Molecular Electromechanics: Modeling Electrostatic Forces Between GAG Molecules by Delphine Marguerite Denise Dean Submitted to the Department of Electrical Engineering and Computer Science in partial fulfillment of the requirements for the degrees of Bachelor of Science in Electrical Engineering and Computer Science and Master of Engineering in Electrical Engineering and Computer Science at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY May 2001 Copyright 2001 Delphine Marguerite Denise Dean. All rights reserved. The author hereby grants to M.I.T. permission to reproduce and distribute publicly paper and electronic copies of this thesis and to MAS T grant others the right to do so. S NOLOGY JUL 112001 LIBRARIES Author .......................................... Department oj.Electrical Engineering and Computer Science May 23, 2001 Certified by ....... Professor Accepted by... ............. . ...... ........... Alan J. Grodzinsky Electric ,echanicl, and Bioengineering ThesisSupervisor ..................... Arthur C. Smith Chairman, Department Committee on Graduate Students 2 Molecular Electromechanics: Modeling Electrostatic Forces Between GAG Molecules by Delphine Marguerite Denise Dean Submitted to the Department of Electrical Engineering and Computer Science on May 23, 2001, in partial fulfillment of the requirements for the degrees of Bachelor of Science in Electrical Engineering and Computer Science and Master of Engineering in Electrical Engineering and Computer Science Abstract Electrostatic repulsions between glycosaminoglycan (GAG) molecules are the principle contributors to the compressive strength of articular cartilage tissue. These molecular-level repulsion forces can be measured in aqueous solutions under various conditions with a molecular force probe (MFP). In order to determine the contribution of electrostatic repulsion and to understand the significance of the components of the total force, accurate models need to be developed. In this thesis, three different models of the electrostatic interactions are investigated. The first models a GAG "brush"-like layer as a smooth flat surface charge. This model has a simple analytical solution in certain limits and provides a qualitative comparison of ionic strength dependent trends. The second model treats a GAG layer as a uniform volume charge density. Since the length of the GAG molecule is comparable to the distance at which the MFP measures forces, this volume charge model is a more appropriate model than the flat surface model for force calculations. The third is a novel model that represents the GAG molecules as rods of uniform charge density, which better describes the geometry of the system. Because the equations relating the electrostatic force to the known system parameters (charge density and ionic strength of the buffer) are nonlinear, there are no known analytical solutions for the last two models. Therefore, numerical methods were developed and their feasibility and results are reported. Thesis Supervisor: Alan J. Grodzinsky Title: Professor of Electrical, Mechanical, and Bioengineering 3 4 Acknowledgments This thesis would not have been possible without the help of many people. I would like to thank Professor Alan Grodzinsky, my thesis and academic advisor, for all the insight, support, and encouragement he has given me during the course of my studies. I would also like to acknowlege the crucial work of Joonil Seog in determining the chemistry for attaching the GAG to gold, running all the MFP experiments, and basically getting this project off the ground. Professor Christine Ortiz provided useful advice and references, and helped introduce me to fields in which I had little prior experience. I would like to thank Dr. Anna Plaas and Shirley Wong-Palms at the Shriners Hospital in Tampa, who provided the GAG chains and helped with the attachment chemistry. I also would like to thank Professor Laibinis and Ivan Lee for their insights on surface chemistry. To everyone in Prof. Grodzinsky's lab: thanks for your friendship and for making the lab such a nice place to work. N6ra Szisz helped me work out my theory derivations, read through my thesis and in general provided lots of moral support and friendship. I want to credit Moonsoo Jin for correctly predicting that my numerical simulation would take at least half a year to develop, and also for his valuable suggestions along the way. I've enjoyed the company of my office mate Han-Hwa Huang, who was always willing to lend a hand in lab-related matters. Linda Bragman provided assistance with several administrative problems and supplied my inbox with many humorous emails. Thanks also to Jiang-Ti Kong, who started the work on this project. I also want to thank everyone in Prof. Ortiz's lab for all the great Monday meetings and discussions, especially Monica Rixman who put up with listening to me rant about current publications on electrostatics. Professor Jacob White taught a great numerical methods class and helped me with the Jacobian-free Newton method. Jonathan Blanford, Mathew Cain, Shannon Cheng, Timothy Garnett, Douglas Heimburger, May Lim, Michelle Nadermann, James Paris, Natalia Toro, Qian Wang, and Boriz Zbarsky are all very nice people who gladly donated computer processor time to run my program in exchange for chocolate 5 cake. Justin Cave, Anne Gaumond, Raffi Krikorian, Julie Park and Michael Spitznagel are also very nice and not only did they run my program on their computers but they also proofread this thesis and found all (hopefully!) of my errors. Benjamin Yoder is especially nice and read over my thesis, helped me debug my program and ran it on his "little" computer farm in the Media Lab. I want to thank Cecile Le Cocq, my sister, who let me take naps at her dorm on campus during those long writing sessions, and who gave me lots of moral support. Catherine and Christian Le Cocq, my parents, who are all-around really cool and super parents, ran my program, proofread this document, listened to my late night rants about GAGs and still managed to stay sane! Finally, I would like to thank Brian Dean, my husband, who helped me debug, read my thesis, discussed my ideas, made me dinner every night for the last three months, and was always there for me when I stressed out. And to you the reader, thanks for reading! 6 Contents 1 Introduction 17 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.2 O bjective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.3 O verview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2 Articular Cartilage 2.1 2.2 3 19 Cartilage Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1.1 Extracellular Matrix . . . . . . . . . . . . . . . . . . . . . . . 19 Proteoglycans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.2.1 21 Glycosaminoglycan . . . . . . . . . . . . . . . . . . . . . . . . Molecular Level Forces 23 3.1 Electrostatics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 3.1.1 The Poisson-Boltzmann Equation . . . . . . . . . . . . . . . . 23 3.1.2 Electrostatic Force . . . . . . . . . . . . . . . . . . . . . . . . 26 Non-Electrostatic Forces . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2 4 Experimental Methods 5 29 4.1 High Resolution Force Spectroscopy . . . . . . . . . . . . . . . . . . . 29 4.2 Experimental Setup and Results . . . . . . . . . . . . . . . . . . . . . 30 Modeling of Electrostatic Force 35 5.1 Smooth Surface Charge Model . . . . . . . . . . . . . . . . . . . . . . 35 5.1.1 35 Analytical Solution to Linearized Flat Surface Charge Model 7 . 5.1.2 5.2 37 . . . . . . . . . . . . . . . . . . . . . . . . . . 39 . . . . . . . . . 41 . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Numerical Method for Solving Charged Rod Model . . . . . . 44 Volume Charge Model 5.2.1 5.3 Numerical Method for Flat Surface Charge Model . . . . . . . Numerical Method for Volume Charge Model Charged Rod M odel 5.3.1 51 6 Results and Discussions 6.1 Smooth Surface Charge Model . . . . . . . . . . . . . . . . . . . . . . 51 6.2 Volume Charge Model . . . . . . . . . . . . . . . . . . . . . . . . . . 55 6.3 Charged Rod Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 6.4 Comparison of Results . . . . . . . . . . . . . . . . . . . . . . . . . . 67 71 7 Conclusions 7.1 Sum m ary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 7.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 A Electrostatic Interactions Between MFP Tip and Substrate: Generalized 2-D Models for "Neutral" and Charged Surfaces 73 . . . . . . . . . . . . . . . . . . . . . . . . . 75 A.2 Electrostatic Force from Induced Charge . . . . . . . . . . . . . . . . 76 A.1 Induced Surface Charge 79 B Detailed Experimental Methods 8 List of Figures 2-1 The structure of cartilage (not drawn to scale). . . . . . . . . . . . . 20 2-2 Schematic drawing of aggrecan, a proteoglycan. . . . . . . . . . . . . 21 2-3 (a) The structure of GAG; (b) Molecular structure of Chondroitin-6sulfate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3-1 Schematic of some of the forces felt by GAG molecules. . . . . . . . . 24 4-1 The molecular force probe (MFP). . . . . . . . . . . . . . . . . . . . 30 4-2 (a) An example of deflection curve; (b) An example of a force curve. . 31 4-3 Schematics of the experiments: (a) GAG coated substrate versus monolayer coated tip; (b) GAG coated substrate versus GAG coated tip. . 4-4 Experimental data of force between a charged monolayer coated tip and a GAG coated substrate at various concentrations of NaCl. 4-5 . . . . . . . . . . . . 36 Potential between two charged surfaces. The control box is shown with the dashed line. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3 33 (a) Two approaching planes with charged surfaces; (b) Charged hemisphere tip approaching a flat charged planar surface. 5-2 32 Experimental data of force between a neutral monolayer coated tip and a GAG coated substrate at various concentrations of NaCl. . . . . . . 5-1 31 39 (a) Flat surface charge over a fixed charge volume; (b) Two approaching surface with fixed charge volumes. . . . . . . . . . . . . . . . . . . 9 40 5-4 (a) A flat surface charge over a fixed charge volume a distance D > brush height apart; (b) A flat surface charge over a fixed charge volume . . . . . . . . . . . . . . . . . . . . . . 42 5-5 Approximated hemisphere of charge over a fixed volume of charge. . . 44 5-6 Two parallel plates with charged rods. . . . . . . . . . . . . . . . . . 44 5-7 Repeating unit in the parallel charged rod plate model. D is the dis- a distance D' < brush height. tance between the square base plates. This unit is discretized into cubic elements of size 6, by Jy by 62. 6-1 . . . . . . . . . . . . . . . . . . A comparison of the force predicted by the surface charge model using the full solution and the linearized approximation. . . . . . . . . . . . 6-2 54 ..................................... Force between a smooth volume of fixed charge and a flat charged surface (hnm x 1nm) at 0.01M. . . . . . . . . . . . . . . . . . . . . . 