeled 3M. Al-

though di�usion coe�cients for 3M on the silicone oil-water interface in these samples were approximately

four-fold smaller than those measured in samples containing only labeled 3M at a bulk concentration of 10-6

mg/mL (presumably due to hindrance of di�usion by other protein molecules at the crowded interface),

di�usion was much more rapid than in corresponding samples without polysorbate 20 (Figure 5.3). Over

the range of 3M bulk concentrations (1.0-7.6 mg/mL) tested, the mean di�usion coe�cients were roughly

constant when the surfactant:protein molar ratio was above ca. 1.

79

Figure 5.3: Mean di�usion coe�cients (μm2/s) of trace amounts of labeled 3M at the silicone oil-water

interface as a function of the polysorbate 20:unlabeled 3M molar ratio. The open symbols correspond to

formulations with a bulk unlabeled 3M concentration of 1.0 mg/mL. The closed symbols correspond to

formulations with a bulk unlabeled 3M concentration of 7.6 mg/mL. The solid line represents the mean

di�usion coe�cient of interfacially-adsorbed labeled 3M molecules in the absence of polysorbate 20 and

without addition of any unlabeled 3M.

5.5 Discussion

5.5.1 Particle Generation in Antibody Formulations in PFS

In a previous study[58], large numbers of particles were observed when surfactant-free formulations of

3M were agitated in siliconized syringes containing headspace. We proposed a mechanism wherein surface

tension forces at the three-phase (silicone oil-water-air) contact line ruptured layers of adsorbed, gelled

proteins leading to the creation of particles.

We hypothesized that the addition of nonionic surfactant would decrease the number of particles

generated by agitation in siliconized PFS by inhibiting the formation of a gelled protein layer at the silicone

oil-water interface. This hypothesis was �rst tested in a 3M formulation containing 0.01 % v/v polysorbate

20. At this concentration of polysorbate 20, which is above the reported CMC of 0.007% v/v[84], the

80

interfaces present in the pre-�lled syringe are expected to be saturated with surfactant. Addition of 0.01 %

v/v polysorbate 20 inhibited the formation of particles, and there were only minimal di�erences in particle

concentrations between formulations agitated in siliconized and un-siliconized syringes (Figure 5.1c and

5.1d, closed symbols). In contrast, our previous study demonstrated that in the absence of polysorbate 20,

particle generation was at least one order of magnitude greater in siliconized syringes than in un-siliconized

syringes[58].

5.5.2 In�uence of Polysorbate 20 Concentration on Particle Generation in PFS

In order to investigate further the mechanism by which surfactants inhibit interface-induced particle

generation, the polysorbate 20 concentration and the polysorbate 20:3M molar ratio in the protein formu-

lation were varied. For the 1 mg/mL 3M formulation with 0.01 % v/v polysorbate 20 used in the �rst part

of this study, the polysorbate 20:3M molar ratio was 13.1 (Table 5.1), and the polysorbate 20 concentration

was above its CMC. For polysorbate 20:3M molar ratios ≥ 0.7, the 3M concentration used in the formulation

was 1.0 mg/mL, and the molar ratio was manipulated by changing the polysorbate 20 concentration (Table

5.1). To obtain ratios 3M:polysorbate 20 molar ratios < 0.7, the 3M concentration was increased while the

polysorbate 20 concentration was held constant at 0.0005 % v/v (Table 5.1).

If saturation of the silicone oil-water interface with surfactant were necessary to inhibit interfacial

particle generation in a pre-�lled syringe, then particle generation would not be anticipated to be inhibited

by the presence of polysorbate 20 at polysorbate 20 concentrations below the CMC, where the interface is not

saturated with surfactant. However, we observed that particle generation in 1 mg/ml 3M formulations was

almost completely inhibited at polysorbate 20 concentrations well below the CMC (Figure 5.2). Interestingly,

the ability of sub-CMC levels of surfactant to inhibit particle formation decreased as the molar ratio of

polysorbate 20:3M decreased. At a constant polysorbate 20 concentration of 0.0005 % v/v, particle formation

was inhibited at a polysorbate 20:3M ratio of 0.7, but not at ratios of 0.1 and 0.3. A potential explanation

for this behavior is that polysorbate 20 binding to 3M is responsible for inhibiting 3M aggregation at the

silicone oil-water interface.

81

5.5.3 In�uence of Polysorbate 20 Concentration on Gelation of 3M Molecules at the

Silicone Oil-Water Interface

We hypothesized that the addition of polysorbate 20 would inhibit the gelation of 3M molecules at the

silicone oil-water interface. Therefore, we used TIRF microscopy to directly monitor the interfacial di�usion

of �uorescently-labeled 3M molecules on the silicone oil-water interface. As a control, the mean di�usion

coe�cient of labeled 3M (bulk concentration 10-6 mg/mL) was measured in the absence of polysorbate

20 and without added unlabeled 3M. In the absence of both polysorbate 20 and unlabeled 3M, the mean

di�usion coe�cient, ca. 0.2 μm2/s, re�ected the mobility of 3M molecules at the silicone oil-water interface

without any crowding due to the adsorption of polysorbate 20 or other 3M molecules (Figure 5.3, solid line).

In all other TIRF microscopy experiments, the amount adsorbed at the interface was much higher due

to the higher bulk concentrations of polysorbate 20 and unlabeled 3M. At these higher concentrations, the

mean di�usion coe�cient was slower than that measured for the control due to crowding at the interface and

resulting interactions between adsorbed molecules. Furthermore, when the adsorbed layer formed gels, large-

scale translational motions of 3M molecules were no longer observed, and the interfacial di�usion coe�cients

were e�ectively zero.

In formulations with polysorbate 20:3M molar ratios above ca. 1, observed values of the mean di�usion

coe�cient of 3M molecules at the silicone oil-water interface were between 0.06-0.08 μm2/s (Figure 5.3).

Because the protein molecules were observed to be di�using relatively rapidly on the interface, we inferred

that, under these conditions, the protein adsorbed to the silicone oil-water interface did not form gels.

Furthermore, in the TIRF microscopy method that we used, the motion of any protein molecules that were

present in the liquid near, but not adsorbed to, the interface was so fast that these protein molecules could

not be tracked individually and registered simply as background �uorescence. On the interface, individual

3M molecules could be observed to di�use, and thus, we may infer that the presence of polysorbate 20 did

not completely prevent 3M molecules from adsorbing to the interface, even at polysorbate 20 concentrations

above its CMC. However, interfacial di�usion of 3M in formulations with polysorbate 20:3M molar ratios

< 1 was dramatically decreased, suggesting that interfacial gels formed under conditions with very low

82

polysorbate 20:3M ratios. This correlates with the higher particle concentrations observed in agitated 3M

formulations with polysorbate 20:3M molar ratios < 1. The number of particles generated in the formulations

increased (Figure 5.2) because polysorbate 20 could not inhibit gelation at polysorbate 20:3M molar ratios

< 1 as e�ectively as it could at molar ratios > 1. In a study by Courthaudon, et al., surfactant:protein molar

ratios as low as 1 were observed to inhibit gelation of β-lactoglobulin at the n-hexadecane-water interface[34].

They also noted that the surfactant:protein molar ratio required to a�ect protein gelation was much lower

than the molar ratio necessary to completely displace protein from the interface. In a later study, Kragel,

et al. demonstrated the ability of polysorbate 20, SDS, and cetyl trimethyl ammonium bromide (CTAB) to

increase the surface di�usion of β-lactoglobulin at the air-water interface at concentrations below the CMC

of each surfactant[92].

Furthermore, the di�erence in gelation between the low and high polysorbate 20:3M molar ratios was

not due to a di�erence in 3M concentration between the formulations. In addition to the aforementioned

experiments that used a 1 mg/mL 3M formulation, the interfacial di�usion of labeled 3M in a formulation

that contained 7.6 mg/mL 3M was measured at both a low and a high polysorbate 20:3M molar ratio. This

formulation showed di�usion indicative of a non-gelled protein layer at the high molar ratio but showed

di�usion indicative of a gelled protein layer at the low molar ratio (Figure 5.3, closed symbols).

