Structural Heterogeneity in Silk Fibers and its Effects on Failure Mechanics and Supercontraction ARCHIVES MASSACHUSETTS by Tristan Giesa Dipl.-Ing. Mechanical Engineering RWTH Aachen University, 2011 JUL 02 2015 LIBRARIES Submitted to the Department of Civil and Environmental Engineering in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 2015 2015 Massachusetts Institute of Technology. All rights reserved. Signature of Author Signature redacted Department of C il nd Environmental Engineering May 21, 2015 Certified by Signature redacted ___ Markus J. Buehler Professor of Civil and Environmental Engineering, Department Head Thesis Supervisor Accented by _ __ _ _ Signature redacted, idi Nepf Donald and Martha Harleman Professor of Civil and Environmental En neering Chair, Graduate Program Committee 1 INISTITUTE OF TECHNOLOLGY 2 Structural Heterogeneity in Silk Fibers and its Effects on Failure Mechanics and Supercontraction by Tristan Giesa Dipl.-Ing. Mechanical Engineering RWTH Aachen University, 2011 Submitted to the Department of Civil and Environmental Engineering on May 21, 2015 in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Field of Mechanics and Materials Abstract Spider dragline silk is a protein material that has evolved over millions of years to become one of the strongest and toughest natural fibers known. Silk features a heterogeneous structure that comprises P-sheet crystals embedded in an amorphous matrix. However, it is not fully understood how the heterogeneity of silk affects its mechanical properties. First, the origin of the nanoscale heterogeneity during the Nephila Clavipes dragline silk assembly is investigated. Using molecular dynamics simulations, a shear flow at natural pulling speeds is modelled and the secondary structure transitions as well as shear stresses in the silk protein chains are determined. It is shown that under shear stresses beyond the elastic regime, silk undergoes an a - P-transition in the spinning duct. The stability of the assembled P-sheet structure seems to arise from a close proximity of the a-helices in the silk solution. The smallest molecule size that might give rise to a silk-like structure is determined to comprise four to six repeats of the silk sequence. Establishing the 3 molecular details of the assembly can guide the design of microfluidic devices and the synthesis of bioinspired protein materials. Second, it is shown how the heterogeneity of silk fibers, specifically its crystalline phase, relates to its fracture mechanical properties: strength and toughness. Analytical fracture mechanical arguments are presented to illustrate the relation between fracture strength and heterogeneity in silk and other biopolymers. Nanoconfinement and flaw tolerance are presented as natural strategies to increase the mechanical performance of the entire material system. It is shown that the consideration of interatomic interactions alone cannot explain the fracture strength observed in biological fibers. Instead, structures at multiple length-scales must be considered to explain the remarkable mechanical performance and resilience of silk. Third, the interaction of water with silk's heterogeneous nanostructure is investigated. At high humidity, some spider dragline silks will shrink up to 50%, a phenomenon known as supercontraction. The molecular origin of dragline silk supercontraction is explored using a full-atomistic model and molecular dynamics supported by in situ Raman spectroscopy and mechanical testing performed at the Max Planck Institute in Potsdam, Germany. Tyrosine and Arginine are identified as the key residues in the Nephila Clavipes silk sequence that control supercontraction. A genetic engineering strategy to alter silk's behavior to industrial requirements is proposed, where sequence mutations reduce or even reverse the supercontraction mechanism. Thesis Supervisor: Markus J. Buehler Title: Professor and Department Head of Civil and Environmental Engineering 4 Acknowledgements I would like to take this opportunity to express my gratitude to those who have supported me throughout my graduate studies at MIT. I would like to thank my advisor, Prof. Markus J. Buehler, for his mentorship, advice, and for the opportunity to work and learn in such a great collaborative and inspiring environment. I am grateful for the advice and support of my committee members, Prof. Oral Buyukozturk, Prof. Pedro Reis, and Prof. Niels HoltenAndersen. I am indebted to my amazing collaborators with whom I had the pleasure working on a diverse set of exciting research topics: Dr. David Spivak, Prof. Carole Perry, Prof. David Kaplan, Prof. Joyce Wong, Prof. Nicola Pugno, Dr. Admir Masic, and Dr. James Weaver. I am very grateful to Leon Dimas, Dieter Brommer, and Francisco Martinez, for their friendship and everlasting support, even in most difficult times. Many thanks also to my past and present colleagues from the Laboratory for Atomistic and Molecular Mechanics at MIT: Talal Al-Mulla, Melis Arslan, Laura Batty, Graham Bratzel, Shu-Wei Chang, Chun-The Chen, Chia-Ching Chou, Steve Cranford, Baptiste Depalle, Nina Dinjaski, Davoud Ebrahimi, Andre Garcia, Alfonso Gautieri, Greta Gronau, Grace Gu, Kai Jin, Gang-Seob Jung, Shangchao Lin, Shengjie Ling, Flavia Libonati, Reza Mirzaeifar, Arun Nair, Seunghwa Ryu, Max Solar, Anna Tarakanova, Olena Tokareva, Steve Palkovic, and Zhao Qin. Finally, I