6-4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 Force between rods of fixed charge and a flat charged surface (1nm x 1nm ) at 0.01M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-9 59 The volume charge model of a GAG-coated tip approaching a GAGcoated surface. 6-8 58 The volume charge model compared to experimental data for GAG surface versus neutral monolayer-coated tip. . . . . . . . . . . . . . . 6-7 57 The volume charge model compared to experimental data for GAG surface versus charged monolayer-coated tip. . . . . . . . . . . . . . . 6-6 56 Force between a smooth volume of fixed charge and a flat charged hemisphere (radius=25nm) at 0.01M. . . . . . . . . . . . . . . . . . . 6-5 53 Surface charge model (nonlinear and linearized) versus experimental data........ 6-3 45 63 The charged rod model compared to the experimental results of a charged monolayer-coated tip approaching GAG surface. . . . . . . . 64 6-10 The charged rod model compared to experimental data for a neutral monolayer-coated tip approaching a GAG surface. . . . . . . . . . . . 10 65 6-11 Force between two approaching 1nm x Inm surfaces with charged rods at 0.01M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 6-12 The charged rod model of a GAG-coated tip approaching GAG surface. 66 6-13 Charged monolayer versus GAG force at 0.01M NaCl compared to 2 models with added approximated steric effect. . . . . . . . . . . . . . 67 6-14 Neutral monolayer versus GAG experimental results at 3M NaCl compared to the de Gennes steric model. . . . . . . . . . . . . . . . . . . 68 6-15 Comparison of the volume charge and charged rod models of GAGcoated tip approaching a GAG-coated substrate. . . . . . . . . . . . . A-i Model of tip approaching a charged surface. 70 . . . . . . . . . . . . . . 74 A-2 General interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 B-1 Schematic of the chemistry to graft GAG onto substrate surface. . . . 80 11 12 List of Tables 6.1 The Debye length at various NaCL concentrations. . . . . . . . . . . . 52 6.2 The flat surface charge model parameters. . . . . . . . . . . . . . . . 52 6.3 The smooth volume charge model parameters. . . . . . . . . . . . . . 56 6.4 The rod model parameters . . . . . . . . . . . . . . . . . . . . . . . . 62 13 14 List of Symbols AFM atomic force microscope CS chondroitin sulfate ECM extracellular matrix GAG glycosaminoglycan HRFS high resolution force spectroscopy MFP molecular force probe PB Poisson-Boltzmann PG proteoglycan SAM self-assembled monolayer r1 Debye length <D) electrostatic potential E electric field F force W energy T temperature CO bulk salt concentration p volume charge density a surface charge density 15 fo permittivity of free space 8.85 x 10-1 22 cf permittivity of fluid 6.923 x 10- N2 R gas constant Nm 8.314 molK F Faraday constant C 96484.6 mol k otmanconstant 1.381 x -1 0 3N 16 Chapter 1 Introduction 1.1 Background Articular cartilage is the load bearing tissue found in joints. Its strength in response to different loads is highly dependent on the composition of its extracellular matrix. It has been shown with macroscopic tissue-level experiments that much of its compressive strength is due to highly charged polyelectrolyte molecules known as glycosaminoglycans (GAGs). The GAGs are packed very closely in the tissue (with spacing approximately 2 - 4nm) and the electrostatic repulsion between them is thought to contribute 50-75% of the tissue's equilibrium modulus in compression. However, molecular level measurements of these forces have been only recently reported [24, 23, 20]. The development of high resolution force spectroscopy (HRFS) instruments, like the atomic force microscope (AFM), has made it possible to measure the nanoNewton scale interactions of molecules [14, 3]. In HRFS, the force is measured by moving a small cantilever or spring with an attached tip of known geometry towards a flat surface. As the tip approaches the surface, the interaction forces will deflect the cantilever. From this deflection measurement, one can calculate the force of interaction between the tip and the surface. Using a new type of HRFS known as a molecular force probe (MFP), the interactions between GAGs can be measured directly. 17 1.2 Objective The force between chondroitin-sulfate glycosaminoglycans (CS-GAGs) attached to a substrate and a charged cantilever tip was measured using a molecular force probe. To interpret these forces, models are needed which accurately represent the system. The goal of this thesis is to develop models for the electrostatic component of the force. By comparing these to the experimental results, one can determine which factors play the most important role in the interaction between GAGs. There has been some previous work done on modeling brush layers of polyelectrolytes [3, 19] although not specifically for the case of GAG molecules. These previous models are simple and lend themselves to efficient numerical solutions. However, their simplifications often do not reflect well the actual molecular geometry of the GAGGAG electrostatic interactions. In this thesis, a new and potentially more accurate model is described, and a different numerical technique is employed for its solution. These models are compared and evaluated with respect to initial experimental data in order to determine the optimal approach for modeling GAG-GAG interactions. 1.3 Overview This thesis is structured as follows. Chapter 2 gives background on the biological properties of cartilage. Chapter 3 describes the theory of some of the molecular level forces. In chapter 4, experimental methods and results are discussed. Chapter 5 describes in detail the three models which were implemented and tested. Chapter 6 gives the results of these models and compares them to some preliminary experimental results. Finally, chapter 7 summarizes the findings and touches on directions for future research. 18 Chapter 2 Articular Cartilage Articular cartilage is the load bearing connective tissue found on the surface of movable joints. It normally sustains high compressive loads, 10 - 20MPa, without damage. The biomechanical properties of cartilage, such as its high compressive resistance, are directly related to the molecular structure of extracellular macromolecules. 2.1 Cartilage Biology Cartilage is a very complex avascular and alymphatic tissue (Figure 2-1). Articular cartilage tissue is composed of 70-80% water by weight and contains only 20 - 40 thousand cells, called chondrocytes, per cubic millimeter [16, 17]. However, this low density of cells maintains the extracellular matrix (ECM) under normal conditions. 2.1.1 Extracellular Matrix The extracellular matrix provides most of the strength of cartilage tissue. The ECM is produced and maintained by the chondrocytes. While it is composed mostly of water, it is surprisingly strong mechanically. The different macromolecular structures of the ECM account for different mechanical properties. The collagen molecules contribute to the tissues' shear and tensile strength, while the highly charged proteoglycan molecules provide most of the com19 Hyaluronate Agggrecan ollagen Chondrocyte GAG chain ~~ Figure 2-1: The structure of cartilage (not drawn to scale). pressive strength. 2.2 Proteoglycans Proteoglycans make up 5-10% of the cartilage wet weight (35% by dry weight) [17, 25]. They are composed of a long core protein to which one or more glycosaminoglycan (GAG) chains are covalently bound. Aggrecan is the most abundant proteoglycan in cartilage (Figure 2-2). The core protein of aggrecan contains three globular domains. The first, GI, is found near the amino-terminal and uses a link protein to attach to the binding region of hyaluronan, a glycosaminoglycan consisting of several thousand repeating disaccharide units. The second globular domain, G2, is found further down on the core protein and the third globular domain, G3, is found near the carboxyl-terminal of the core protein. Between G2 and G3 is a highly charged glycosaminoglycan rich region. The core protein is several hundred nanometers long and ~ 300kDa. Proteoglycans, such as aggrecans, form large aggregates with hyaluronan (Figure 2-2). 20 CS / N-linked oligosaccharide Hyaluronate binding region 0-linked oligosaccharide Link protein Hyaluronate H a-r CS = Chondroitin sulfate KS-rich region KS = Keratan sulfate Figure 2-2: Schematic drawing of aggrecan, a proteoglycan. 2.2.1 Glycosaminoglycan Aggrecan contains three major types of glycosaminoglycans (GAG): chondroitin-6sulfate, chondroitin-4-sulfate, and keratan-sulfate. The keratan-sulfate chains, which are - 5kDa, are the shortest consisting of about 10 repeating disaccharides. These chains are located mainly near the G2 region of the core protein. The chondroitinsulfate (CS) chains occupy most of remaining GAG-rich region of aggrecan and comprises 95% of the molecular weight of the entire proteoglycan molecule (Figure 2-3). The chondroitin-sulfate chains are 30 - 40nm long and are spaced approximately 2 - 4nm apart on the core protein. One end of the CS is covalently linked to the core protein while the reducing end points into the intra-tissue space (Figure 2-3a). The CS-GAG chain is composed of alternating glucuronic acid and N-acetyl-6(or 4)-sulfate galactosamine (Figure 2-3b). Under normal physiological conditions, the carboxylic acid and the sulfate groups are negatively charged. It has been found that the sulfation of GAG chains, and therefore its charge, is decreased in cartilage disease, such as osteoarthritis [22]. Studies have shown with macroscopic measurements [16, 2] that the high negative charge of the GAG molecules is the major determinant of cartilage compressive loading properties, responsible for 21 Ser Xyl = 00- Serine H2OSO- HH = Xylose = Galactose Gal = Glucuronate GluA GlcNAc = N-Acetyl-galactosamine \ HH H HH H H==== H Sr-O-Xyl-Gal-Gal--fO-GluA-O-GlcNAc -n HH Glucuronate Chondroitin sulfate H NHCOCH3 N-Acetyl-galactosamine-6-sulfate (b) (a) Figure 2-3: (a) The structure of GAG; (b) Molecular structure of Chondroitin-6sulfate. 50-75% of the equilibrium modulus in compression. 22 Chapter 3 Molecular Level Forces Intermolecular interactions are complex and must be modeled differently depending on the distance and charge states of the molecules. Far from trying to give a general model that takes every kind of interaction into consideration, this short description will concentrate on the forces that are of interest for the modeling of the cartilage GAG molecules (Figure 3-1). 