5.5.4 E�ects of Surfactants on Protein Gelation at Interfaces

Surfactants have been observed to protect proteins against interface-induced aggregation through sev-

eral di�erent mechanisms. The most commonly cited mechanism is preferential adsorption of the surfactant

to an interface which inhibits protein adsorption at that same interface. In several di�erent systems, surfac-

tants at concentrations above their CMC's were shown to decrease protein aggregation, and these protective

e�ects were attributed to the preferential adsorption mechanism[93, 83, 103, 153]. This explanation is rea-

sonable because at concentrations above their CMC, surfactants will saturate the interface. Furthermore, at

the silicone oil-water interface, Ludwig, et al. observed signi�cantly less adsorption of four di�erent proteins

to the interface in the presence of polysorbate 20 above its CMC[104].

The other common mechanism by which surfactants may protect proteins against aggregation at

83

interfaces is the stabilization of protein conformation that may result from surfactant binding to protein

native-state structures. Binding of nonionic surfactants to proteins has been documented in several studies[5,

31, 38]. Surfactant molecules can interact with hydrophobic patches on a protein's surface which then inhibits

the protein-protein interactions that lead to aggregation[5]. Alternatively, preferential binding of surfactants

to a protein molecule's native state (as opposed to binding to non-native, unfolded states) stabilizes the native

state and increases the protein's free energy of unfolding[31]. The formation of surfactant-protein complexes

has been observed to protect protein formulations against interfacial damage at surfactant concentrations

below the CMC[5, 31, 38] because, in this mechanism, the interface does not need to be saturated with

surfactant. However, these interactions between surfactant and protein that provide protective e�ects have

not been observed for every surfactant-protein system[93, 83, 103], and in systems where surfactant-protein

complexes are not formed, it has been observed that the surfactant must be present above the CMC in order

to decrease interface-induced protein aggregation by the preferential adsorption mechanism.

In addition, surfactants will a�ect the gelation of an adsorbed protein layer. Liu, et al. showed that

KGF-2 did not form an interfacial gel in the presence of 0.01% w/v polysorbate 20[103] because polysorbate

20 inhibited adsorption of KGF-2 to the air-water interface. In this case, polysorbate 20 was present above

its CMC, and binding of polysorbate 20 to KGF-2 was not observed. However, other studies have shown

that surfactants a�ect the di�usion of protein molecules at an interface at concentrations well below their

CMC's[34, 32, 92, 129, 72, 7]. In these cases, the protein was adsorbed at the interface, but it did not form

a gel. Instead of complete displacement of protein by surfactant at the interface, the inhibition of gelation

at sub-CMC levels was likely due to the interaction of surfactant molecules with hydrophobic patches on the

protein's surface which inhibited the protein-protein interactions necessary for interfacial gel formation.

In our agitated, siliconized syringe system, particles were formed by rupture of the protein gel layer[58].

Therefore, inhibition of protein gelation was hypothesized to decrease particle formation in this system. The

protective e�ects of polysorbate 20 were seen at concentration both above and below the CMC of polysorbate

20. At surfactant:protein molar ratios > 1, interfacial di�usion coe�cient measurements indicated that the

adsorbed protein layer was not gelled, but protein was still present at the silicone oil-water interface. Thus,

at sub-CMC levels, it was likely that gelation was inhibited, not by complete displacement of the protein

84

from the interface, but by the interaction of polysorbate 20 with 3M which inhibited the protein-protein

interactions required for gelation. Even at polysorbate 20 concentrations above the CMC, interfacial di�usion

coe�cient measurements indicated that 3M molecules were still present at the silicone oil-water interface.

This was consistent with quartz crystal microbalance measurements of the adsorption and viscoelastic nature

of an Fc-fusion protein and polysorbate 20 at the silicone oil-water interface[46]. For the Fc-fusion protein-

polysorbate 20 system, both the protein and the surfactant adsorbed to the silicone oil-water interface in

solutions with a polysorbate 20 concentration of 0.02 % w/v (above the CMC). However, in the presence of

0.02 % w/v polysorbate 20, the viscoelastic nature of the adsorbed protein layer was signi�cantly di�erent

than the viscoelasticity of the adsorbed protein layer in the absence of polysorbate 20[46].

5.6 Conclusions

Polysorbate 20, at concentrations above and below the CMC, was observed to decrease particle gen-

eration in formulations of a model antibody that were agitated in siliconized syringes with headspace. In

PFS �lled with formulations that contained polysorbate 20:antibody molar ratios above ca. 1, no increase

in particle concentration could be detected after 24 hours of agitation. Also, interfacial di�usion coe�cient

measurements showed that, at polysorbate 20:antibody molar ratios above ca. 1, the presence of polysorbate

20 inhibited the gelation of antibody molecules adsorbed at the silicone oil-water interface. At polysorbate

20:antibody molar ratios below ca. 1, particle concentrations were increased after agitation, and polysorbate

20 was not e�ective in inhibiting gelation of adsorbed protein molecules at the silicone oil-water interface.

The lack of correlation between the CMC of the surfactant and its protective e�ect against aggregation,

the apparent stoichiometric dependence of the polysorbate-induced inhibition of aggregation as well as the

TIRF observations of antibody adsorbed to the silicone oil-water interface even at surfactant concentrations

above the CMC suggest that polysorbate 20 does not inhibit gelation solely by displacing the protein from

the interface. Rather, we speculate that polysorbate 20 binding to the protein interferes with protein-protein

interactions required for protein gelation at the silicone oil-water interface.

Chapter 6

Conclusions

The work presented in this thesis demonstrated how a clear understanding of non-covalent interactions

a�ect interfacial biopolymer dynamics. Electrostatic and hydrophobic interactions were both found to be

critically important in the formation of a ssDNA-surfactant complex. It was proposed that the ssDNA-

surfactant complex is con�ned to two dimensions, increasing the surfactant area per molecule. The increased

surfactant molecular area was shown as the driving force in the observed LC anchoring transition. The

dramatic changes in DNA structure upon hybridization was proposed to allow the surfactant to reorganize

at the interface, resulting in the subsequent LC anchoring transition.

Electrostatic repulsion was shown to dramatically in�uence BSA elementary adsorption rates on fused

silica, but demonstrated minimal e�ect on desorption rates or interfacial di�usion. As predicted by DLVO

theory, high ionic strength conditions greatly minimized the e�ects electrostatics had on adsorption rates.

Alternatively, adsorption rates were not apparently increased by like charge attraction, suggesting that

other short-range barriers to adsorption are the limiting factor under these conditions. The apparent lack

of in�uence electrostatic played in desorption and interfacial di�usion suggest that these phenomena are

primarily in�uenced by non-electrostatic short-range interactions.

Single-molecule techniques were used at the silicone-oil water interface to examine protein layer for-

mation mechanisms. By employing trajectory analysis approaches traditionally applied to micro-rheological

studies of colloidal systems, we showed that dynamic single molecule/particle analysis provided direct sig-

natures of interfacial crowding, in addition to measuring the evolution of mobility. By comparing these

dynamic signatures to macroscopic measurements of dynamic surface tension, we readily distinguished be-

86

tween a system where layer formation (and slowing down) was dominated by interfacial cluster formation

and one where layer formation was dominated by crowding of individual protein molecules. These mech-

anisms are, of course, not mutually exclusive. We believe that the methods can be applied generally, e.g.

to understand how changing experimental conditions may alter mechanisms of interfacial layer formation.