express my deepest gratitude to my family and friends who have supported and encouraged me in all my endeavors. This research was funded by grants by NSF, ARO, BASF-NORA and NIH. Their support is greatly appreciated. Cambridge, United States, June 2015 5 6 List of Journal Publications I am the author of all the work presented in this thesis. Research was conducted in the Department of Civil and Environmental Engineering at the Massachusetts Institute of Technology. Part of the work presented here has been published in peer-reviewed journal papers and book chapters (chronological): 1. T. Giesa, M. Arslan, N.M. Pugno, M.J. Buehler, Nanoconfinement of Spider Silk Fibrils Begets Superior Strength, Extensibility and Toughness. Nano Letters 11 (11), 5038-5046, 2011 2. D.I. Spivak, T. Giesa, E. Wood, M.J. Buehler, Category Theoretic Analysis of HierarchicalProtein Materials and Social Networks. PLoS One 6 (9), e23911, 2011 3. T. Giesa, D.I. Spivak, M.J. Buehler, Reoccurring Patterns in Hierarchical Protein Materials and Music: The Power of Analogies. BioNanoScience 1 (4), 153-161, 2011 4. T. Giesa, G. Bratzel, M.J. Buehler, Modeling and Simulation of Hierarchical Protein Materials. In: Nano and Cell Mechanics: Fundamentals and Frontiers, John Wiley & Sons, Ltd, 389-409, 2012 5. G. Gronau, S.T. Krishnaji, M.E. Kinahan, T. Giesa, J.Y. Wong, D.L. Kaplan, M.J. Buehler, A review of combined experimental and computational procedures for assessing biopolymer structure-process-propertyrelationships. Biomaterials 33 (33), 8240-8255, 2012 6. T. Giesa, D.I. Spivak, M.J. Buehler, Category Theory Based Solution for the Building Block Replacement Problem in Materials Design. Advanced Engineering Materials 14 (9), 810-817, 2012 7. T. Giesa, N.M. Pugno, M.J. Buehler, Natural Stiffening Increases Flaw Tolerance of Biological Fibers. Physical Review E 86 (4), 041902, 2012 7 8. T. Giesa, M.J. Buehler, Spidermans Geheimnis. Physik in unserer Zeit 44 (2), 72-79, 2013 9. T. Giesa, M.J. Buehler, Nanoconfinement and the Strength of Biopolymers. Annual Reviews of Biophysics 42, 651-673, 2013 10. L.S. Dimas, T. Giesa, MJ Buehler, Coupled Continuum and Discrete Analysis of Random Heterogeneous Materials: Elasticity and Fracture. Journal of the Mechanics and Physics of Solids 63, 481-490, 2014 11. T. Giesa, N.M. Pugno, J.Y. Wong, D.L. Kaplan, M.J. Buehler, What's Inside the Box? - Length Scales that Govern Fracture Processes of Polymer Fibers. Advanced Materials 26 (3), 412-417, 2014 12. L.S. Dimas, D. Veneziano, T. Giesa, M.J. Buehler, Random Bulk Propertiesof Heterogeneous Rectangular Blocks with Lognormal Young's Modulus: Effective Moduli. Journal of Applied Mechanics 82 (1), 011003, 2015 Journal papers in submission or revision: 13. T. Giesa, R. Schuetz, P. Fratzl, A. Masic, M.J. Buehler Molecular Origin of Supercontraction in Spider Dragline Silk Revealed by Simulation and Experiment. In submission, 2015. 14. T. Giesa, C.C. Perry, M.J. Buehler, Secondary Structure Transition and CriticalStress during the Assembly of Spider Silk Fibers. In submission, 2015. 15. L.S. Dimas, D. Veneziano, T. Giesa, M.J. Buehler, ProbabilityDistributionof Fracture Elongation, Strength and Toughness of Notched Rectangular Blocks with Lognormal Young's Modulus. In revision, 2015. 16. T. Giesa, R. Jagadeesan, D. I. Spivak., M.J. Buehler, A Python Library for MaterialsArchitecture. In submission, 2015. 8 Table of Contents A b stra ct .............................................................................................................................. 3 Acknowledgements ..................................................................................................... 5 List of Journal Publications........................................................................................ 7 Table of Contents ....................................................................................................... 9 1 In trod u ction ............................................................................................................. 1.1 Biological Materials as Template for Structural Design ......................... 13 1.2 Spider Silk as Model Material ................................................................... 20 1.2.1 Properties of Silk Fibers ..................................................................... 21 1.2.2 Tensile Behavior of Silk Dragline Fibers........................................... 23 1.2.3 Defects and Flaws in Silk Fibers ........................................................ 25 1.3 2 Research Hypothesis.................................................................................... Meth od s.................................................................................................................... 2.1 Atomistic Modeling ..................................................................................... 26 29 30 2.1.1 Classical Molecular Dynamics .......................................................... 31 2.1.2 Quantities determined from Molecular Dynamics ............. 32 2.1.3 Steered Molecular Dynamics............................................................. 34 2.1.4 Molecule Shape Calculation ............................................................... 34 2.1.5 Replica Exchange Molecular Dynamics............................................. 35 2.1.6 Entropy calculation............................................................................... 