3.1 3.1.1 Electrostatics The Poisson-Boltzmann Equation The Poisson-Boltzmann equation gives an expression for the potential in space in an electrolyte solution [26, 10, 25]. This potential can then be used to compute the total electrostatic force. There are several simplifying assumptions made: the ions in solution are treated as point charges and therefore take up no volume. The permittivity, C, is everywhere the same as that of the bulk solution, ef, and is independent of any electric field. Start with Gauss's and Faraday's laws from Maxwell's equations, V-cfE V xE p = - 23 (3.1) at (pH) (3.2) Electrostatic force between opposing GAGs Electrostatic force between interdigitating GAGs Electrostatic force between neighboring GAGs Steric force Figure 3-1: Schematic of some of the forces felt by GAG molecules. where E is the electric field, p is the volume charge density, p is the magnetic permeability, and H is the magnetic field. If time varying magnetic fields are assumed to be negligible then the curl of the electric field is zero and a scalar potential, <b) (volts), can then be defined. This is known as the electro quasi- static approximation: V x E ~ 0, E =_- V D. (3.3) From 3.1 and 3.3, the potential and space charge distribution are then related by Poisson's equation: V2(b = (3.4) For biological tissues, the volume charge density will be due to the mobile ion con- centrations in solution, pi plus any added fixed charge density pfi., associated with macromolecules of the ECM. When the system is in thermodynamic equilibrium, pi can be related to the potential by the Boltzmann distribution, which represents the balance between ionic diffusion and electrical migration. Fick's first law describes the relationship between the flux of the solute due to 24 diffusion and its concentration gradient [9, 27]. Also, if the solute is charged, it will move in the presence of an electric field. The component of the flux due to this electrical migration is proportional to the concentration, ci. The total flux of species i, Fi, is due to both electrical migration and diffusion: zi -i (3.5) -piciE - DjVcj where zi is the valence, pi is the electrical mobility, Di is the ionic diffusion coefficient and ci is the concentration of species i. In equilibrium, the flux of each ionic species is zero, and so the electrical migration and diffusion fluxes must balance: z. piciE = DjVc. (3.6) lzil As stated above, the electric field can be written as the gradient of a potential: zi piciv4 =-DiVci, lzil and this equation can be rearranged then integrated to get an expression for the concentration profile of species i, ci : i = jV4) -Di( 1)Vci Ci l~il piV_ = -DjV(ln(c )) Pi4 + Constant = -Di ln(ci) _ lzil lzil using the Einstein relation, D = RT: zjFD + Constant = -RT ln(ci) =i (Constant)e-RT a z=F 25 When D = 0, the concentration of i must be the bulk concentration, so the concentration profile of species i at steady state will be: (3.7) ,T Ci = Cioe~ where cio is the bulk concentration of species i. Using 3.7, the charge density can be expressed as: ziFci + pix p = S ziFc 0 6tY ±TPfix (3.8) where zi is the valence and ci(x) is the concentration profile, cio is the bulk concentration of ion i, F is Faraday's constant, R is the gas constant, T is the temperature, and pfix is the fixed volume charge density. For a solution of monovalent salt such as NaCl having bulk concentration Co, equations 3.4 and 3.8, combine to give the Poisson-Boltzmann equation: 2 = 2FIC0 -1s . 2 F(D )h pf ix Pf (3.9) The Debye length, r,-, is defined to be: 6f RT r 2F 2 C, (3.10) This is the characteristic length over which the potential will decay. As shown in the equation, it varies inversely with the square root of salt concentration. 3.1.2 Electrostatic Force The total force on a surface can be calculated one of two ways: either by balancing the forces on a control volume, or by differentiating the total energy with respect to 26 the direction (the energy method). The force will be due to two terms: the osmotic term due to the ion concentrations gradient, and the Maxwell stress term due to the force of the electric field acting on the ion charge. The osmotic pressure due to the ion concentration relative to the bulk concentration at a distance r from the charged surface, where (D (=(r),for a monovalent salt solution is given by van't Hoff's law: Posm - P (ci(r) - ci (rrf)) = RT -';, The outward swelling pressure is: P.Smotic = RT(Coe- "+ Coe+-T) - RTCO = 2RTCO (cosh ( . The Maxwell stress tensor, defined as acting outward from the surface, is: Tig = EEJ Tie -c~jj j 1 Tzz -64 -EE, 2 = 1~ ~x EkEk) 2 6 ij (-I 1 2 E(E,2+ EY) The Maxwell pressure (defined as acting in on the surface) is the normal force that a charge feels due to the electric field: PMaxwell 2(V - 2 2. Thus, the total electrostatic force, Fe, per unit area in the z-direction will be the sum of these two pressure terms: Fe Area - 2RTCo cosh F \RT - 1) + 2 (V))2 .(3.11) Force, as always, can be written as the derivative of the energy, We: a Fe = -We Fe)z 27 (312) The energy will be the sum of the terms due to the fixed charges, the osmotic pressure due to the ion concentration gradient, and the Maxwell stress: W = Jf dS + p101 - -dV 2RTCo(cosh T) - 1) + 'f(V4)2) dV (3.13) where the first term is a surface integral and the other two terms are volume integrals. 3.2 Non-Electrostatic Forces There are of course other types of forces between molecules besides the electrostatic forces, including steric and other short range forces. Steric forces are associated with molecules that are being bent or stretched. Steric force may, however, be hard to seperate from electrostatic components because they can also be caused by intramolecular repulsion. Therefore, the steric component is sometimes hard to isolate. The repulsive force per area between two uncharged brushes has been previously described. According to the theory by de Gennes [6], the force per unit area on an uncharged brush of thickness h at separation D is: Ps(D) = F Area ~ S3 D _ 2h ,h)D<2h (3.14) where the brush is composed of end grafted polymers and the polymers are modeled as chains of s-sized sections. The first term is due to the non-ionic osmotic repulsion between the brushes while the second term comes from the elastic interactions of the chains. Since the GAG is a charged polymer, the total force could be approximated as the sum of the electrostatic force and this uncharged brush steric force. As molecules get closer and closer, their structure becomes of utmost importance. Their geometry, their charge distribution and their mechanical degrees of freedom are all affected. There is no general tool to describe very close interaction of long chain charged polymers. Close polymer interaction is a vast field of study in itself [6, 19]. 28 Chapter 4 Experimental Methods 4.1 High Resolution Force Spectroscopy With the development of high-resolution force spectroscopy (HRFS) instruments like the molecular force probe (MFP), it is possible to measure the nanoscale interaction forces between molecules [14, 7, 11, 24, 23]. AFMs have been popular tools for imaging surfaces, and recent work has shown that they are reliable tools for force measurements as well [1, 5]. The MFP (Figure 4-1) is very similar to the AFM but has been optimized specifically for the measurement of force. It overcomes some of the optical interference problems inherent to standard AFMs and has slightly better resolution. In HRFS, a fixed surface is slowly approached by a small cantilever or spring with a tip of known geometry and size (in these experiments, a hemisphere of 50nm in diameter) held by a piezoelectric ceramic translator. As the tip feels repulsive or attractive forces, the cantilever bends and its deflection is measured by a laser. Thus, a cantilever deflection versus piezoelectric actuator distance curve (deflection curve) is obtained. The cantilever spring constant is measured by analyzing its resonant frequencies. The data is then converted into a force versus distance curve (force curve) using the known cantilever spring constant (Figure 4-2). These measurement techniques are particularly useful for biological applications because they can be carried out in an aqueous environment without harming the sample. Therefore, using these 29 HEAD 15 rnLD IR laser mirror laser focusing te tis ho dr I tical and cariener UP ationt Xra lquid meniscus m-pl-- --- stage X o.ptical BASE microscope 0 objective Figure 4-1: The molecular force probe (MFP). Thsterasaientswroeb onlSo 2,2, 0 n ealdepaaino techniques, it is now possible to make molecular level measurements of the repulsion forces between the GAG molecules, leading to a better understaning of how cartilage behaves in a variety of conditions. 4.2 Experimental Setup and Results These experiments were done by Joonil Seog [24, 23, 20] and a detailed explanation of these methods is given in Appendix B. Glycosaminoglycan molecules (approximately 32nm in length) from rat cells were end grafted to a gold substrate (Figure 4-3). The density of GAG on the surface was approximately one chain per 6nm x 6nm area and the charge per GAG chain was approximately 1.097 x 10-17C. A gold coated AFM tip was coated with either a charged monolayer or a neutral monolayer (Figure 4-3a). One substrate was also coated with charged self-assembled monolayer (SAM). The force between it and the charged monolayer coated tip was measured in order to determine the monolayer surface charge density. It was found to be approximately 30 0 tip touching surface force 1-) Z 42 .,. .. ......... .. , repul ive force 0 piezoelectric distance (nm) (a) 0 0 no force " tip to surface distance (nm) (b) Figure 4-2: (a) An example of deflection curve; (b) An example of a force curve. Gold GAr D monolayer backfill SA t-GAG7 Gold(b) (a) Figure 4-3: Schematics of the experiments: (a) GAG coated substrate versus monolayer coated tip; (b) GAG coated substrate versus GAG coated tip. 31 ...... Charged tip versus GAG surface 2.5\ 2. 1.0M 1.5 0.01 M 0.001 M 0.0001 M 8 0.5 \- 0 50 100 150 distance (nm) Figure 4-4: Experimental data of force between a charged monolayer coated tip and a GAG coated substrate at various concentrations of NaCl. -8.16 x 10- . The forces between the modified tips and the GAG coated surface were measured at different NaCl concentrations (from 0.0001M to 3.OM) using a molecular force probe. Future experiments will be carried out with the tip also coated with GAG (Figure 4-3b). The measured force between the charged tip and GAG surface is shown in Figure 4-4. The majority of the force seems to be electrostatic in origin as the force decreases with increasing ionic strength. The force between the neutral tip and GAG surface is shown in Figure 4-5. The forces observed are smaller than those with the charged monolayer-coated tip. However, it is interesting to note that there still appears to be a long range force of electrostatic origin. Except for 0.0001M, the force gets smaller as the ionic strength increases. The forces measured at 0.0001M are slightly smaller 32 Neutral tip versus GAG surface 2 1.8 1.6 1.4 -3.OM 1.2 -0.1M 0.01 M S .. 001 M 2 0.0001 M- - 0.8 - 0.6 -- 0.4 -- 0.2 --- 0 10 20 40 30 50 60 70 distance (nm) Figure 4-5: Experimental data of force between a neutral monolayer coated tip and a GAG coated substrate at various concentrations of NaCl. than those at 0.001M possibly because there is an offset of the curve due to an incompressible layer of GAGs. This would result in the 0 distance being offset by the thickness of this incompressible layer. The 0.1M-3M results are very close together because the forces are only significant at a range when steric and non-electrostatic effects occur. Therefore, the change in ionic strength only produces a small change in the total force. The origin of this electrostatic force is discussed in detail in Appendix A. 33 34 Chapter 5 Modeling of Electrostatic Force To better understand the forces measured in the high resolution force spectroscopy experiments, models were developed to more accurately represent the system [23]. Because the electrostatic forces are related by a nonlinear differential equation, these models do not usually lead to a closed form analytical solution and, therefore, numerical techniques need to be employed to solve them. 5.1 Smooth Surface Charge Model The simplest model is one in which the glycosaminoglycan (GAG) coated surfaces are modeled as smooth, infinite planar surface charges (Figure 5-1a). Due to symmetry, this system is essentially a one-dimensional problem. 