Similar techniques were used to examine how the surfactant, Polysorbate 20, a�ected protein layer forma-

tion. Polysorbate 20 displayed protective e�ects at polysorbate 20:antibody molar ratios above ca. 1, when

a model antibody was agitated in siliconized syringes. Microscopically, interfacial di�usion coe�cient mea-

surements demonstrated elevated di�usion coe�cients polysorbate 20:antibody molar ratios above ca. 1,

correlating with the agitation observations. Polysorbate 20:antibody molar ratios below ca. 1 demonstrated

non-protective e�ects within siliconized syringe agitation studies as well as being ine�ective in inhibiting

gelation of adsorbed protein molecules at the silicone oil-water interface. These results led to the hypothe-

sis that polysorbate 20 forms a complex with the protein a�ecting protein-protein interactions required for

protein gelation at the silicone oil-water interface.

The implementation of single-molecule techniques along with traditional ensemble-averaging tech-

niques provide scientists with unique perspectives at biological systems. The development of micro-rheological

analysis techniques using single-molecule TIRFM provides an exciting new method to study viscoelastic layers

that has been previously unexplored. The implementation of such a technique may address many unanswered

questions traditional rheological techniques are unable to explore.

Bibliography

[1] J L Adgate, A Bartekova, P C Raynor, J G Griggs, A D Ryan, B R Acharya, C J Volkmann, D DMost, S Lai, and M D Bonds. Detection of Organophosphate Pesticides Using a Prototype LiquidCrystal Monitor. J. Environ. Monit., 11(1):49�55, 2009.

[2] O Albrecht, H Gruler, and E Sackmann. Polymorphism of Phospholipid Monolayers, 1978.

[3] Advait Badkar, Amanda Wolf, Leigh Bohack, and Parag Kolhe. Development of BiotechnologyProducts in Pre-Filled Syringes: Technical Considerations and Approaches. AAPS PharmSciTech,12(2):564�572, June 2011.

[4] Stefania G Baldursdottir, Maria Stoier Fullerton, Signe Hougaard Nielsen, and Lene Jorgensen. Ad-sorption of Proteins at the Oil/Water Interface-Observation of Protein Adsorption by Interfacial ShearStress Measurements. Colloids Surf., B, 79(1):41�46, August 2010.

[5] N B Bam, J L Cleland, J Yang, M C Manning, J F Carpenter, R F Kelley, and T W Randolph. TweenProtects Recombinant Human Growth Hormone Against Agitation-Induced Damage via HydrophobicInteractions. J. Pharm. Sci., 87(12):1554�1559, 1998.

[6] N B Bam, T W Randolph, and J L Cleland. Stability of Protein Formulations: Investigation ofSurfactant E�ects by a Novel EPR Spectroscopic Technique, 1995.

[7] Grigor B Bantchev and Daniel K Schwartz. Surface Shear Rheology of β-Casein Layers at the Air /Solution Interface : Formation of a Two-Dimensional Physical Gel. Langmuir, 19(7):2673�2682, 2003.

[8] F Bashforth and J C Adams. Theories of Capillary Action. Cambridge University Press, London, 1883.

[9] P. Basu. Particle Formation in IgG1 Formulations During Rupture of Air-Water and Silicone Oil-WaterInterfaces in IgG1 Samples. Press, 2013.

[10] Pinaki Basu, Sampathkumar Krishnan, Renuka Thirumangalathu, Theodore W Randolph, and John FCarpenter. IgG1 Aggregation and Particle Formation Induced by Silicone-Water Interfaces on Sil-iconized Borosilicate Glass Beads: A Model for Siliconized Primary Containers. J. Pharm. Sci.,102(3):852�865, March 2013.

[11] Jared S Bee, David Chiu, Suzanne Sawicki, Jennifer L Stevenson, Koustuv Chatterjee, Erwin Freund,John F Carpenter, and Theodore W Randolph. Monoclonal Antibody Interactions with Micro- andNanoparticles: Adsorption, Aggregation, and Accelerated Stress Studies. J. Pharm. Sci., 98(9):3218�3238, September 2009.

[12] Jared S Bee, Theodore W Randolph, John F Carpenter, Steven M Bishop, and Mariana N Dimitrova.E�ects of Surfaces and Leachables on the Stability of Biopharmaceuticals. J. Pharm. Sci., 100(10):4158�4170, 2011.

88

[13] Jared S Bee, Daniel K Schwartz, Siwar Trabelsi, Erwin Freund, Jennifer L Stevenson, John F Car-penter, and Theodore W Randolph. Production of Particles of Therapeutic Proteins at the Air-WaterInterface During Compression/Dilation Cycles. Soft Matter, 8:10329�10335, 2012.

[14] C J Beverung, C J Radke, and H W Blanch. Protein Adsorption at the Oil/Water Interface: Char-acterization of Adsorption Kinetics by Dynamic Interfacial Tension Measurements. Biophys. Chem.,81:59�80, September 1999.

[15] Xinyan Bi and Kun Lin Yang. Real-Time Liquid Crystal-Based Glutaraldehyde Sensor. Sens. ActuatorB-Chem., 134(2):432�437, 2008.

[16] Je�rey M Brake and Nicholas L Abbott. An Experimental System for Imaging the Reversible Ad-sorption of Amphiphiles at Aqueous-Liquid Crystal Interfaces. Langmuir, 18(16):6101�6109, August2002.

[17] Je�rey M Brake, Maren K Daschner, Yan-Yeung Luk, and Nicholas L Abbott. Biomolecular Interac-tions at Phospholipid-Decorated Surfaces of Liquid Crystals. Science, 302(5653):2094�2097, December2003.

[18] Je�rey M Brake, Andrew D Mezera, and Nicholas L Abbott. E�ect of Surfactant Structure on theOrientation of Liquid Crystals at Aqueous-Liquid Crystal Interfaces. Langmuir, 19(16):6436�6442,August 2003.

[19] Keith A. Britt, Daniel K. Schwartz, Christine Wurth, Hanns Christian Mahler, John F. Carpenter,and Theodore W. Randolph. Excipient E�ects on Humanized Monoclonal Antibody Interactions withSilicone Oil Emulsions. J. Pharm. Sci., 101(12):4419�4432, 2012.

[20] Anna Bujacz. Structures of Bovine, Equine and Leporine Serum Albumin. Acta Crystallogr. Sect.D-Biol. Crystallogr., 68(10):1278�1289, 2012.

[21] Stas Burov, Jae-Hyung Jeon, Ralf Metzler, and Eli Barkai. Single Particle Tracking in Systems ShowingAnomalous Di�usion: the Role of Weak Ergodicity Breaking. Phys. Chem. Chem. Phys., 13:1800�1812,2011.

[22] Hans-Jürgen Butt, Karlheinz Graf, and Michael Kappl. Physics and Chemistry of Interfaces. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, second edition, 2008.

[23] Maria Monica Castellanos, Jai A Pathak, and Ralph H Colby. Both Protein Adsorption and Aggrega-tion Contribute to Shear Yielding and Viscosity Increase in Protein Solutions. Soft Matter, 10:122�131,2014.

[24] Maria Monica Castellanos, Jai A. Pathak, William Leach, Steven M. Bishop, and Ralph H. Colby.Explaining the Non-Newtonian Character of Aggregating Monoclonal Antibody Solutions Using Small-Angle Neutron Scattering. Biophys. J., 107(2):469�476, 2014.

[25] V Chan, D.J. Graves, P Fortina, and S.E. McKenzie. Adsorption and Surface Di�usion of DNAOligonucleotides at Liquid/Solid Interfaces. Langmuir, 13(2):320�329, 1997.

[26] V Chan, S E McKenzie, S Surrey, P Fortina, and D J Graves. E�ect of Hydrophobicity and Electro-statics on Adsorption and Surface Di�usion of DNA Oligonucleotides at Liquid/Solid Interfaces. J.Colloid Interface Sci., 203(1):197�207, 1998.