35 2.1.7 Free Energy of Solvation...................................................................... 37 2.1.8 Simulated Raman and Infrared Spectrum......................................... 38 2.1.9 Data Analysis........................................................................................ 41 Analytical approaches to materials failure/ fracture.............................. 41 2.2 3 13 2.2.1 Linear elastic fracture mechanics...................................................... 41 2.2.2 Polymer fracture mechanics ............................................................... 42 Creation of Heterogeneity - Silk Assembly.................................................... 45 9 3.1 Backgrou nd ................................................................................................... 46 3.1.1 Fiber formation- spinning from solution.............................................. 47 3.1.2 Silk structure after spinning ............................................................... 50 3.1.3 Simulation Setup ................................................................................. 52 3.2 Size Dependence of Structural Transition ............................................... 58 3.3 Structural Stability of the Assembled Silk Molecules ............................. 62 3.4 Molecular Details of the Transition Mechanism...................................... 65 3.5 Layered Structures ...................................................................................... 70 3.6 C onclu sion .................................................................................................... 72 4 Heterogeneity and Nanoconfinement of Fibrils to Increase Strength and Tou gh n ess ....................................................................................................................... 4.1 Background ................................................................................................... 74 4.1.1 Nanoconfinement Strategy .................................................................. 75 4.1.2 Confinement of Polymers .................................................................... 80 4.1.3 Strength of Polymers .......................................................................... 82 4.1.4 The Relation between Confinement and the Strength of Polymers .87 4.1.5 Nanoconfinement in Silk .................................................................... 94 Fracture Mechanics Analysis...................................................................... 98 4.2 5 73 4.2.1 Importance of the Process Zone......................................................... 99 4.2.2 Derivation of the Process Zone Size .................................................... 100 4.2.3 Fracture Length Scales in Silk Fibers...................................................106 4.2.4 Continuum Fracture Mechanics Analysis ................... 108 4.3 The Importance of Heterogeneity in Silk Fibers....................................... 110 4.4 C on clu sion ...................................................................................................... 113 Supercontraction - Silk's Interaction with Water ............................................ 115 5.1 Back grou nd .................................................................................................... 116 5.1.1 Mechanism of Supercontraction .......................................................... 5.1.2 Combined Simulation and Experimental Approach........................117 5.1.3 Molecular Dynamics Setup................................................................... 10 117 119 Supercontraction of the silk wildtype ........................................................ 5.2 120 5.2.1 Silk Supercontraction in Simulation and Experiment ...................... 120 5.2.2 Secondary Structure Change during Supercontraction ................... 123 Molecular Origin of Supercontraction ....................................................... 5.3 124 5.3.1 Raman Spectroscopy and Hydrogen Bonding .................................. 124 5.3.2 Simulated Infrared Spectrum and Vibrational Density of States .... 127 5.3.3 Energy Balance and Supercontraction Stress.....................................129 5.3.4 Conformational Changes ...................................................................... 5.3.5 Hydrogen Bonding of Tyrosine and Arginine...................................148 5.3.6 Key Residues for Supercontraction..................................................... 136 152 5.4 Controlling Supercontraction...................................................................... 153 5.5 C on clu sion ...................................................................................................... 157 6 Summary and Outlook ........................................................................................ 159 7 A p p en d ix ............................................................................................................... 163 7.1 Secondary Structure and Shear Stress Trajectories .................................. 7.2 Probability for the a-p-Transition...............................................................167 7.3 N omenclature ................................................................................................ 172 7.4 R eferences ....................................................................................................... 