5.1.1 Analytical Solution to Linearized Flat Surface Charge Model The Poisson-Boltzmann equation (Equation 3.9) is difficult to solve because it is nonlinear. However, one can obtain an analytical solution to this problem by first taking a linear approximation as follows [21, 3]: ,24 2=2FCo sinh Ef ( FG\ RT 35 2F 2 C "I = rC K cf RT (5.1) r "D -_ --- - - -- -------- (a) - G, 0- - -- (b) Figure 5-1: (a) Two approaching planes with charged surfaces; (b) Charged hemisphere tip approaching a flat charged planar surface. where r,-' is the Debye length at the salt concentration Co. This is a good approximation if F'I is small. The solutions to this linearized equation has the form 4D = Ae"z + Be-". Since there is a surface charge density on both of plates (Figure 5-1a), the boundary conditions at z = 0 and z = D are <-2 = -12 and b =El respectively. Using these boundary conditions and the linearized equation, the solution for the potential between two infinite sheets of charge (Figure 5-1a) at a distance D apart is: <(Z) = 92(cs D)cosh(rz) - sinh(Kz) l cosh(rz) cfsinh(rD) g ) cf r, sinh(KD) (5.2) The force per unit area at any position z acting in the z-direction is equal to the sum of two terms: the osmotic pressure due to the ion concentrations that are distributed according to Boltzmann statistics and the Maxwell electric field stress (Equation 3.11). The calculation for the force becomes trivial once the potential is known and a ground (reference potential) is well defined. Here, the charged surfaces are infinite in the x and y-direction but are finite in the z-direction. At z -+ 00, the potential and the electric field are zero since this is very far from the charge (see Appendix A). Therefore, the full solution for the linearized approximation of the 36 pressure, force per area, between infinite sheets of charge is [21]: + SFfl2t PfAlat - Area 2cr1 j cosh(iD) + aO 2 (2 2cf sinh sinh2(rD) 2 (5.3) Using the above equation, one can integrate this force on small circular cross-sections and get an expression for a hemisphere tip of radius r (Figure 5-1b) [3]. If the surface charge on the tip and on the substrate are of the same order and KD is small, then some terms may be dropped and the resulting force is: 2w where j Pflatrdr = Fhemisphere 4wuou E2 r -D efK Jo (5.4) -1 is the charge on the tip and o 2 is the effective surface charge of the GAG modified substrate. The solution to the nonlinear Poisson-Boltzmann equation for this geometry can be solved numerically. 5.1.2 Numerical Method for Flat Surface Charge Model Since this geometry is simple, one can use a Newton method on finite differences to solve the nonlinear Poisson-Boltzmann equation. There are only boundary conditions at z = 0 and z = D as discussed above. Since the problem is one dimensional, the potential in space can be represented as a one-dimensional matrix or vector, <D, where each entry in the vector is the potential at evenly spaced points in the z-direction. The derivative in the z-direction can be written as differences between neighboring discretizations. a 41jontk point k -+ ([k + 1] - Oz 02 19Z2 <D[k] 6Z [k + 1] + point [k - 1] - 2D[k] 6en +2 (5.5) (5.6) where 6z is the distance between points k and k + 1. The Poisson-Boltzmann equation (Equation 3.9) for each discrete entry plus the 37 boundary conditions leads to a set of N nonlinear equations, where N is the number of discretizations, all equal to zero if the potential at each point is correct. These can be rearranged and represented as a vector-valued function A as follows: <D[1] - (1[2] + 6x L 0 2D[2] - o[1] - (D[3] + 62(2FO sinh(F"42])) A (D 2i] -Di_1]-_ '14i + 1] + -1 2(D[N - 1] - ([N - 2] - (<[N] + <D[N] 62 (2Fyo 6 2(2FCo 0 sinh(FEi])) sinh(F[N-1])) 0 - (D[N - 1] - 6Ol (5.7) If a close enough initial guess for the value of the potential at all points is given, then that guess can be refined using a Taylor series expansion. This is repeated until the change in potential at each step is small enough. This algorithm is known as a Newton method for solving multidimensional systems (here the dimension refers to the number of elements in the vector, N). Newton algorithm where (Dk 1. (Do = Initial Guess, k = 0 2. Compute A((Dk) and the Jacobian matrix of A, 3. A 4. (Dk+1 = 5. k =k + 1 6. Error 7. Loop back to 2 until Error is below some threshold. = ( OA (A(D = k + A JIA1 2 is the vector of the calculated potential values at each of the discretized points at the kth iteration. The potential is then converted to a force by taking a bounding box with one surface at point i where the derivative of the potential is zero and the other surface 38 .- . - --. - ---- -- . ------ . I E I I z = US1 S1 Z=0 z =O iz=D The control box is shown with Figure 5-2:IFPotential between two charged surfaces. F the dashed line. at infinity where the potential and electric fields are zero. The force on the enclosed plate is then the osmotic pressure at i (Figure 5-2): = 2RTCO cosh <I[i] (RT Area (5.8) The hemisphere tip geometry is approximated by using the force results between the flat surfaces and summing up the force on appropriately sized concentric cylinders. In effect, this is the numerical version of the integral done for the closed form hemisphere tip solution. This numerical method was run in Matlab. Space was discretized to 200 increments (i.e. N=200) and the maximum error threshold was 10-10. 5.2 Volume Charge Model Since GAG molecules are approximately 30nm long, the "brush"-like layer takes up some volume above the gold substrate. Without taking into account the shape of the separate molecules, this region can be modeled as a fixed uniform volume charge density [19]. This new geometry should be a significant improvement over the the 39 AD z 1 h (I) P (I) h =h brush (II) brush height 0 x (I) height 0 (a) (b) Figure 5-3: (a) Flat surface charge over a fixed charge volume; (b) Two approaching surface with fixed charge volumes. previous flat surface model because the flat surface model is only appropriate if the separation distance D is much greater than the dimension of the charged molecules grafted to the substrate (e.g., the length of the GAG chain) but the forces were measured at forces at distances on the order of the GAG length. Therefore, the experiments where the tip is coated with sulfate, the charged monolayer, can be modeled as shown in Figure 5-3a and the experiments where the substrate and tip are both coated with GAG molecules can be modeled as shown in Figure 5-3(b). The model has two different regions. In the region outside the fixed volume charge (I), the Poisson-Boltzmann equation has the same form as in the previous model. In the region inside the fixed volume charge (II), the Poisson-Boltzmann equation has an extra term added to take into account the fixed volume density. I) V24 = 2F0o sinh Ef II) V2 D = 2F0O sinh Ef (I) (RT (I) RT (5.9) (5.10) - Ef The boundary conditions are at the two surfaces and at the edge of the volume. At the surfaces, the boundary conditions are the same as before: the derivative of the 40 potential is proportional to the surface charge density. At the edge of the volume charge, there is a continuity condition so that the potential and its derivative must be continuous at that point. Of course, since the Poisson-Boltzmann equation is nonlinear, these equations are very difficult to solve analytically. However, due to symmetry, the problem is one-dimensional and thus it can be solved numerically using a Newton method. 5.2.1 Numerical Method for Volume Charge Model For the model shown in Figure 5-3a, there are three different boundary regions. At z = D, there is a surface charge density so the condition is -24 az -2-.Ef At z = 0, there 0 (see Appendix C for a more general is no surface charge so the condition is discussion of this boundary condition). Finally, at the boundary between region (I) and (II) at z = h, the potential and electric fields must be continuous. The model in Figure 5-3b has one extra condition on the edge of the second volume but overall the equations look the same. Since the problem is one dimensional because of its symmetry, the potential in space can be represented as a vector (Figure 5-4a), 4D, where each entry is the potential at evenly spaced points in the z-direction, as in the previous model. This again leads to a set of N nonlinear equations, where N is the number of discretizations. These can be rearranged to define a vector valued function A as follows (where the kth point 41 D "z D'<h (I) y NIX = (Pflx) (h/D) D>h h= brush height NPix (a) (b) Figure 5-4: (a) A flat surface charge over a fixed charge volume a distance D > brush height apart; (b) A flat surface charge over a fixed charge volume a distance D' < brush height. corresponds to the point on the edge of the volume charge): P[1] - D[2] + 2D[2] - 1[1] - 1[3] + 21[i] - 4[i - 1] - A(D) = - Ef '1[i + 1] + 6 6 0 Pf-) 6 (2"O sinh( F"d2]) - - 2] - 1[k] + 2-[k -1]- sinh(F",l]) 5 (2"O RT Pfi2) Ef ) (2FCo sinh(ETE) - 2(2FCO sinh(F11[k-1]) _ ___ 0 24)[k] - 4)[k - 1] - D[k + 1] 21[k + 1] - 4[k] - 4[k + 2] + 62(2FCO sinh(F 21[i] - _ 1] -[i - 1[i + 1] + 21[N - 1] - 1[N - 2] - [N] + 62(2F00 6 2(2FCo 4[N] - P[N - 1] - 6x [k+l])) sinh(ETE)) sinh(FlN-1])) 0 If the plates are close enough together and the volume charge takes up the whole space, then there are only two boundary conditions and the system simplifies to the 42 following (Figure 5-4b): 1[1] - (D[2] + 21[2] A(f) = 2[i] - - [i- 62( D[1] - D[3] + 1] - X 6 sinh(F("] RT Ef 2 (2Fo XEf 1 [i + 1] + 24[N - 1] - @[N - 2] - P[N] + sinh( F"2 T 62(2FCo 6 -] - PfE sinh(!f E) - 2 ( 2FCo X 0 P-ix Ef ) sinh( F4D[N-1) Ef 4[N] - 4[N - 1] - 6x(Ewhere p' = pfix- 0 0TE 0 is the adjusted volume charge density now that the plates are a distance D' < h apart. These two systems of equations can both be solved by a Newton method (as in previous section). The solution to the linearized equation is used to obtain a good initial guess at the first distance. The solution for the potential at each distance is used as the initial guess for the potential at the next distance. The potential is used to compute a force by taking a Gaussian bounding box with one surface in region (I) where the derivative of the potential is zero and the other surface at infinity where the potential and electric fields are zero. Then, the force on the enclosed surface is then due only to the osmotic pressure term at the first surface in region (I). Finally, the force due to the hemisphere tip geometry is approximated by summing the forces due to appropriately sized concentric cylinders (Figure 5-5). This numerical method was implemented in Matlab. Space was discretized into 200 increments (N=200) and the maximum error threshold was 10-10 5.3 Charged Rod Model Since the chondroitin sulfate chains on our surface are approximately 30nm long but only about Inm wide [2], the long chain polymeric shape might have an effect on the force profile that is not taken into account by the smooth volume charge model above. To account for certain aspects of molecular shape, the GAG coated surfaces are 43 (I) D h (II) Pfix Figure 5-5: Approximated hemisphere of charge over a fixed volume of charge. Figure 5-6: Two parallel plates with charged rods. modeled as two parallel plates with charged rods (Figure 5-6). Once more, because the Poisson-Boltzmann equation is nonlinear, this can only be solved numerically. However, due to of its lack of symmetry, the model of two approaching brush surfaces (Figure 5-6) is now a three-dimensional problem. This must be solved by a slightly modified method, as the matrices that would result from a normal Newton method would be too large (trillions or more entries) to be manipulated and stored efficiently on normal modern computers. 