[27] Ernst A Chantelau and Michael Berger. Pollution of Insulin with Silicone Oil, a Hazard of DisposablePlastic Syringes. Lancet, 1(8443):1459, 1985.

[28] Chih-Hsin Chen and Kun-Lin Yang. Detection and Quanti�cation of DNA Adsorbed on Solid Surfacesby Using Liquid Crystals. Langmuir, 26(3):1427�1430, February 2010.

89

[29] Eva Y Chi, Sampathkumar Krishnan, Theodore W Randolph, and John F Carpenter. Physical Stabilityof Proteins in Aqueous Solution: Mechanism and Driving Forces in Nonnative Protein Aggregation.Pharm. Res., 20(9):1325�36, September 2003.

[30] Eva Y. Chi, Joachim Weickmann, John F. Carpenter, Mark C. Manning, and Theodore W. Ran-dolph. Heterogeneous Nucleation-Controlled Particulate Formation of Recombinant Human Platelet-Activating Factor Acetylhydrolase in Pharmaceutical Formulation. J. Pharm. Sci., 94(2):256�274,2005.

[31] Danny K. Chou, Rajesh Krishnamurthy, Theodore W. Randolph, John F. Carpenter, and Mark CornellManning. E�ects of Tween 20 and Tween 80 on the Stability of Albutropin During Agitation. J. Pharm.Sci., 94(6):1368�1381, 2005.

[32] David C. Clark, Alan R. Mackie, Pete J. Wilde, and David R. Wilson. Di�erences in the Structureand Dynamics of the Adsorbed Layers in Protein-Stabilized Model Foams and Emulsions. FaradayDiscuss., 98:253, 1994.

[33] P. J. Collings and M Hird. Introduction to Liquid Crystals. C R C Press LLC, Bristol, 1997.

[34] J.L. Courthaudon, E. Dickinson, and D.G. Dalgleish. Competitive Adsorption of β-Casein and NonionicSurfactants in Oil-in-Water Eemulsions. J. Colloid Interface Sci., 145(2):390�395, 1991.

[35] L Cristofolini, T Berzina, S Erokhina, O Konovalov, and V Erokhin. Structural Study of the DNADipalmitoylphosphatidylcholine Complex at the Air-Water Interface. Biomacromolecules, 8(7):2270�2275, 2007.

[36] Shuxun Cui, You Yu, and Zhangbi Lin. Modeling Single Chain Elasticity of Single-Stranded DNA: AComparison of Three Models. Polymer, 50(3):930�935, January 2009.

[37] R. M. C. Dawson, D. C. Elliot, W. H. Elliot, and K. M. Jones. Data for Biochemical Research. OxfordSciencePubl., 3rd edition, 1986.

[38] Songpon Deechongkit, Jie Wen, Linda O. Narhi, Yijia Jiang, Sungae S. Park, Justin Kim, and Bruce A.Kerwin. Physical and Biophysical E�ects of Polysorbate 20 and 80 on Darbepoetin Alfa. J. Pharm.Sci., 98(9):3200�3217, 2009.

[39] Roberto A Depaz, Tzvetelina Chevolleau, Sébastien Jou�ray, Roja Narwal, and Mariana N. Dimitrova.Cross-Linked Silicone Coating: A Novel Pre�lled Syringe Technology That Reduces Subvisible Particlesand Maintains Compatibility with Biologics. J. Pharm. Sci., 103(5):1384�1393, 2014.

[40] M. A. Despósito and A. D. Viñales. Subdi�usive Behavior in a Trapping Potential: Mean SquareDisplacement and Velocity Autocorrelation Function. Phys. Rev. E, 80:021111, 2009.

[41] Rita Dias, Sergey Mel'nikov, Bjorn Lindman, and Maria G. Miguel. DNA Phase Behavior in thePresence of Oppositely Charged Surfactants. Langmuir, 16(24):9577�9583, 2000.

[42] Rita S. Dias, Luís M. Magno, Artur J M Valente, Dibyendu Das, Prasanta K. Das, Souvik Maiti,Maria G. Miguel, and Björn Lindman. Interaction Between DNA and Cationic Surfactants: E�ect ofDNA Conformation and Surfactant Headgroup. J. Phys. Chem. B, 112(46):14446�14452, 2008.

[43] E. Dickinson, DS Horne, JS Phipps, and RM Richardson. A Neutron Re�ectivity Study of the Ad-sorption of β-Casein at Fluid Interfaces. Langmuir, 9(1):242�248, April 1993.

[44] Ken A Dill and Sarina Broomberg. Molecular Driving Forces: Statistical Thermodynamics in Chemistry and Biology.Garland Science, New York, �rst edition, 2003.

[45] Nitin Dixit, Kevin M. Maloney, and Devendra S. Kalonia. Application of Quartz Crystal Microbalanceto Study the Impact of pH and Ionic Strength on Protein-Silicone Oil Interactions. Int. J. Pharm.,412(1-2):20�27, 2011.

90

[46] Nitin Dixit, Kevin M Maloney, and Devendra S Kalonia. The E�ect of Tween 20 on Silicone Oil-FusionProtein Interactions. Int. J. Pharm., 429(1-2):158�67, June 2012.

[47] Jesper Donsmark, Lene Jorgensen, Susanne Mollmann, Sven Frokjaer, and Christian Rischel. Kineticsof Insulin Adsorption at the Oil-Water Interface and Di�usion Properties of Adsorbed Layers MonitoredUsing Fluorescence Correlation Spectroscopy. Pharm. Res., 23(1):148�155, January 2006.

[48] Jesper Donsmark and Christian Rischel. Fluorescence Correlation Spectroscopy at the Oil-Water Inter-face: Hard Disk Di�usion Behavior in Dilute β-Lactoglobulin Layers Precedes Monolayer Formation.Langmuir, 23(12):6614�6623, June 2007.

[49] F Eisenhaber and P Argos. Hydrophobic Regions on Protein Surfaces: De�nition Based on HydrationShell Structure and a Quick Method for their Computation. Protein Eng., 9(12):1121�1133, 1996.

[50] Svetlana Erokhina, Tatiana Berzina, Luigi Cristofolini, Oleg Konovalov, Victor Erokhin, and Marco PFontana. Interaction of DNA Oligomers with Cationic Lipidic Monolayers: Complexation and Splitting.Langmuir, 23(8):4414�4420, 2007.

[51] Flora Felsovalyi, Sebastien Janvier, Sebastien Jou�ray, Herve Soukiassian, and Paolo Mangiagalli.Silicone-Oil-Based Subvisible Particles: Their Detection, Interactions, and Regulation in Pre�lled Con-tainer Closure Systems for Biopharmaceuticals. J. Pharm. Sci., 101(12):4569�4583, 2012.

[52] S P Fodor, R P Rava, X C Huang, a C Pease, C P Holmes, and C L Adams. Multiplexed BiochemicalAssays with Biological Chips. Nature, 364(6437):555�556, 1993.

[53] Erik M Freer, Kang Sub Yim, Gerald G Fuller, and Clayton J Radke. Interfacial Rheology of Glob-ular and Flexible Proteins at the Hexadecane/Water Interface: Comparison of Shear and DilatationDeformation. J. Phys. Chem. B, 108(12):3835�3844, March 2004.

[54] Erik M Freer, Kang Sub Yim, Gerald G Fuller, and Clayton J Radke. Shear and Dilatational Relax-ation Mechanisms of Globular and Flexible Proteins at the Hexadecane/Water Interface. Langmuir,20(23):10159�10167, November 2004.

[55] Arivalagan Gajraj and Robert Y. Ofoli. E�ect of Extrinsic Fluorescent Labels on Di�usion and Ad-sorption Kinetics of Proteins at the Liquid-Liquid Interface. Langmuir, 16(21):8085�8094, October2000.