177 7.5 List of Figu res ................................................................................................ 197 7.6 List of Tables .................................................................................................. 211 11 163 12 1 Introduction Parts of the review presented in this chapter have been published in: " T. Giesa, M. Arslan, N.M. Pugno, M.J. Buehler, Nanoconfinement of Spider Silk Fibrils Begets Superior Strength, Extensibility and Toughness. Nano Letters, 11, 11, pp. 5038-5046, 2011 " T. Giesa, G. Bratzel, M.J. Buehler, Modeling and Simulation of Hierarchical Protein Materials. In: Nano and Cell Mechanics: Fundamentals and Frontiers, John Wiley & Sons, Ltd, pp. 389-409, 2012 * T. Giesa, M.J. Buehler, Spidermans Geheimnis. Physik in unserer Zeit 44 (2), 72-79, 2013 * T. Giesa, M.J. Buehler, Nanoconfinement and the Strength of Biopolymers. Annual Reviews of Biophysics, 42, pp. 651-673, 2013 1.1 Biological Materials as Template for Structural Design Most biological materials are organized in complex hierarchical structures [1], some of which have already been mimicked by engineers using synthetic pathways such as transcription, synergetic assembly, morphosynthesis (including chemical transformations in confined geometries to produce complex structures), and integrative synthesis. Latter combines all previously mentioned methods to produce materials with complex morphologies [2]. Due to advances in observation methods during the last few centuries, scientists have gained deeper insight into the way structures and materials are fabricated in nature. Sanchez et al. [2] state that new materials and systems produced by man must aim at higher levels of sophistication and miniaturization, be recyclable and reduce the environmental footprint, implying also increased reliability and 13 durability while consuming less energy in fabrication. By elucidating the design and construction principles of living organisms, new material synthesis pathways become accessible. It has become possible to understand and even imitate the superior performance, variability, and effectiveness seen in biological systems. This development has culminated into the field of biomimetics [3]. By controlling material properties on the length scales of Angstroms and nanometers through new processing methods, it is now possible to create functional surfaces and polymer nanostructures with tailored mechanical, optical, thermal, or electrical responses. Still, the understanding (and mimicking) of the complex assembly processes that form polymer-based/composite biomaterials like bone, silk, collagen, and many biological tissues remains a major challenge. Protein materials are found in complex structures such as cells, organs, or organisms. An analysis of their composition reveals universal basic constituents (building blocks such as a-helices, -sheets) but also highly specific features (filament assemblies, nanocrystals in spider silk or tendon fascicles, etc.) [4]. These examples illustrate that the coexistence of universality and diversity is an overarching feature in protein materials, as characterized by the UniversalityDiversity paradigm, wherein universality tends to dominate at smaller levels and diversity is found predominantly at larger levels and used to create many different functional properties [5]. Biological materials such as spider silk and diatom algae arrange in structural hierarchies and optimize their behavior in regard to the environmental requirements. Diatoms exhibit high structural stiffness combined with high robustness to protect against predators, and spider webs show high energy absorption and extensibility for catching prey while localizing web damage [6, 7]. Similarly, in the glassy sponge Eucleptella, silica nanospheres are arranged at multiple levels of hierarchy to constitute a skeleton with high structural stability 14 at minimum cost [8]. The teeth of sea urchins and the lamellar structure of mollusk shells are other examples for structural hierarchies in biomaterials that lead to extremely strong and tough structures [9]. Earlier studies showed that in materials like bone or wood, for example, the structural assembly of the basic building blocks collagen, water, hydroxyapatite minerals, hemicelluloses, and lignin governs the mechanical properties at different length scales with similar mechanisms, despite the differences in the building blocks and the overall material properties [10, 11]. Figure 1 shows the hierarchical structure of diatom algae and glassy sponges as examples of complex assembled natural systems. a b Figure 1 I Complex hierarchical structures found in natural materials. (a) Scanning electron microscopy (SEM) pictures displaying the intricate hierarchical porous silica wall structure of diatoms. Figure adapted from [12], with permission from Elsevier. (b) SEM pictures of the mineralized skeletal system of Eucleptella. The caged structure consists of struts (bundled spicules) which themselves are a ceramic composite with laminated silica layers and organic interlayers. Figure adapted from [8], copyright @ 2005, with permission from the American Association for the Advancement of Science. It is remarkable how the exceptionally complex functionality found in natural biological systems is created despite (i) a limited number building blocks, e.g. the 