5.3.1 Numerical Method for Solving Charged Rod Model The surfaces can be broken down into small repeating units (Figure 5-7). The space potential of one unit is solved numerically and the force corresponding to that potential is calculated. The rods are modeled as volumes with a fixed charge density 44 charged rod ~3 nm Figure 5-7: Repeating unit in the parallel charged rod plate model. D is the distance between the square base plates. This unit is discretized into cubic elements of size 6o, by 6o, by 62. and the substrates onto which they are attached are modeled as flat surfaces with no surface charge. The dimensions of the unit are determined by the coating density of the GAG on the surface and the distance between the tip and surface. The total force for the two surfaces at the current spacing, D, will be the force for one unit times the number of units needed to represent the surface. The hemisphere geometry of the tip can be approximated by numerically integrating in the same manner as with the previous -models. The units are discretized into small elements with dimensions 6x, 6g, and 6 2. The fixed volume charge and potential for each cubic element are stored in a threedimensional matrix. As before, the continuous derivatives in Cartesian coordinates 45 are discretized to finite differences: a2 02 92 OZ24 ([i + 1, J, k] + 4)[1 - 1, j, k] - 24)[i, j, k] VD point(ij,k) 62 P[i, j + 1, k] + + [ij - 1, k] - 2 [i, j, k] 62 y [i, j, k + 1] + 1[i, j, k - 1] - 2)[i, j, k] 62 z Then, the continuous Poisson-Boltzmann equation can be written in discrete form (where pfix[x, y, z] is the volume charge for the discrete volume at (x, y, Z)): f (4[i, j, k]) f (a) P[i-7j,k] = P[i, j, k], where = 1 2a( x + 1 ) + A sinh(Ba) z =(1~i±1,i~kl±1[i-1,jik])(- k] + +([i,j+±1, 1 [i, j - 1, k]) y +(1[i,j,k + 1] + 1[i,j, k - 1])(62) z Pf ix [i7j, k] For each element that is not a boundary, this equation is used to update the potential. Boundary conditions determine which values of 4[x, y, z] and which differences, P[x, y, z], are fixed: a ai -,a- where i is the direction of the boundary. (5.11) For example, if there is a charged surface between the element at (i, j, k) and the ones at (i + 1,] , k) and (i, j + 1, k), Equation 5.11 leads to the following two equations for 46 the potential at point (i, j, k): TD[i + 1, j, k] - (1[i,j, k] _- 1[i, j + 1, k] - 1[i, j, k] o- If the equation is not satisfied, then (D[i, j, k] must be changed by some amount, A, approximated using the Taylor series approximation as follows: f (4b +,A) ~f (,D) + A f '(<D) f(14i,j,k]+A) P[i, j, k] = ~ P[i, j, k] f((D[i, j, k]) + Af'((b[i, j, k]) P[i, j, k] - f (4b[i, , k]) f'(1b[i, j, k]) For the boundary conditioned elements, A is taken to be the average of the A calculated for each of the equations. Therefore, for the example boundary above, A can be defined by these equations: A, = A2 A D[i + 1,jk] - <[ij, <b[Zjj+1,k]- <[ijk]+ Al + A2 2 The full algorithm is as follows: 47 k]+ Ef Ef 6x 6y 1. Initialize D (set everything to 0 for example) 2. For all (i, j, k), calculate A[i, j, k] based on current 1[i,j, k]; check if it is a boundary and calculate A[i, j, k] appropriately if it is. 3. D <- 4+ A 4. Error = 5. Loop back to 2 until il 2 a. a certain fixed number of iterations is reached b. until Error < MAXERROR 6. Calculate force from 1 at current distance D 7. D = D - d (decrease the distance between the two plates by d) 8. Loop back to 2 until D = 0 For these experiments, the maximum number of iterations was 10000 and the MAXERROR was 10'0. This is a Newton method like the one used in the volume charge model above but in this case the Jacobian matrix is too large to store in memory. Therefore, this method employs a "matrix-free" Newton method approach. Because the potential needs to be solved for several values of D in order to get a full force curve, the solution of the potential at D is used as the initial condition for the potential D - d, the next distance. If d is made small enough, the potential at D will be very close to the potential at D - d and therefore the algorithm should converge fairly rapidly at each iteration. The energy method is used to calculate the force in this model. The total energy at distance D is calculated by numerical integration. The energy contribution of one 48 element that is not a boundary is: j, k] = (pfix[ii , k]f[i, j, k] )6x6y64 w[ij,k] = -(2RTCo(cosh( Wf [i, M . e wmz,,k k] [i) = TI[i,, k]) - 1) 6X6 6z [i -+ 1') k, 2 k] 6X + 4D[i, j + 1, k] - D[i, j, k] 6Y + w[ij,k] [i,j,k + 1] - [, j, k]) 6 6 wf +wo+m where Wf is the energy contribution due to the fixed charge density, wo is the energy contribution due to the osmotic pressure, and wm is the energy contribution due to the electric field stress. The energy contribution of an element with boundary conditions is the average of the potential at (i, j, k) and the potential on the other side of the boundary multiplied by the surface charge density and by the surface area. Hence, for the example boundary element described above, the contribution to the energy will be: . . w[i, j, k] =- ((D[iI j, k] _+(D[z + 1, j, k]) 6j +or(4)[Z, j, k] + (1[t., j + 1, k]) 2 + 66 The total energy at distance D will be the sum of all these contributions. Wtotal = w[i, j, k] Z: Z x y z The force is the derivative of the energy in the z-direction, which is approximated by finding a second order polynomial of best fit for the local energy. The force at point z = D is approximated as the derivative of the quadratic fit of the energies calculated at distances D-( < z < D+( evaluated at z = D, where ( is 1.5nm unless otherwise 49 stated. The numerical method to solve for the potential and the energy calculation were implemented in C and run in parallel on several machines so as to calculate the energy at every increment of one nanometer. There were two machines allocated per ionic strength run. For example, there were two machines calculating the energies at 0.01M NaCl starting at 100nm down to 2nm where the first machine was calculating at even number distances and the second one at the odd distances. It was determined that discretizing by 50 increments in the x- and y-directions and 100 in the z-direction was enough to ensure convergence since more discretizations produced the same results. At this discretization, it takes about 24 - 48 hours to run the program at a single concentration for all the distances on one 1GHz PentiumIII. Therefore, doing the calculations in parallel, getting the results for all the concentrations at all the distances takes about 24 hours. The conversion of the energy to force was done in Matlab. This model assumes the rods are stiff so that they do not move until there is contact. After contact, the rods maintain the same volume by getting shorter and wider and maintaining their total charge. If the GAG molecules are not really completely rods and move as the surfaces approach one another, the rod behavior as the surfaces approach could be altered. Of course, the description of this simulation by no means models all that is happening at the surfaces. There are other interaction forces besides the electrostatics. However, because GAG molecules are very highly charged, the contribution of these forces to the total force should be small, especially at distances greater than the length of the GAG chain. 50 Chapter 6 Results and Discussions Measurements were taken using the charged and neutral monolayer-coated tips. The tip radii were measured to be approximately 25nm. In this chapter, the different models are compared to the experimental data (see Chapter 4). Because of difficulties in coating the tip with GAG chains, the experiments of GAG versus GAG are not yet completed. However, it is interesting to see what the different models predict for these experiments since modeling GAG-GAG interactions is the ultimate goal of the project. 6.1 Smooth Surface Charge Model The smooth surface charge model is a useful tool as it is relatively easy to solve and there is a closed form analytical solution to it in the linearized approximation limit (see Chapter 5). However, although the model predicts the correct trend, it does not fit the data well. The Debye length at various salt (NaCl) concentrations is summarized in Table 6.1. The values calculated using the model are compared to the experimental results when all model parameters are fixed at the values determined experimentally or from theory (see Table 6.1). The Debye length is the length scale over which the force should decay away when two charged surfaces approach each other. However, since the GAG "brush"-like layer takes up a volume, is not flat, and the forces are acting 51 NaCl concentration (M) Ir (nm) 3M 0.18 iM 0.30 0.1M 0.96 0.01M 3.04 0.001M 9.61 Table 6.1: The Debye length at various NaCl concentrations. Parameter NaCI Concentration Debye Length Value 1-0.001M See Table 6.1 Tip Radius Tip Surface Charge Density Effective GAG Surface Charge Density 25nm -8.16 x 10 4 6 Method for determining Known quantity of salt added Theoretical value from NaCl concentration (see Chapter 3) Measured from SEM picture From previous experiments (see Chapter 4) -0.18964 Calculated from experimental values of chain length, number of repeating charge units and the GAG coating density on the surface Table 6.2: The flat surface charge model parameters. 52 x 10-4 Comparison of nonlinear surface charge model to its linearized approximation 1 - 0.9 - I - 0.01 nonlinear - M linear 0.01M linear --- 0) C.) M noniinear 0.M nonlinear 0.8 0. 1M linear 0.7 -M linear 0.6 0.01M nonlinear 0.5 0.1M nonlinear 0.1M linear 0.01M linear 0 1M nonlinear 0.4 0.3 0.2 - . 0.1 ......... .. 0 0 10 20 30 40 50 60 70 distance (nm) Figure 6-1: A comparison of the force predicted by the surface charge model using the full solution and the linearized approximation. 