[56] Arivalagan Gajraj and Robert Y. Ofoli. Quantitative Technique for Investigating MacromolecularAdsorption and Interactions at the Liquid-Liquid Interface. Langmuir, 16(9):4279�4285, May 2000.

[57] Alana Gerhardt, Kurt Bonam, Jared S Bee, John F Carpenter, and Theodore W Randolph. IonicStrength A�ects Tertiary Structure and Aggregation Propensity of a Monoclonal Antibody Adsorbedto Silicone Oil-Water Interfaces. J. Pharm. Sci., 102(2):429�440, February 2013.

[58] Alana Gerhardt, Nicole R Mcgraw, Daniel K Schwartz, Jared S Bee, John F Carpenter, andTheodore W Randolph. Protein Aggregation and Particle Formation in Pre�lled Glass Syringes. J.Pharm. Sci., 103(6):1601�1612, June 2014.

[59] Claudia R Gordijo, Khajag Koulajian, Adam J Shuhendler, Leonardo D Bonifacio, Hui Yu Huang,Simon Chiang, Geo�rey A Ozin, Adria Giacca, and Xiao Yu Wu. Nanotechnology-Enabled ClosedLoop Insulin Delivery Device: In Vitro and In Vivo Evaluation of Glucose-Regulated Insulin Releasefor Diabetes Control. Adv. Funct. Mater., 21(1):73�82, January 2011.

[60] Y Gu and D Li. The Zeta-Potential of Silicone Oil Droplets Dispersed in Aqueous Solutions. J. ColloidInterface Sci., 206(1):346�349, October 1998.

[61] Juan Guan, Bo Wang, and Steve Granick. Even Hard-Sphere Colloidal Suspensions Display FickianYet Non-Gaussian Di�usion. ACS Nano, 8(4):3331�3336, 2014.

91

[62] W D Harkins and E Boyd. The States of Monolayers. J. Phys. Chem., 45(1):20�43, 1941.

[63] Michael J Heller. DNA microarray Technology: Devices, Systems, and Applications. Annu. Rev.Biomed. Eng., 4:129�153, January 2002.

[64] R. Hogg, T. W. Healy, and D. W. Fuerstenau. Mutual Coagulation of Colloidal Dispersions. Trans.Faraday Soc., 62(615):1638, 1966.

[65] Andrei Honciuc, Denver Jn Baptiste, Ian P Campbell, and Daniel K Schwartz. Solvent Dependenceof the Activation Energy of Attachment Determined by Single Molecule Observations of SurfactantAdsorption. Langmuir, 25(13):7389�7392, July 2009.

[66] Andrei Honciuc, Adam W Harant, and Daniel K Schwartz. Single-Molecule Observations of SurfactantDi�usion at the Solution-Solid Interface. Langmuir, 24(13):6562�6566, June 2008.

[67] Andrei Honciuc, Alexander L. Howard, and Daniel K. Schwartz. Single Molecule Observations of FattyAcid Adsorption at the Silica/Water Interface: Activation Energy of Attachment. J. Phys. Chem. C,113(6):2078�2081, February 2009.

[68] Fredrik Höök, J Väräs, M Rodahl, R Kurrat, P Böni, J J Ramsden, M Textor, N D Spencer, P Tengvall,J Gold, B Kasemo, and János Vörös. A Comparative Study of Protein Adsorption on Titanium OxideSurfaces using in situ Ellipsometry, Optical Waveguide Lightmode Spectroscopy, and Quartz CrystalMicrobalance/Dissipation. Colloids Surf., B, 24(2):155�170, 2002.

[69] B. D. Hughes, B. A. Pailthorpe, and L. R. White. The Translational and Rotational Drag on a CylinderMoving in a Membrane, 1981.

[70] J. N. Israelachvili. Intermolecular and Surface Forces. Academic Press Inc., San Diego, second edition,1992.

[71] Jing Jin, Wei Jiang, Jinghua Yin, Xiangling Ji, and Paola Stagnaro. Plasma Proteins AdsorptionMechanism on Polyethylene-Grafted Poly(Ethylene Glycol) Surface by Quartz Crystal Microbalancewith Dissipation. Langmuir, 29(22):6624�6633, 2013.

[72] Daniel B Jones and Anton P J Middelberg. Direct Determination of the Mechanical Properties of anInterfacially Adsorbed Protein Film. Chem. Eng. Sci., 57:1711�1722, 2002.

[73] Latoya S Jones, Allyn Kaufmann, and C Russell Middaugh. Silicone Oil Induced Aggregation ofProteins. J. Pharm. Sci., 94(4):918�927, April 2005.

[74] Omkar Joshi, Liping Chu, Joseph McGuire, and D. Q. Wang. Adsorption and Function of RecombinantFactor VIII at the Air-Water Interface in the Presence of Tween 80. J. Pharm. Sci., 98(9):3099�3107,2009.

[75] Omkar Joshi, Joseph Mcguire, and D. Q. Wang. Adsorption and Function of Recombinant Factor VIIIat Solid-Water Interfaces in the Presence of Tween-80. J. Pharm. Sci., 97(11):4741�4755, 2008.

[76] Mark Kastantin, Blake B Langdon, Erin L Chang, and Daniel K Schwartz. Single-Molecule Resolutionof Interfacial Fibrinogen Behavior: E�ects of Oligomer Populations and Surface Chemistry. J. Am.Chem. Soc., 133(13):4975�4983, April 2011.

[77] Mark Kastantin, Blake B Langdon, and Daniel K Schwartz. A Bottom-Up Approach to UnderstandingProtein Layer Formation at Solid-Liquid Interfaces. Adv. Colloid Interface Sci., 207:240�252, May 2014.

[78] Mark Kastantin and Daniel K Schwartz. Distinguishing positional uncertainty from true mobility insingle-molecule trajectories that exhibit multiple di�usive modes. Microsc. Microanal., 18(4):793�797,August 2012.

92

[79] Mark Kastantin and D.K. Schwartz. Connecting Rare DNA Conformations and Surface DynamicsUsing Single-Molecule Resonance Energy Transfer. ACS Nano, 5(12):9861�9869, 2011.

[80] Mark Kastantin, Robert Walder, and Daniel K Schwartz. Identifying Mechanisms of Interfacial Dy-namics using Single-Molecule Tracking. Langmuir, 28(34):12443�12456, August 2012.

[81] Maneephan Keerati-u rai, Matteo Miriani, Stefania Iametti, Francesco Bonomi, and Milena Corredig.Structural Changes of Soy Proteins at the Oil-Water Interface Studied by Fluorescence Spectroscopy.Colloids Surf., B, 93:41�48, May 2012.

[82] Sharon M Kelly, Thomas J Jess, and Nicholas C Price. How to Study Proteins by Circular Dichroism.Biochim. Biophys. Acta, 1751(2):119�139, August 2005.

[83] B A Kerwin, M C Heller, S H Levin, and T W Randolph. E�ects of Tween 80 and Sucrose on AcuteShort-Term Stability and Long-Term Storage at -20 C of a Recombinant Hemoglobin. J. Pharm. Sci.,87(9):1062�1068, 1998.

[84] Bruce A. Kerwin. Polysorbates 20 and 80 Used in the Formulation of Protein Biotherapeutics: Struc-ture and Degradation Pathways, 2008.

[85] Sylvia Kiese and Astrid Papppenberger. Shaken, Not Stirred: Mechanical Stress Testing of an IgG1Antibody. J. Pharm. Sci., 97(10):4347�4366, 2008.

[86] David T Kim, Harvey W Blanch, and Clayton J Radke. Direct Imaging of Lysozyme Adsorption ontoMica by Atomic Force Microscopy. Langmuir, 18(15):5841�5850, 2002.

[87] M I Kinsinger, B Sun, Nicholas L Abbott, and D M Lynn. Reversible Control of Ordering Transitions atAqueous/Liquid Crystal Interfaces Using Functional Amphiphilic Polymers. Adv. Mater., 19(23):4208�4212, December 2007.