20 amino acids in protein materials, (ii) constraints in available material volume and energy for synthesis, and (iii) only a handful of simple chemical interactions, generally referred to as interaction rules [5, 13-17]. The manipulation of single building blocks has fundamental influences on the overall system behavior. A single point-mutation in a DNA strand can make the difference between health 15 and disease [18]. On the other hand, the localized failure of larger structural elements does not influence the total system behavior, in accordance with the robustness of hierarchical systems. This is related to a high redundancy within the structure. Computational modeling and simulation of materials seeks to bridge analytical theory and experimental observation in order to both explain these phenomena and predict materials behavior. The computational tools available for multiscale simulation and engineering of protein materials now cover a similar length scale range as experimental tools (see chapter 0) and can provide crucial insights to deformation mechanisms, particularly at the nanoscale. As computer performance continues to advance, multiscale computational modeling of synthetic and biological materials will further enable to improve the understanding of processes and provide a rigorous basis for the bottom-up design of advanced materials, coupled with multiscale experiment. This comprehensive view of structures and materials forms the foundation to a new field of study, biomateriomics [19]. Biomateriomics aims at elucidating the basic components and building principles selected by evolution to propose more reliable, efficient and environmentally benign materials and requires a multidisciplinary approach. Within this framework, the most important and interesting features of biological materials are: * Reuse of structural constituents Protein materials can be found in vastly complex structures such as cells, organs or organisms. The coexistence of universality and diversity is an overarching feature in protein structures, where universality tends to dominate at smaller levels whereas diversity is found predominantly at 16 larger, functional levels [5]. The universal building blocks can be defined depending on the depth of analysis, e.g., the 20 amino acids, the chemical elements, or quantum level elements like strings. * Multi-constituent hierarchical build-ups Many materials and structures engineered by humans bear a conflict between strength and toughness; strong materials are often fragile, while robust materials tend to be soft [4]. In extreme conditions, only high safety factors and thus bigger amount of resources can guarantee the strength of engineered materials. This can be overcome by the arrangement of structures at multiple scales. One of nature's remarkable features is its ability to combine (bio)organic and inorganic components at the nanoscale [20]. In materials like bone, universal patterns form a nanocomposite of strong but brittle minerals (hydroxyapatite) and soft but ductile biopolymers (collagen) through the assembly into complex shapes on seven levels of hierarchy, giving rise to improved mechanical properties. Figure 2 shows the hierarchical build-up of two prominent natural materials, spider silk and bone. The strength and toughness observed in these materials serves as model to engineering composites. Advances in 'soft chemistry' during the past ten years have produced original hybrid materials with controlled porosity and/or texture resulting in easy-to-process materials [2]. These offer many advantages such as tunable physical properties, high photochemical and thermal stability, chemical inertness and negligible swelling, both in aqueous and organic solvents with applications to smart devices, sensors and catalysis. 17 Silk 1PM n-m 15 Fiber Heteronanocomposite Fibril f-sheet and semiamorphous phase Web Bone Amino acids - A Tropocollagen - 300 nrn Bone tissue - 50 cm Mineralized collagen fibrils -mm Osteons and Harvesian canals - 100mm Fibril arrays - 10 MM Fiber pattems -50mm Figure 2 Hierarchical structures of biological materials such as spider silk and bone. Biological materials are designed bottom-up to overcome fundamental strength limits at the nanoscale. Spider dragline silk, specifically the protein MaSpi, consists of -sheet nanocrystals embedded in an amorphous phase. These are aligned along the fiber axis to form fibrils of size 20-150 nm. Hundreds of fibrils are spun together in a fiber that eventually forms the frame of an orb web. Compact bone is composed of osteons that surround and protect blood vessels. Osteons have a lamellar structure. Each individual lamella is composed of fibers arranged in geometrical patterns. These fibers are the result of several collagen fibrils, each linked by an organic phase to form fibril arrays. Each array makes up a single collagen fiber. The mineralized collagen 2010, fibrils are the basic building blocks of bone. Bone figure adapted from Reference [20], copyright James Dr. of courtesy Annual Reviews. Web image courtesy of Charles J. Sharp. Silk figure composition Weaver, Harvard University. 