53 Comparison of nonlinear and linearized surface charge models to experimental data I F 1M I UMnled[ - - a.M nonlinear 0.01 M nonlineai_ 1 M data 0.1 0.09- - M linear 0.1M data 0.08 0.1M linear - - 0.01M - data 0.07- 0.01M linear 1 M data 0.1M data Q0.1M data 0.01M linear 0.06 k 0' 0.05 0.1M nonline ar 0 0.04 0.1 M nonlinea r 0.03 - M linear and 1 M nonlinear 0.02 0.01 0 0 0 10 20 30. 40 10 20 30 40 50 60 50 60 distance (nm) 80.... 90.100 70...... 70 80 90 100 Figure 6-2: Surface charge model (nonlinear and linearized) versus experimental data. 54 on the length scale of the GAG molecules, the force measured does not decay in the same way as it would for flat charged surfaces (Figure 6-2). The nonlinear model does not fit as well as the linearized approximation model because the linearized approximation overpredicts what the force should be between two charged flat sheets (Figure 6-1). However, since in this case the GAG molecules take up a volume and the forces observed are larger than those for flat surface charges, the error in the linearized approximation is in the right direction and therefore the linearized model is closer to the data then the nonlinear model (Figure 6-2). Of course, some of the model parameters can be changed so that the data and the surface charge model results are closer. For example, one can change the model Debye length in order for it to fit the data more closely. However, the discrepencies between this model and the experimental data are largely due to the geometry of the model not reflecting the geometry of the experimental setup. Artificially changing the Debye length to fit the data reveals no new information and can be misleading. Although the flat surface charge model predicts the right trends, the GAG surface is not well modeled by a flat surface charge density and so the model results are not expected to be close to the experimental data. 6.2 Volume Charge Model The model parameters for the volume charge model were fixed to the previously known values (see Table 6.2). The volume charge model fits the data reasonably well. The model results for the volume of charge approaching a flat charged surface are shown in Figure 6-3. There is a cusp at the point where the flat surface touches the top of the volume (32nm). However, when the tip is made to be a hemisphere, the cusp is not as conspicuous as it is smoothed out in the numerical integration (Figure 6-4). The model tends to underpredict the experimental results when compared to the GAG versus charged monolayer data (Figure 6-5). This is probably because modeling the GAG surface as a volume charge does not take into account the discrete 55 Method for determining Value Parameter Known quantity of salt added Theoretical value from NaCl concentration 1-0.001M See Table 6.1 NaCl Concentration Debye Length Tip Radius Tip Surface Charge Density Monolayer Neutral (see Chapter 3) Measured from SEM picture From previous experiments (see Chapter 4) 25nm -8.16 x 10-4 From literature [13, 18] 1-2 times Eo Permittivity GAG Volume Charge Density -5.925 * 106 Calculated from experimental values of chain length, number of repeating charge units and the GAG coating density on the C surface Charge Height Determined from GAG length 32nm Volume Table 6.3: The smooth volume charge model parameters. 1. X 10' Volume charge model of flat surface approaching volume charge 5 Volume height = 32nm Force at 0.01 M 0.5 0 10 20 30 40 distance (em) 50 60 70 Figure 6-3: Force between a smooth volume of fixed charge and a flat charged surface (1nm x 1nm) at 0.01M. 56 Volume charge model of charged hemisphure approaching volume charge 0.5 0.450.4- Force at 0.01 M 0.35- 0.34 - Volume height 32nm a 0.25 0.2- 0.15- 0.1 0.05- 0 10 20 30 40 distance (cm) 50 60 70 Figure 6-4: Force between a smooth volume of fixed charge and a flat charged hemisphere (radius=25nm) at 0.01M. nature of the GAG chains. The discrepancies are most apparent at distances less than 10nm. This is because at distances smaller than 10nm, the steric and nonelectrostatic components of the force also tend to be more significant. Therefore, any model which only considers electrostatic forces should deviate from experimental results at small distances. The smooth volume charge model can be made to fit the data more closely by increasing the surface charge density parameter from the value previously calculated. Although this will make the lower concentrations fit better, it will not change the fit at the higher concentrations notably since the distance on which the surface charge makes significant differences to the force is very small (see Table 6.1). The model tends to overpredict the force for the neutral monolayer-coated tip (Figure 6-6) at distances greater than 10nm. This discrepancy in the force is most likely due to the modeling of the tip monolayer as a perfect insulator; in reality, the layer is not a perfect insulator and has a finite non-zero conductance. Therefore, the actual induced surface charge on the surface will be smaller than that predicted by the model. This in turn leads to a smaller force. For distances less than 10nm, the model predicts forces which are smaller than the experimental value. This is because 57 Comparison of volume charge model to charged monolayer vs GAG experimental data 0.5 - M model 0.1M model 0.01M model 0.45 S1M data - 0.1M data - 0.01M data 1 M data 0.4 0.1M data 0.35 0.01 M data 0.3 -M model Volume height =32nm C 0.25 1M model 0 0.2 0.01M model 0.15 0.1 0.05 C 0 5 10 15 25 20 30 35 40 45 50 distance (nm) Figure 6-5: The volume charge model compared to experimental data for GAG surface versus charged monolayer-coated tip. 58 Comparison of volume charge model to neutral monolayer versus GAG experimental data 0.5 omMcidel 1 M model 0.45 0.1M model 3M data 0.01M model - 3M data S1M data - 0.1M data 1 M data 0.4 0.1M data 0.351 0.01 M data ''- 0.3 Volume height 3M model = 32nm (D C.) 0.25 1 M model 0 0.2 0.1 M model 0.15 0.01M model 0.1 0.05 0 0 ' ''' ' 5 10 15 20 25 30 35 40 45 50 distance (nm) Figure 6-6: The volume charge model compared to experimental data for GAG surface versus neutral monolayer-coated tip. 59 at small distances nonelectrostatic effects become very significant. This is especially true at the higher ionic strengths where the electrostatic component of the force will be relatively small. There appears to be a certain nonelectrostatic component of the force which is independent of ionic strengths and begins at about 10nm. This is evident since the experimental data at 1-3M for distances less than 10nm lie rather close. The electrostatic component of the force for these ionic strength should be small. The model, which overpredicts the actual electrostatic component, predicts forces which are about 5-10 times smaller than the experimental results at IM and 3M NaCl. Therefore, most of the force measured is due to steric and nonelectrostatic effects. For the volume charge versus volume charge (which models the GAG versus GAG surface), the cusp is at 64nm when the two parallel volumes (i.e. flat tip versus a parallel flat substrate) approach each other since that is when the two volumes touch. Of course, the cusp is smoothed out when the model assumes a hemisphere tip. This model assumes that the GAG brushes do not interpenetrate at all. Once the fixed charge volumes touch, the fixed charge density is constant everywhere between the tip and the bottom of the substrate. The results are illustrated in Figure 6-7. Because of difficulties with the GAG tip modification chemistry, there are not yet any experimental data to which these model results can be compared. Nonetheless, a few interesting remarks can be made about these results. The force at 32nm and 0.1M NaCl is about 0.03nN. This translates roughly to a pressure of 0.1MPa. This predicted force can be compared to the macroscopic measurements of cartilage as the volume charge density at that distance is equivalent to a GAG-GAG spacing of 3nm, which is approximately physiological conditions. The value reported in literature for the measured modulus of cartilage tissue is about 0.7MPa and it is thought that electrostatic interactions of GAG account for about 0.29MPa [8, 15, 2]. 60 Volume charge model of GAG-tip approaching GAG surface 0I 0. I 3M iM 0.1M - 0.01 M 0.5 32nm 64nm 0.01M 0.4 U.IM 0 I vi 0.3 - 3M 0.2 - 0.1 0 0 - 10 I - 20 30 40 50 60 70 80 90 100 distance (nm) Figure 6-7: The volume charge model of a GAG-coated tip approaching a GAG-coated surface. 61 Parameter NaCl Concentration Debye Length Value 1-0.001M See Table 6.1 Tip Radius Tip Surface Charge 25nm -8.16 x 10-4c Method for determining Known quantity of salt added Theoretical value from NaCl concentration (see Chapter 3) Measured from SEM picture From previous experiments (see Chapter 4) Density Monolayer Neutral Permittivity GAG Volume Charge Density Charge Rod Height Charge Rod Spacing Charge Rod Width From literature [13, 18] 1-2 times co 106__ Calculated from experimental values of chain length and number of repeating charge units From GAG length From GAG coating density Approximated from GAG chain diameter in literature -5.925 * 106 32nm 6nm 1nm Table 6.4: The rod model parameters. 6.3 Charged Rod Model The model parameters for the volume charge model were fixed to the previously known values (see Table 6.3). The model results for charged rods approaching a flat charged surface are shown in Figure 6-8. Similar to the charged volume model, there is a cusp where the rods first touch the flat surface (32nm) but when the tip is made to be a hemisphere this cusp is smoothed out in the numerical integration. The charged rod model predicts higher forces than the smooth volume model because the same amount of charge is in a much smaller volume. It overpredicts the force at 0.1M but fits the data at the other concentrations better than the volume charge model (see Figure 6-9). The model results predict smaller forces than the data at distances less than 10nm as that seems to be the range where steric and nonelectrostatic forces are more significant. Also, the model tends to slightly underpredict the force at - 32nm. This is probably due to polydispersity, which means that not all the GAG chains are 32nm long but have an average length of 32nm with some standard deviation. Therefore, the longer chains will cause the force to increase sooner than predicted. 