[88] Michael I Kinsinger, Maren E Buck, Fernando Campos, David M Lynn, and Nicholas L Abbott.Dynamic Ordering Transitions of Liquid Crystals Driven by Interfacial Complexes Formed BetweenPolyanions and Amphiphilic Polyamines. Langmuir, 24(23):13231�13236, December 2008.

[89] Janine Kirby, Paul R Heath, Pamela J Shaw, and Freddie C Hamdy. Gene Expression Assays, 2007.

[90] C M Knobler. Recent Developments in the Study of Monolayers at the Air-Water-Interface. Adv.Chem. Phys., 77:397�449, 1990.

[91] Kaloian Koynov and Hans-Jürgen Butt. Fluorescence Correlation Spectroscopy in Colloid and InterfaceScience. Curr. Opin. Colloid Interface Sci., 17(6):377�387, December 2012.

[92] J Kragel, R Wustneck, F Husband, Pj Wilde, Av Makievski, Do Grigoriev, and Jb Li. Properties ofMixed Protein / Surfactant Adsorption Layers. Colloids Surf., B, 12:399�407, 1999.

[93] L Kreilgaard, L S Jones, T W Randolph, S Frokjaer, J M Flink, M C Manning, and J F Carpenter.E�ect of Tween 20 on Freeze- Thawing- and Agitation-Induced Aggregation of Recombinant HumanFactor XIII. J. Pharm. Sci., 87(12):1597�1603, December 1998.

[94] Siok Lian S.L. Lai, Deny Hartono, and Kun-Lin K.L. Yang. Self-Assembly of Cholesterol DNA at LiquidCrystal/Aqueous Interface and its Application for DNA Detection. Appl. Phys. Lett., 95(15):153702,2009.

[95] Siok Lian S.L. Lai and Kun-Lin K.L. Yang. Detecting DNA targets through the formation ofDNA/CTAB complex and its interactions with liquid crystals. Analyst, 136(16):3329�34, August2011.

[96] Blake B Langdon, Mark Kastantin, and Daniel K Schwartz. Apparent Activation Energies Associatedwith Protein Dynamics on Hydrophobic and Hydrophilic Surfaces. Biophys. J., 102(11):2625�2633,June 2012.

93

[97] Blake B. Langdon, Mark Kastantin, Robert Walder, and Daniel K. Schwartz. Interfacial protein-proteinassociations. Biomacromolecules, 15:66�74, 2014.

[98] Helén Larsericsdotter, Sven Oscarsson, and Jos Buijs. Structure, Stability, and Orientation of BSAAdsorbed to Silica. J. Colloid Interface Sci., 289(1):26�35, September 2005.

[99] Hsien-Ming Lee, Yeong Woo Kim, and James K Baird. Electrophoretic Mobility and Zeta-Potential ofLysozyme Crystals in Aqueous Solutions of Some 1 : 1 Electrolytes. J. Cryst. Growth, 232:294�300,2001.

[100] Hyo Jin Lee, Arnold McAuley, Karl F. Schilke, and Joseph McGuire. Molecular Origins of Surfactant-Mediated Stabilization of Protein Drugs. Adv. Drug Deliv. Rev., 63(13):1160�1171, 2011.

[101] Jinjiang Li, Swathi Pinnamaneni, Yong Quan, Archana Jaiswal, Fredrik I Andersson, and XiaochunZhang. Mechanistic Understanding of Protein-Silicone Oil Interactions. Pharm. Res., 29(6):1689�1697,June 2012.

[102] Chia-Sheng Liao, Gwo-Bin Lee, Hsiao-Sheng Liu, Tsung-Min Hsieh, and Ching-Hsing Luo. MiniatureRT-PCR System for Diagnosis of RNA-Based Viruses. Nucleic Acids Res., 33(18):e156, January 2005.

[103] Juanjuan Liu, Shihui Zou, Sha Li, Xiaofeng Liao, Yejiang Hong, Liping Xiao, and Jie Fan. A GeneralSynthesis of Mesoporous Metal Oxides with Well-Dispersed Metal Nanoparticles via a Versatile Sol-GelProcess. J. Mater. Chem. A, 1(12):4038�4047, 2013.

[104] D Brett Ludwig, John F Carpenter, Jean-bernard Hamel, and Theodore W Randolph. Protein Adsorp-tion and Excipient E�ects on Kinetic Stability of Silicone Oil Emulsions. J. Pharm. Sci., 99(4):1721�1733, April 2010.

[105] D Brett Ludwig, Joseph T Trotter, John P Gabrielson, John F Carpenter, and Theodore W Randolph.Flow Cytometry: A Promising Technique for the Study of Silicone Oil-Induced Particulate Formationin Protein Formulations. Anal. Biochem., 410(2):191�199, March 2011.

[106] Shlomo Magdassi. Surface Activity of Proteins. CRC Press, Boca Raton, �rst edition, 1996.

[107] Hanns Christian Mahler, Robert Müller, Wolfgang Frieÿ, Aurelie Delille, and Susanne Matheus. Induc-tion and Analysis of Aggregates in a Liquid IgG1-Antibody Formulation. Eur. J. Pharm. Biopharm.,59(3):407�417, 2005.

[108] Hanns-Christian Mahler, Frank Senner, Karsten Maeder, and Robert Mueller. Surface Activity of aMonoclonal Antibody. J. Pharm. Sci., 98(12):4525�4533, December 2009.

[109] Sumit Majumdar, Brandi M Ford, Kevin D Mar, Vince J Sullivan, Robert G Ulrich, and Ajit Joseph MD'Souza. Evaluation of the E�ect of Syringe Surfaces on Protein Formulations. J. Pharm. Sci.,100(7):2563�2573, July 2011.

[110] Julia Maldonado-Valderrama, Valentin B. Fainerman, Eugene Aksenenko, M. Jose Gálvez-Ruiz,Miguel A. Cabrerizo-Vilchez, and Reinhard Miller. Dynamics of Protein Adsorption at the Oil-WaterInterface: Comparison with a Theoretical Model. Colloids Surf., A, 261:85�92, July 2005.

[111] S.M. Malone and D.K. Schwartz. Macroscopic Liquid Crystal Response to Isolated DNA Helices.Langmuir, 27(19):11767�11772, 2011.

[112] M Mandelkern. The Dimensions of DNA in Solution. J. Mol. Biol., 152(1):153�161, October 1981.

[113] Johannes Mathes and Wolfgang Friess. In�uence of pH and Ionic Strength on IgG Adsorption to Vials.Eur. J. Pharm. Biopharm., 78(2):239�247, 2011.

94

[114] Shyam B. Mehta, Rachael Lewus, Jared S. Bee, Theodore W. Randolph, and John F. Carpenter.Gelation of a Monoclonal Antibody at the Silicone Oil-Water Interface and Subsequent Rupture ofthe Interfacial Gel Results in Aggregation and Particle Formation. J. Pharm. Sci., 104(4):1282�1290,2015.

[115] Maria-Victoria Meli, I-Hsin Lin, and Nicholas L Abbott. Preparation of Microscopic and Planar Oil-Water Interfaces that are Decorated with Prescribed Densities of Insoluble Amphiphiles. J. Am. Chem.Soc., 130(13):4326�4333, April 2008.

[116] R Miller, V B Fainerman, a V Makievski, J Krägel, D O Grigoriev, V N Kazakov, and O V Sinyachenko.Dynamics of Protein and Mixed Protein/Surfactant Adsorption Layers at the Water/Fluid Interface.Adv. Colloid Interface Sci., 86:39�82, May 2000.

[117] Dietmar Möbius and Reinhard Miller, editors. Proteins at Liquid Interfaces. Elsevier, New York,seventh edition, 1998.