18 0 Self-assembly under moderate conditions In order to control structure and shape below the scale where most engineering techniques become unfeasible, structural self-assembly to control the chemical synthesis of small-scale structures upwards provides a promising bottom-up approach. Key to engineering applications is the thorough understanding of how simple chemical complexes, i.e., building blocks like amino acids, DNA, viruses, nanoparticles, or enzymes as catalysts, interact under moderate conditions (room-temperature and atmospheric pressure) to form complex 3D structures. Current models cannot predict protein folding unambiguously [21]. Self-healing Self-healing is the functional repair to mitigate damage, for example from an impact event, thus progressing from a conventional damage tolerance philosophy. It is mainly related to two biological mechanisms [22]: On the molecular level, sacrificial bonds, such as hydrogen bonds, dynamically break and reform giving rise to a quasi-plastic behavior without fracture (related for example to spider silk's second deformation stage, see chapter 1.2.2). On the macroscale, cyclic replacement of material, for instance in bone and plants by specialized cells or in tissues by intermediate tissues, forms the governing repair mechanism. * Lightweight design The density of biological materials such as silk is generally less than 3 g/cm 3 (compare to steel: 8 g/cm 3) due to the hierarchical structure that strongly reduces the amount of needed resources [23]. 19 0 High robustness and strength (damage tolerance) In most applications, catastrophic failure needs to be avoided by using strong and tough (fracture-resistant) materials. Hard materials tend to be brittle because the high stresses at the crack tip cannot be dissipated, whereas soft materials dissipate the stress by plastic deformation. Natural materials rely on intrinsic and extrinsic toughening mechanisms (see chapter 4.1.4.2). Intrinsic mechanisms (based on plastic deformation, e.g., fiber sliding) are usually originated from smaller length scales (akin to dislocations in metal), whereas extrinsic toughening mechanisms (e.g. fiber bridging) usually take place on micrometer scales [24]. 1.2 Spider Silk as Model Material Spider and silkworm silk are among the most studied natural protein materials, since they exhibit most of the desired features of a composite material discussed in the previous chapter. Silk is a hierarchically structured protein fiber with a high tensile strength and great extensibility, making it one of the toughest materials known [25-27], in spite of the material's simple protein building blocks [28-30]. In contrast to synthetic polymers based on petrochemicals, silk is spun into strong and totally recyclable fibers at ambient temperatures, low pressures, and with water as the solvent. However, biomimetic reproductions of silk remain a challenge because of silk's characteristic microstructural features that can only be achieved by controlled self-assembly of protein polymers with molecular precision [31, 32]. Unlike silkworms, some spiders can use different glands to create up to seven types of silk, from the strong dragline to the viscoeleastic capture silk and tough eggsack casing [27]. 20 1.2.1 Properties of Silk Fibers Dragline silk, containing a high fraction of densely hydrogen-bonded (Hbonded) domains, is used to provide the structural frame for the web and has an elastic modulus of up to 10 GPa [25]. Capture silk, on the other hand, is a viscid biofilament containing cross-linked polymer networks and has an elastic modulus that is comparable to that of other elastomers [33]. A comparison of silk to other composite and classical engineering materials is shown in the Ashby plot in Figure 3. 10000 Engineering Natural Polyrmers and Polymer Composites Ceramics Engineering 1000 W Alloys 100 Natural Cellular Materials BenOO ineering ,0, C posites Engir P CO 10 Natural Ceramics and Ceramic Comnposites 1 Elastomers 0.1 0.01 0.1 1 10 Density [g/cm3I Figure 3 1 Density vs. failure strength of synthetic and natural materials as Ashby plot. Adapted from Reference [34]. While there are many types of silk with different properties, this thesis focuses on the dragline silk (specifically the protein MaSpi) of orb-weaving spiders that is known to be extremely strong, extensible and tough [35-38]. Silk fibers 21 typically feature an initial modulus up to 10 GPa, a high extensibility exceeding 50-60% strain at failure, and a tensile strength of 1-2 GPa [37-41], which results in toughness values of several times that of KevlarTM [42]. In addition to the relatively large ultimate strength of spider silk, comparable to that of steel, silk features a strength-to-density ratio that is up to ten times higher than that of steel because of the material's small density (-1.3 g/cm 3). As shown in Figure 4, silk features a hierarchical structure, where the nanoscale geometry is characterized by a network of silk repeat units that each consist of a P-sheet nanocrystal embedded in semi-amorphous protein domains. The sequence of silk repeat units (in one-letter amino acid codes, here as an example for Major Ampullate Spidroin 1, MaSpi, of Nephila clavipes) is (GAGAAAAAAGGAGQGYGLGS QGGRGLGGQ)a where the bolded 'A' (Alanine) identifies the region that forms -sheet nanocrystals and the rest forms semi-amorphous domains. poly-alanine region P-sheet crystal ... QGAG AAAAAA GGAGQ... semi-amorphous phase - * 0.2 nm hydrogen bonding 20 - 150 nm pm-sheet 10 nm content 15-50 % fiber 1 fibril Figure 4 | Silk features a hierarchical structure, where P-sheet crystals play a key role in defining the mechanical properties by providing stiff and orderly cross-linking domains embedded in a semi-amorphous matrix that consists predominantly of less ordered structures. These f-sheet nanocrystals, bonded by means of assemblies of H-bonds, have dimensions of a few nanometers and constitute roughly 15-50% of the silk volume. Adapted from Reference [43]. 