62 I- Charge rod model of flat charged surface approaching a surface with charged rods 0.9 - 0.8 - 0.7 Rod height = 32nm 0.6 Force at 0.01 M 0.50.4 - 0.3- 0.2- 0.1 - 0 10 20 30 40 distance (nm) 50 60 70 Figure 6-8: Force between rods of fixed charge and a flat charged surface (Inm x 1nm) at 0.01M. There might also be some initial resistance that is not electrostatic in nature when the tip hits the top of the GAG chains that cannot be taken into account by the charged rod model. This model also tends to overpredict the force for the neutral monolayer-coated tip (Figure 6-10). This is expected, as explained in the previous section, since the monolayer is not a perfect insulator. For distances less than 10nm, the model predicts forces which are less then the experimental value. Again, this is because, at those short distances, non-electrostatic components of the force become much more significant. Because of difficulties with the GAG tip modification chemistry, there are not yet any experimental data to which these model results can be compared. However, these data can be compared to macroscopic modulus measurements [8, 15, 2]. The force at 32nm and 0.1M NaCl is about 0.03nN or about 0.1MPa (Figures 6-11 and 6-12). The value reported in the literature for the measured modulus of cartilage tissue is about 0.7MPa and the electrostatic interactions of GAG is about 0.29MPa of the total modulus. This is essentially the same as the result for the volume charge model since the two models' result curves cross at approximately 32nm. 63 Comparison of rod model to charged monolayer vs GAG experimental data 1 model 0.1 M model -- lM data 0.45 S 0.1M data . 0.01M model .Mdata 0.1 M data 0.4 - 0.01M data 0.35 - M model 0.3 - >0.25 rod height = 32nm - model -0.1M 0 0.2 0.01M model -. -. 0. 15 0.1 0.051 - - - -. 0 0 5 10 - 15 20 30 25 distance (nm) 35 40 45 50 Figure 6-9: The charged rod model compared to the experimental results of a charged monolayer-coated tip approaching GAG surface. 64 Comparison of charged rod model to neutral monolayer versus GAG experimental data 0.5 - M model 0.1M model 0.01M model S1M data 0.1 M data 0.01 M data 0.45| 1 M data 0.4F 0.1M 0.35F data 0.01 M data 0.31 M model Rod height= 32nm 0.25 0. 1M 0.2 model 0.01 M model 0.150.1 - * '. 0.05 0 4 8 12 16 20 24 28 distance (nm) 32 36 40 44 48 Figure 6-10: The charged rod model compared to experimental data for a neutral monolayer-coated tip approaching a GAG surface. 65 1 Charge rod model of two approaching surfaces with charge rods x 10' 0.9 0.8 0.7 Rod height = 32nm 2x Rod height o- 64nm 0.6 0.5 Force at 0.01 M - 0.4 0.3 0.2 0.1 0 10 20 40 30 50 distance (nm) 70 60 80 100 90 Figure 6-11: Force between two approaching 1nm x 1nm surfaces with charged rods at 0.01M. 0. ID I I" I Rod charge model of GAG-tip approaching GAG surface I : iM -- 0.5 411 32nm 0.4 0.1MA 0.01 M 64nm 001M i.M 0.3 U 1M 0.2 0.1 01 0 10 20 30 40 50 distance (nm) 60 70 80 90 100 Figure 6-12: The charged rod model of a GAG-coated tip approaching GAG surface. 66 Force between charged monolayer and GAG compared to models with added steric component 1 - __ 0.9 0.01M data Steric model (de Gennes) Volume charge model + StE c Charged rod model + steric 0.8 0.01M data 0.70.6 0 0.5 charge rod model + steric mod el L volume charge model + steric nodel 0.4 steric model 0.30.2 0.1 U. I 0 10 20 30 distance (nm) 40 50 60 Figure 6-13: Charged monolayer versus GAG force at 0.01M NaCl compared to 2 models with added approximated steric effect. 6.4 Comparison of Results All the models predict forces smaller than the experimental results at short distances (less than 10nm) because no steric or other non-electrostatic effects were taken into account. Using the equation in Chapter 3 for the force between uncharged brush layers, the non-electrostatic component of the force can be approximated. As previously mentioned in Chapter 3, the steric component is hard to isolate from the electrostatic effects in theory and therefore it is not surprising that although the new models with added steric effects fit the data better they still show large discrepancies at the smaller distances. The de Gennes steric model [6] does not take into account intra-GAG electrostatic repulsion and therefore is not really appropriate to use for these experiments for anything more than rough approximations. This is also noticeable if the predicted 67 0.3 Neutral monolayer tip versus GAG at 3M NaCl experimental data compared to de Gennes steric theory . - 3M experimental dat6 -de Gennes theory 0.25 0.2experimental data z )0. 15 - P 0 0.1 steric theory 0.05- 0 0 10 30 20 40 50 60 distance (nm) Figure 6-14: Neutral monolayer versus GAG experimental results at 3M NaCl compared to the de Gennes steric model. steric force is compared to the experimental results of the neutral monolayer coated tip and GAG surface at 3M NaCl. At this ionic strength the inter-GAG electrostatic forces should be very small since the Debye length is only about 0.2nm and therefore the steric and nonelectrostatic effects should dominate. However, as shown in Figure 6-14, the de Gennes steric model does not exhibit the behavior that would match the data. From the very poor fit of the flat surface charge model to the experimental data for the charged monolayer-coated tip versus GAG surface in comparison to the other models, it can be concluded that the space that the GAG molecules occupy is very important in determining interaction forces. This is reasonable considering that the length of the GAG chain is of the same order as the distance between the tip and the surface. It is not yet clear which of the two other models, the volume charge or the 68 charged rod, is more appropriate for the GAG surface. It seems that, in the range where electrostatic forces are dominant, the charge volume model provides a lower bound while the charged rod model provideds an upper bound on the experimental data. Since the way the charged rods move as they are pressed on by the tip can make significant changes in the resulting force, it might be that the GAG chains are best modeled by rods which lie down instead of compress as the two surfaces approach. Experimental data for the GAG versus GAG has not yet been obtained as there were difficulties in coating the GAG onto the tip surface. However, observations can be made about each of the different models' behavior. As can be seen in Figure 6-15, there are significant differences in the behavior of the predicted curves for the volume charge and the charged rod models. The main reason is that the volume charge model assumes that there is no interdigitation of GAG molecules whereas the charge rod model assumes that there is. The predicted force at 32nm in O.1M NaCl is about the same in both models which means they compare about the same way to the macroscopic modulus measurements. Until some measurements are done on the molecular level, there is no clear indication of precisely how the GAG molecules behave at physiological ionic strength. 69 Comparison of 2 models of a GAG-coated tip approaching a GAG-coated surface 0.7 - 0.6- 0.5 m 1M volume model 0.1M volume model 0.01 M volume model - rod model 1M M rod model 0.o1e 0.01M rod model 0.01M rod model - 0.01M volume model 0.4 - 0. 1M rod model C 0 0. 1M volume model 0.3 r 1 M rod model 0.2- 0.1 1 M volume model - .... ... . 0 10 20 40 30 50 60 70 distance (nm) Figure 6-15: Comparison of the volume charge and charged rod models of GAG- boated tip approaching a GAG-coated substrate. 70 Chapter 7 Conclusions 7.1 Summary Articular cartilage is the load bearing tissue found on joint surfaces. Macroscopic tissue measurements have shown that its ability to withstand large compressive forces is primarily due to the highly charged glycosaminoglycan (GAG) molecules found in its extracellular matrix. With tools such as the molecular force probe (MFP), the molecular forces between thesis GAG molecules can be measured [24, 23, 201. In this thesis, different models for the electrostatic repulsion between GAG "brush"like layers were developed and compared to experimental MFP force data for a charged monolayer coated tip and a GAG coated surface. It was shown that modeling the surface coated with a "brush" -like layer of GAG as a flat surface charge predicted the trends in the data but greatly underestimates the actual force measured. Modeling the "brush" as a smoothed volume of fixed charge only slightly underestimated the actual experimental results. This second model is a useful tool in that it fits the data relatively well, since the resultant force from the volume model has a shape much closer to the experimental data than the flat surface charge model, and it is an easy model to implement. However, modeling the "brush" as a surface covered with rods of fixed charge predicts a force which was even closer to the experimental results. This model, however, tends to slightly overestimate the force in the region where electrostatic forces are dominant. This would seem to indicate that the "brush"-like 71 layer behaves like a surface covered with rods that are not completely rigid. It was also shown that there can be an electrostatic component of the force when the MFP tip is coated with a neutral monolayer. The monolayer can be modeled as a perfect insulator to predict the forces. However, it appears that this is too strong a condition and that incoporating a conductance into the model would be more appropriate. It was noted, however, that the electrostatic component of the force only dominates at distances greater than 10nm. At smaller distances, there appears to be a large steric and non-electrostatic component of the force. Since there are not yet any MFP results for a GAG-coated tip interacting with a GAG coated surface, both the volume and the rod charge models were compared to macroscopic results. They both found forces which were of the same order as these experimental results. 7.2 Future Work Because the rod model has more parameters and is more general than the smooth volume model (a smooth volume is just one big rod), it can be better modified to fit experimental results and give more insight into the behavior of the GAG chains. Since the behavior of the rods as the distance gets smaller than their height greatly determines the shape of the force curve, more studies need to be carried out to fully understand how these GAG molecules behave in tissue. Experiments need to be carried out to measure the forces between GAG "brush"-like layers in order to relate these to previous macroscopic measurements. In this thesis, the rod model has been shown to be a feasible alternative to the simpler models previously discussed in the literature [19, 3]. Accurate models give good insight into the behavior of GAG chains in physiological conditions. Therefore, the continued development of more realistic models will ultimately provide much needed insight into the microscopic and macroscopic properties of articular cartilage. 