[118] H Mohwald. Phospholipid and Phospholipid-Protein Monolayers at the Air/Water Interface. Annu.Rev. Phys. Chem., 41:441�476, 1990.

[119] Jon H Monserud and Daniel K Schwartz. E�ects of Molecular Size and Surface Hydrophobicity onOligonucleotide Interfacial Dynamics. Biomacromolecules, 13(12):4002�4011, 2012.

[120] Grant A Myers, Daniel A Gacek, Eric M Peterson, Christopher B Fox, and Joel M Harris. MicroscopicRates of Peptide-Phospholipid Bilayer Interactions from Single-Molecule Residence Times. J. Am.Chem. Soc., 134(48):19652�19660, December 2012.

[121] Nathaniel Nelson and Daniel K Schwartz. Speci�c Ion (Hofmeister) E�ects on Adsorption, Desorption,and Di�usion at the Solid-Aqueous Interface. Phys. Chem. Lett., pages 4064�4068, 2013.

[122] Angelika Niemz, Tanya M Ferguson, and David S Boyle. Point-of-Care Nucleic Acid Testing forInfectious Diseases. Trends Biotechnol., 29(5):240�250, May 2011.

[123] Patrick S Noonan, Amit Shavit, Bharat R Acharya, and Daniel K Schwartz. Mixed Alkylsilane Func-tionalized Surfaces for Simultaneous Wetting and Homeotropic Anchoring of Liquid Crystals. ACSAppl. Mater. Interfaces, 3(11):4374�4380, November 2011.

[124] Vaheh Oganesyan, Melissa M. Damschroder, William Leach, Herren Wu, and William F. Dall'Acqua.Structural Characterization of a Mutated, ADCC-Enhanced Human Fc Fragment. Mol. Immunol.,45(7):1872�1882, 2008.

[125] Nr Pallas and Ba Pethica. Liquid-Expanded to Liquid-Condensed Transitions Monolayers at theAir/Water Interface. Langmuir, 1:509�513, 1985.

[126] Lalit M. Pandey, Sudip K. Pattanayek, and Didier Delabouglise. Properties of Adsorbed Bovine SerumAlbumin and Fibrinogen on Self-Assembled Monolayers. J. Phys. Chem. C, 117(12):6151�6160, 2013.

[127] Seungkyung Park, Yi Zhang, Shin Lin, Tza Huei Wang, and Samuel Yang. Advances in Micro�uidicPCR for Point-of-Care Infectious Disease Diagnostics. Biotechnol. Adv., 29(6):830�839, 2011.

[128] Avanish S Parmar and Martin Muschol. Hydration and Hydrodynamic Interactions of Lysozyme:E�ects of Chaotropic Versus Kosmotropic Ions. Biophys. J., 97(2):590�598, July 2009.

[129] Jordan T. Petkov, Theodor D. Gurkov, Bruce E. Campbell, and Rajendra P. Borwankar. Dilatationaland Shear Elasticity of Gel-Like Protein Layers on Air/Water Interface. Langmuir, 16(3):3703�3711,2000.

[130] Manuel Pinelo, Carme Ferrer, Anne S. Meyer, and Gunnar Jonsson. Controlling the Rejection ofProtein During Membrane Filtration by Adding Selected Polyelectrolytes. Sep. Purif. Technol., 85:54�60, February 2012.

95

[131] Dinesh Prasad and Ambarish Sharan Vidyarthi. DNA Based Methods Used for Characterization andDetection of Food Borne Bacterial Pathogens with Special Consideration to Recent Rapid Methods.African J. Biotechnol., 8(9):1768�1775, 2009.

[132] A.D. Price and D.K. Schwartz. DNA Hybridization-Induced Reorientation of Liquid Crystal Anchoringat the Nematic Liquid Crystal/Aqueous Interface. J. Am. Chem. Soc., 130(26):8188�8194, 2008.

[133] A M Puertas and T Voigtmann. Microrheology of Colloidal Systems. J. Phys. Condens. Matter,26(24):243101, July 2014.

[134] Michael Rabe, Dorinel Verdes, and Stefan Seeger. Understanding Protein Adsorption Phenomena atSolid Surfaces. Adv. Colloid Interface Sci., 162(1-2):87�106, February 2011.

[135] V Rampon, C Genot, A Riaublanc, M Anton, M A V Axelos, and D J McClements. Front-face�uorescence spectroscopy study of globular proteins in emulsions: displacement of BSA by a nonionicsurfactant. J. Agric. Food Chem., 51(9):2482�9, April 2003.

[136] T Randolph and L Jones. Surfactant-Protein Interactions In, Rational Design of Stable Protein Formulations.Springer, 2002.

[137] KB Ranger. The Circular Disk Straddling the Interface of a Two-Phase Flow. Int. J. Multiph. Flow,4:263�277, 1978.

[138] Paul Roach, David Farrar, and Carole C. Perry. Interpretation of Protein Adsorption: Surface-InducedConformational Changes. J. Am. Chem. Soc., 127(22):8168�8173, 2005.

[139] Christopher J. Roberts. Non-Native Protein Aggregation Kinetics, 2007.

[140] Sergii Rudiuk, Laetitia Cohen-Tannoudji, Sylvain Huille, and Christophe Tribet. Importance of theDynamics of Adsorption and of a Transient Interfacial Stress on the Formation of Aggregates of IgGAntibodies. Soft Matter, 8(9):2651, 2012.

[141] PG Sa�man and M. Delbrück. Brownian Motion in Biological Membranes. Proc. Natl. Acad. Sci.U.S.A., 72(8):3111, August 1975.

[142] Sema Salgin, UÇ�ur Salgin, and Seda Bahadir. Zeta Potentials and Isoelectric Points of Biomolecules:The E�ects of Ion Types and Ionic Strengths. Int. J. Electrochem. Sci., 7:12404�12414, 2012.

[143] Murali Sastry, Ashavani Kumar, Vidya Ramakrishnan, Krishna N. Ganesh, and Mrunalini Pattarkine.Cationic Surfactant Mediated Hybridization and Hydrophobization of DNA Molecules at the Liq-uid/Liquid Interface and their Phase Transfer. Chem. Commun., (16):1434�1435, 2001.

[144] Murali Sastry, Vidya Ramakrishnan, Mrunalini Pattarkine, Anand Gole, and K. N. Ganesh. Hybridiza-tion of DNA by Sequential Immobilization of Oligonucleotides at the Air-Water Interface. Langmuir,16(24):9142�9146, November 2000.

[145] M Schena, D Shalon, R W Davis, and P O Brown. Quantitative Monitoring of Gene ExpressionPatterns with a Complementary DNA Microarray. Science, 270(5235):467�470, 1995.

[146] K. Sharp, A Nicholls, R. Fine, and B Honig. Reconciling the Magnitude of the Microscopic andMacroscopic Hydrophobic E�ects. Science, 252(5002):106�109, April 1991.

[147] S K Singh, N Afonina, M Awwad, K Bechtold-Peters, J T Blue, D Chou, M Cromwell, H J Krause,H C Mahler, B K Meyer, L Narhi, D P Nesta, and T Spitznagel. An Industry Perspective on theMonitoring of Subvisible Particles as a Quality Attribute for Protein Therapeutics. J. Pharm. Sci.,99(8):3302�3321, 2010.

[148] Michael J. Skaug, Joshua Mabry, and Daniel K. Schwartz. Intermittent molecular hopping at thesolid-liquid interface. Phys. Rev. Lett., 110(25):1�5, 2013.

96

[149] Michael J Skaug, Joshua N Mabry, and Daniel K Schwartz. Single-Molecule Tracking of PolymerSurface Di�usion. J. Am. Chem. Soc., 136(4):1327�1332, January 2014.

[150] V Sluzky, J a Tamada, a M Klibanov, and R Langer. Kinetics of Insulin Aggregation in AqueousSolutions Upon Agitation in the Presence of Hydrophobic Surfaces. Proc. Natl. Acad. Sci. U.S.A.,88(21):9377�9381, 1991.