22 While rubber is extensible, and KevlarTM is stiff and strong, silks feature a combination of strength and toughness not typically found in synthetic materials. It is known that -sheet crystals at the nanoscale play a key role in defining the mechanical properties of silk by providing stiff and orderly crosslinking domains embedded in a semi-amorphous matrix that consists predominantly of less ordered structures [44, 45]. These P-sheet nanocrystals, bonded by means of assemblies of H-bonds, have dimensions of a few nanometers and constitute roughly 15-50% of the silk volume. When silk fibers are stretched, the P-sheet nanocrystals reinforce the partially extended and oriented macromolecular chains by forming interlocking regions that transfer the load between chains under lateral loading, similar to their function in other structural proteins [46]. The hierarchical network of spider draglines, contrasting synthetic elastomers like rubber, enables quick energy absorption and efficiently suppresses vibration during an impact [42]. Rubbers, composed of random polymer chains, display an elasticity regime that is primarily due to the change in conformational entropy of these chains. In contrast, the amorphous chains in silk filaments are extended and held in partial alignment with respect to the fiber axis in its natural dry state, resulting in remarkably different mechanical behaviors from rubbers [45, 46]. In chapter 5, the transition of hydrated silk into a rubbery state will be discussed (supercontraction effect). 1.2.2 Tensile Behavior of Silk Dragline Fibers The mechanical behavior of silk is explained in detail in Reference [44, 47]. Molecular-level studies elucidated the structure and role of P-sheet nanocrystals and semi-amorphous protein domains during deformation, and the mechanical parameters for the behavior of a single repeat unit of silk were extracted [44, 48, 49]. The mechanical behavior of silk fibers under tensile stretching is highly nonlinear. Beyond an initial high-stiffness regime spider silk softens at the so23 called 'yield point' where the stress-strain response gives way to a plateau, eventually leading to a stiffening regime prior to failure [50]. These mechanisms result in the characteristic softening-stiffening stress-strain response that is found for many different types of silk [37-39]. Figure 5 displays schematically a typical stress-strain curve of spider silka dn classifies its deformation regimes. Regime I is characterized by a linear-elastic response dominated by homogenous stretching before protein unfolding begins. The transition from Regime I to Regime II is marked by the beginning of the rupture of hydrogen bonds (H-bond) in 3 1 0-protein helices that make up the semi-amorphous domains, and the unraveling of these proteins continues until all hidden length is exhausted. Regime II is the key to the extensibility of silk. Regime III reflects the stiffening behavior that sets in after the exhaustion of unfolding events and the alignment of polypeptide chains. This facilitates the deformation of P-sheet nanocrystals that leads to a significant stiffening of the material in Regime III. Regime IV involves a brief softening behavior as P-sheet nanocrystals fail under stick-slip deformation leading to the breakdown of Psheet nanocrystal cross-links, and eventual material failure. The mechanical stability of P-sheet nanocrystals is the key to the ultimate strength molecularlevel silk since they are the last molecular elements that break. 24 1500 - Experiment/Simulation Quad-Linear Fit 1000 III C 500 a'' 0 0.1 0.2 0.4 0.3 Strain f; 0.5 0.6 Figure 5 | Stress-strain behavior for a defect-free silk fiber, noting the key transition points between the four regimes marked by molecular events at the molecular scale. The transition from Regime I to Regime II marks the onset of unfolding of the semi-amorphous phase of silk; the transition from Regime II to Regime III marks the onset of stretching of the -sheet nanocrystal phase. In Regime IV (-sheet nanocrystals fail via a stick-slip mechanism, eventually leading to failure. Figure adapted from Reference [431. 1.2.3 Defects and Flaws in Silk Fibers At a much larger scale, experimental studies have shown that silk fibers contain many defects that act as stress concentrators, including cavities, surfaces or tears [51-53]. For example, Figure 6 shows images of crack-like cavities in Nephila madagascariensis dragline silk [51]. These defects feature sizes that reach several hundred nanometers and are crucial in the consideration of mechanical properties as they serve as seeds for material failure through localized deformation (in fracture mechanics defects are known to lead to local stress concentrations) [54, 55]. Nevertheless, despite the presence of defects, silk fibers display remarkable mechanical properties [56, 57]. 