72 Appendix A Electrostatic Interactions Between MFP Tip and Substrate: Generalized 2-D Models for "Neutral" and Charged Surfaces In this appendix, a model is developed to describe the experiments where the tip was coated with a neutral polymer. Even though the polymer may be neutral, the charge from the GAG on the substrate may induce a surface charge on the tip. Thus, electrostatic forces may still be induced and measured. The tip is modeled as a grounded metal surface coated with an insulator that may also contain surface fixed charge groups (Figure A-1). The substrate surface charge is given as o-s; the charge at the insulator electrolyte interface, U-2, and the charge on the insulator metal interface, 9s3, are both unknown. There are three regions in this model: region 1 between the charged surface and the tip, which is filled with salt solution, region 2 inside the insulator, which has no volumetric charge density, and the region 3 inside the grounded metal, in which the electric fields must be zero. The electric field and potential in each of the three regions are described by the laws of electroquasistatics (Gauss, Faraday, and conservation of charge, respectively) 73 1 P .7(D= + 3 2 ' E=0 + -, f 3 I _- + + 0 z z D zD+d Figure A-1: Model of tip approaching a charged surface. (a) Ea (b) nfb Fb a E Ea interface,S Figure A-2: General interface. and the interface boundary conditions associated with each respective law (Figure V - cE =p n - (,EaE V x E =0 n x (Ea - Eb) V -J = atp n - (JaE, - CbEb) = Us - = 0 (or <1a = <b*) UbEb) = at Os * in the absence of an infinitesimal thin double layer. These equations lead to three boundary conditions for the model, one at each of the 74 interfaces: n o1s1 'a D = -VD - n - (E2E2 - cf E1) at z = 0 at z O-s2, = , - D 0, at z = D + d (A.1) (A.2) (A.3) The electric field inside the insulator (region 2) will be constant since there is no volume charge. From this, an expression for the induced surface charge density at z = D can be derived. A.1 Induced Surface Charge The problem is in steady state so 2 = at 0. Therefore, due to charge conservation, - n - (9 2 E 2 - o-E yUs2 =0 1) where -1 and or2 are the conductivities of region 1 and 2 respectively. But since region 2 is an insulator, -2 ~ 0, and therefore, E12 = 0 (A.4) where E1 2 is the magnitude of the electric field in the z-direction inside region 1. Gauss's law states that discontinuities in the electric fields at the boundary are due to a surface charge: 0 2= n - (f 2 E 2 - f E1) The electric field in region 1 at z = D- is zero as shown above. So this simplifies to: Os2 - -E2E22 75 where E 2, is the magnitude of the electric field in the z-direction inside region 2. The electric field inside region 2 can be obtained by finding the slope of the potential: (z = D + d) - 4(z = D) D+d-D 2- on the insulator will be: Therefore, the surface chargeE~z62 (A.5) f24)2 -' d and similarly, the surface charge on the insulator at z =s3 - 2E22 =- = D + d will be: f2 dd E2E22= ' which means that: O-s3 (A-6) = -Us2. Therefore, the net charge on the insulator is 0, as expected. However, there is a non-zero surface charge density on the insulator at z = D. Notice that the surface charge induced on the insulator at z = D is of the same sign as the potential on its surface. The potential will be positive when a., > 0 and negative when o-i < 0. Therefore, the induced charge on the tip will have the same sign as that on the charged substrate. This will lead to electrostatic repulsive forces even though the tip may not have had charge to begin with. A.2 Electrostatic Force from Induced Charge The potential in region 1 can be calculated numerically in the same manner as for the flat surface charge model described in 5.1. The only difference is that the boundary 76 condition at z = D is (<D = 0 as show above. Once the potential is obtained inside region 1, the force can be calculated by using the same balance of force method used in the calculation of force for the charged tip versus charged substrate models. A Gaussian bounding box is defined where one surface is at z = D_, where -Philz=D_ = 0, and the other surface is at infinity, where the potential and the electric field are 0. Then the force on the tip enclosed inside the box will be due only to the osmotic pressure at z = D. This logic can be used if the surface substrate model is changed from a flat surface charge to a volume charge, or even to charged rods. 77 78 Appendix B Detailed Experimental Methods These experiments were carried out by Joonil Seog at MIT [24, 23, 20]. Silicon wafers (Recticon Enterprises Inc., Pottstown, PA; test grade) and Si 3 N4 cantilevers (Thermomicroscope, Sunnyvale, CA) were coated with 2nm of chromium to promote the adhesion of gold, followed by a 100nm of gold deposited using thermal evaporator at 1.5). Gold-coated silicon wafer was cleaned using piranha solution (1:3 concentrated H2 SO 4 and H2 0 2 (30%)) for five minutes just before further modification with CS-GAG. Metabolically radiolabeled 3 5S-aggrecan was obtained from rat chondrosarcoma cell cultures and digested with proteinase K. The resulting amino-acid-terminated 35 S-CS-GAG chains were precipitated with ethanol, purified on superose-6, and re- suspended in 0.01M phosphate buffer. The 35 S-CS-GAG chains with their terminal reactive amine groups were treated with an excess of dithiobis[sulfosuccinimidyl propionate] (DTSSP, Pierce), and the terminal disulfide bond was reduced to a thiol group using an excess of dithiothreitol (DTT, Pierce). After removal of excess reactants, 5pl aliquots of the thiol-terminated 3 chains were placed on cleaned 1S-CS-GAG 1cm x 1cm gold-coated silicon wafers. The slides were left to react at room temperature for 72 hours. After rinsing, the wafers were placed in 1 mM solution of C1 2H 25-SH for 30 min. This process known as backfilling forms a neutral SAM (self-assembly monolayer) on that part of the gold which was not modified with GAG. GAG surface density on the wafer was assessed by removal of the 79 3 1S-CS-GAG after AFM tests HH -NH COOH H OH H H OH 2 + so; o H CH H SOjNa4 H 0DTSSP0 H NHCOCH, &--10-50 0 0H 0 DTT, 1 h 0 -P PBS buffer, 1 hr Cs H0 backfill with methyl-terminated $AM : C1 "1-5S" ''u" s HN H 0 2S H 0 Cs 24 hrs ~~ SO,-Na HN 0 s§ s Cs end-gmfted CS-G AG H 0 =0 S =100 nm on silicon chip backfill with methyl-terminated SAM: Figure B-1: Schematic of the chemistry to graft GAG onto substrate surface. by sonication, and measuments by scintillation counting. The parking density was about 1 GAG chain per 6.5nm x 6.5nm (Figure B-1). The gold coated tip was chemically modified by forming monolayers in one of two ways. It was immersed for 24 hours in either a 5mM ethanol solution of 2- mercaptoethanesulfonic acid (Aldrich) to create a charged tip or a 5mM ethanol solution of 11-mercaptoundecanol (from Prof. Laibinis lab at MIT) to create a neutral tip. The tip modified with 2-mercaptoethansulfonic acid was backfilled with ethanethiol (Aldrich). Modified tips were rinsed in ethanol and dried under N2 just prior to mounting them. Various ionic strength solutions were prepared by mixing NaCl and dionized H2 0. The pH of the solutions was about 5.6. Repulsive forces between the CS-GAG chains and a chemically modified tip (Figure 4-3a) were then measured in the NaCl solutions ranging in concentration from 0.0001M to 3.0M, using the Molecular Force Probe (MFP; Asylum Res., Santa Barbara, CA). The standard silicon nitride tip had a spring constant of 0.012). The spring constant of triangular, 320pam long silicone nitride cantilever was determined using thermal method [4, 12]. 80 Bibliography [1] T. Abraham, S. Giasson, J. Gohy, and R. Jerome. Direct measurements of interactions between hydrophobically anchored strongly charged polyelectrolyte brushes. Langmuir, 16:4286-4292, 2000. [2] M. Bushmann and A. Grodzinsky. A molecular model of proteoglycan-associated electrostatic forces in cartilage mechanics. Journal of BiomechanicalEngineering, 117:179-192, May 1995. [3] H.J. Butt. Electrostatic interaction in atomic force microscopy. Biophysical Journal, 60:777-785, October 1991. [4] H.J. Butt and M. Jaschke. Calculation of thermal noise in atomic force mi- croscopy. Nanotechnology, 6:1-7, January 1995. [5] B. Capella and G. Dietler. Force-distance curves by atomic force microscopy. Surface Science Report, 34(1):5-104, July 1999. [6] P.G. de Gennes. Polymers at an interface: a simplified view. Advances in Colloid Interface Science, 27:189-209, July 1987. [7] W. Ducker, T. Senden, and R. Pashley. Direct measurement of colloidal forces using an atomic force microscope. Nature, 353:239-241, September 1991. [8] S. Eisenberg and A. Grodzinsky. Swelling of articular cartilage and other connective tissues. Journal of Orthopaedic Research, 3:148-159, 1985. [9] A. Fick. On liquid diffusion. Philosophical Magazine, 10:30-39, 1855. 81 [10] A. Grodzinsky. Fields, forces and flows in biological tissues and membranes. Class notes for 6.561j at M.I.T., Massachusetts Institute of Technology, 1998. [11] H. Hansma. Atomic force microscopy of biomolecules. Journal of Vacuum Science and Technology, 14:1390-1394, March 1996. [12] J. Hutter and J. Bechhoefer. Calibration of atomic force microscope tips. Nanotechnology, 6:1-7, January 1995. [13] G.K. Jennings, J.C. Munro, T.H. Yong, and P.E. Laibinis. Effect of chain length on the protection of copper by n-alkanethiols. Langmuir, 14:6130-6139, August 1998. [141 Jiang-Ti Kong. Measuring the electrostatic repulsion forces between gly- cosaminoglycans using the atomic force microscope. Master of science thesis, Massachusetts Institute of Technology, Materials Science and Engineering Department, September 1999. [15] W. Lai, J. Hou, and V. Mow. A triphasic theory for the swelling and deformation behaviors of articular cartilage. ASME Journal of biomechanical engineering, 3:245-258, 1991. [16] A. Maroudas. Adult Articular Cartilage. Freeman, 215 edition, 1979. [17] A. Maroudas. Physical Chemistry of Articular Cartilage and the Intervertebral Disk in Joints and Synovial Fluid, volume II. Academic Press, New York, L. Sokoloff edition, 1980. [18] C. Miller, P. Cuendet, and M. Gritzel. Adsorbed w-hydroxy thiol monolayers on gold electrodes: evidence of electron tunneling to redox species in solution. Journal of Physical Chemistry, 95:877-886, 1991. [19] H. Ohshima. Electrostatic repulsions between two parallel plates covered with polymer brush layers. Colloid Polym. Science, 277:535-540, 1999. 82 [20] C. Ortiz, J. Seog, D. Dean, and A. Gordzinsky. Molecular mechanics of cartilage: investigation of macromolecular intermolecular interactions via high-resolutioon force spectroscopy. In 221st ACS National Meeting, San Diego, April 2001. ACS. [21] V.A. Parsegian and D. Gingell. On the electrostatic interaction across a salt solution between two bodies bearing unequal charges. Biophysical Journal, 12:11921204, 1972. [22] A. Plaas, L. West, S. Wong-Palms, and F. Nelson. Glycosaminoglycan sulfation in human osteoarthritis. Journal of Biological Chemistry, 273:12642-12649, May 1998. [23] J. Seog, D. Dean, A. Gordzinsky, and C. Ortiz. Modeling and measurement of glycosaminoglycan electrostatic interactions by high-resolutioon force spectroscopy. In 221st ACS National Meeting, San Diego, April 2001. ACS. [24] J. Seog, D. Dean, A. Plaas, S. Wong-Palms, I. Lee, J. Kong, P. Laibinis, A. Gordzinsky, and C. Ortiz. Cartilage molecular mechanics: detection of GAG electrostatic interactions by high-resolutioon force spectroscopy. In 47th annual meeting of the orthopaedic research society, San Francisco, February 2001. ORS. [25] N6ra Szasz. Transport and binding of insulin-like growth factors to articular cartilage. Master of engineering thesis, Massachusetts Institute of Technology, Electrical Engineering and Computer Science Department, May 1999. [26] E. Verwey, J. Overbeek, and K. van Nes. Theory of the Stability of Lyophobic Colloids. Elsevier, New York, 1948. [27] T. Weiss. Cellular Biophysics, volume I. MIT Press, Cambridge, Massachusetts, 1996. 83