[151] DM Soumpasis. Theoretical Analysis of Fluorescence Photobleaching Recovery Experiments. Biophys.J., 41(1):95�97, 1983.

[152] Indira Sriram, Robert Walder, and Daniel K. Schwartz. Stokes-Einstein and Desorptiom-MediatedDi�usion of Protein Molecules at the Oil-Water Interface. Soft Matter, 8(22):6000�6003, 2012.

[153] Renuka Thirumangalathu, Sampathkumar Krishnan, Margaret Speed Ricci, David N Brems,Theodore W Randolph, and John F Carpenter. Silicone Oil- and Agitation-Induced Aggregationof a Monoclonal Antibody in Aqueous Solution. J. Pharm. Sci., 98(9):3167�3181, September 2009.

[154] Elvira Tjipto, Katie D Cadwell, John F Quinn, Angus P R Johnston, Nicholas L Abbott, and FrankCaruso. Tailoring the Interfaces Between Nematic Liquid Crystal Emulsions and Aqueous Phases viaLayer-by-Layer Assembly. Nano Lett., 6(10):2243�2248, October 2006.

[155] Sachin S Vaidya and Robert Y Ofoli. Adsorption and Interaction of Fibronectin and Human SerumAlbumin at the Liquid-Liquid Interface. Langmuir, 21(13):5852�5858, June 2005.

[156] Ank Valstar, Mats Almgren, Wyn Brown, and Marilena Vasilescu. The Interaction of Bovine SerumAlbumin with Surfactants Studied by Light Scattering. Langmuir, 16(3):922�927, February 2000.

[157] M.N. Velasco-Garcia and T. Mottram. Biosensor Technology Addressing Aagricultural Problems.Biosyst. Eng., 84(1):1�12, January 2003.

[158] a. W P Vermeer, M. G E G Bremer, and Willem Norde. Structural Changes of IgG Induced byHeat Treatment and by Adsorption onto a Hydrophobic Te�on Surface Studied by Circular Dichroismspectroscopy. Biochim. Biophys. Acta - Gen. Subj., 1425(1):1�12, 1998.

[159] Christina Vessely, Tia Estey, Theodore W. Randolph, Ian Henderson, Rajiv Nayar, and John F. Car-penter. E�ects of Solution Conditions and Surface Chemistry on the Adsorption of Three RecombinantBotulinum Neurotoxin Antigens to Aluminum Salt Adjuvants. J. Pharm. Sci., 96(9):2375�2389, 2007.

[160] Emanuele Vignati, Roberto Piazza, and Thomas P. Lockhart. Pickering Emulsions: Interfacial Tension,Colloidal Layer Morphology, and Trapped-Particle Motion. Langmuir, 19(6):6650�6656, 2003.

[161] T. A. Vilgis, G. Heinrich, and M. Kluppel. Reinforcement of Polymer Nano-Composites - Theory, Experiments and Applications.Cambridge Univ. Press, 2009.

[162] Erwin A. Vogler. Protein Adsorption in Three Dimensions. Biomaterials, 33(5):1201�1237, 2012.

[163] David Walba, Charles Liberko, Eva Korblova, Matthew Farrow, Thomas Furtak, Bruce Chow, Daniel KSchwartz, Adam Freeman, Kenneth Douglas, Scott Williams, and Others. Self-Assembled Monolayersfor Liquid Crystal Alignment: Simple Preparation on Glass uUing Alkyltrialkoxysilanes. Liq. Cryst.,31(4):481�489, April 2004.

[164] R. Walder, N. Nelson, and D.K. Schwartz. Super-Resolution Surface Mapping Using the Trajectoriesof Molecular Probes. Nat. Commun., 2:515, 2011.

[165] Robert Walder, Mark Kastantin, and Daniel K Schwartz. High Throughput Single Molecule Trackingfor Analysis of Rare Populations and Events. Analyst, 137(13):2987�2996, July 2012.

[166] Robert Walder, Nathaniel Nelson, and Daniel Schwartz. Single Molecule Observations of Desorption-Mediated Di�usion at the Solid-Liquid Interface. Phys. Rev. Lett., 107(15):156102, October 2011.

97

[167] Robert Walder and Daniel K. Schwartz. Single Molecule Observations of Multiple Protein Populationsat the Oil-Water Interface. Langmuir, 26(16):13364�13367, August 2010.

[168] Robert Walder and Daniel K. Schwartz. Dynamics of Protein Aggregation at the Oil-Water InterfaceCharacterized by Single Molecule TIRF Microscopy. Soft Matter, 7(17):7616�7622, 2011.

[169] Robert B Walder, Andrei Honciuc, and Daniel K Schwartz. Phospholipid Di�usion at the Oil-WaterInterface. J. Phys. Chem. B, 114(35):11484�11488, September 2010.

[170] A. F. H. Ward and L. Tordai. Time-Dependence of Boundary Tensions of Solutions I. The Role ofDi�usion in Time-E�ects. J. Chem. Phys., 14(7):453�461, 1946.

[171] Christian F. Wertz and Maria M. Santore. Adsorption and Reorientation Kinetics of Lysozyme onHydrophobic Surfaces. Langmuir, 18(4):1190�1199, 2002.

[172] Stephen R. Williams, G. Bryant, I. K. Snook, and W. Van Megen. Velocity Autocorrelation Functionsof Hard-Sphere Fluids: Long-Time Tails Upon Undercooling. Phys. Rev. Lett., 96(March):3�6, 2006.

[173] Benjamin T. Wong, Jiali Zhai, Søren V. Ho�mann, Marie Isabel Aguilar, MaryAnn Augustin, Tim J.Wooster, and Li Day. Conformational Changes to Deamidated Wheat Gliadins and β-casein UponAdsorption to Oil-Water Emulsion Interfaces. Food Hydrocoll., 27(1):91�101, 2012.

[174] Xiaoyu Wu and Ganesan Narsimhan. E�ect of surface concentration on secondary and tertiary confor-mational changes of lysozyme adsorbed on silica nanoparticles. Biochim. Biophys. Acta, 1784(11):1694�701, November 2008.

[175] Fan Xia, Xiaolei Zuo, Renqiang Yang, Yi Xiao, Di Kang, Alexis Vallee-Belisle, Xiong Gong, Jonathan DYuen, Ben B. Y Hsu, Alan J Heeger, and Kevin W Plaxco. Colorimetric Detection of DNA, SmallMolecules, Proteins, and Ions Using Unmodi�ed Gold Nanoparticles and Conjugated Polyelectrolytes.Proc. Natl. Acad. Sci. U.S.A., 107:10837�10841, 2010.

[176] Galina Yampolskaya and Dimo Platikanov. Proteins at Fluid Interfaces: Aadsorption Layers and ThinLiquid Films. Adv. Colloid Interface Sci., 128-130(2006):159�183, December 2006.

[177] Linling Yu, Lin Zhang, and Yan Sun. Protein Behavior at Surfaces: Orientation, ConformationalTransitions and Transport. J. Chromatogr. A, 1382:118�134, 2015.

[178] Jiali Zhai, Søren V Ho�mann, Li Day, Tzong-Hsien Lee, Mary Ann Augustin, Marie-Isabel Aguilar,and Tim J Wooster. Conformational Changes of α-Lactalbumin Adsorbed at Oil-Water Interfaces:Interplay between Protein Structure and Emulsion Stability. Langmuir, 28(5):2357�2367, February2012.

[179] Jianhua Zou, Tanmay Bera, Alicia a Davis, Wenlang Liang, and Jiyu Fang. Director Con�gurationTransitions of Polyelectrolyte Coated Liquid-Crystal Droplets. J. Phys. Chem. B, 115(29):8970�4, July2011.

understanding how non-coalenvt interactions a ect

Documents