25 Figure 6 1 Microscopic images of Nephila madagascariensis dragline silk fibers showing the skin-core structure as well as flaws and cavities in the material. The white arrows point in the axial fiber direction, and the red ellipses highlight some of the defects found in the structure. Pictures reprinted from [51], C copyright 1998, with permission from John Wiley & Sons, Inc. 1.3 Research Hypothesis As shown in the previous chapter, silk serves as a model fibrous material and has the potential to replace expensive synthetic fibers in composites. Unfortunately, silk has several disadvantages: it is very temperature sensitive and flammable, it interacts strongly with water and the assembly process is quite complex. For industrial purposes, silk has to be either modified (e.g., through genetic engineering) to mitigate some of these problems or a new material has to be created that contains silk's desirable features. To study the nanoscale mechanisms that lead to silk's heterogeneity and that control silk's mechanical properties, a multidisciplinary and multiscale approach is necessary. This work focusses mainly on the computation of silk on the nanoscale using molecular dynamics simulation. However, simulations complement and are validated with experimental observations from literature and from collaborators. First, the origin of the nanoscale heterogeneity during the Nephila Clavipes dragline silk assembly is investigated. This is of interest since artificial spinning of recombinant silk has not yet yielded fibers that compare in strength in 26 toughness to their natural counterparts. It is hypothesized, that shear stresses in the spinning duct need to reach a critical value in order to allow the silk a - Ptransition from solution to fiber. Furthermore, it is hypothesized that the stability of the crystal is determined by the size of the silk molecule in the solution and that there is a minimum size that ensures this stability. Establishing the molecular details of the assembly can guide the design of microfluidic devices and the synthesis of bioinspired protein materials, even beyond silk. Second, it is hypothesized that the heterogeneous structure during the assembly is crucial to the mechanical performance of the fiber. Specifically, the nanoconfinement strategy employed in biomaterials leads to a flaw tolerant structure. Structures at multiple length-scales must be considered to explain the remarkable mechanical performance and resilience of silk. Understanding the trigger of strengthening mechanisms has a tremendous impact on the design of novel hierarchical materials. Third, the interaction of water with silk's heterogeneous nanostructure is investigated. At high humidity, some spider dragline silks will shrink up to 50%, a phenomenon known as supercontraction. This effect is not necessarily desirable. It is hypothesized that a detailed understanding and control over silk's heterogenous nanostructure can help to identify the molecular origin of supercontraction. Furthermore, a genetic engineering strategy (targeted sequence mutations) can reduce or even reverse the supercontraction mechanism while maintaining the mechanical properties. 27 28 2 Methods In this chapter, molecular simulation is described as a multiscale method to elucidate atomistic and molecular mechanisms of protein assembly and deformation. Figure 7 summarizes approximate length and time scale regimes of the tools for multiscale engineering. Computational tools predict and explain phenomena that are observed experimentally, but are limited to certain regimes due to constraints on computational performance [18]. While mesoscale and continuum modeling cannot capture atomistic details, they are trained by atomistic results from Density Functional Theory (DFT) and Molecular Dynamics (MD) simulations. They cover the same length scale range as experimental tools (e.g., atomic force microscopy, optical/ magnetical electromechanical systems and nano-indentation. 29 tweezers, micro- 0(300 nm. ps) Nano - a dentation MEMS testing >minr Collagen pnpette nMicro >m fibril C Continuum models Meso. scale models PSG I OpticalAnagnetic tweezers -g Nonreactive Atomic force microscopy Reactive MD iugraphy PS QM/DFT Transmission eetron NMR diffraction b Length scale pm unm diX-ray I: 10 Nanopaticles (nanowires. carbon nanotubes) DNA polypeptides Secondary protein structures (e.g. f-sheets, a-helices) Cells Tissues organs organisms Figure 7 1 Approximate length and time scale regimes of the tools for multiscale engineering. to Computational tools predict and explain phenomena that are observed experimentally, but are limited continuum and mesoscale While certain regimes due to constraints on computational performance. modeling (subpanel a) cannot capture atomistic details, they are trained by atomistic results from Density scale Functional Theory (DFT) and Molecular Dynamics (MD) simulations. They cover the same length tweezers, magnetical optical/ b), subpanel (AFM, microscopy force atomic (e.g. tools range as experimental classes microelectromechanical systems (MEMS, subpanel c) and nano-indentation. The lower part indicates [18], from reprinted Figure techniques. or scales of protein materials that can be studied with the respective Group. Publishing Nature copyright 2009, with permission from the 2.1 Atomistic Modeling Beginning from first principles, quantum mechanical methods such as the Hartree-Fock Theory solve many-electron wave functions, based on the Schr6dinger equation, to derive bond energies and chemical interactions, but are limited to small molecules and describe processes lasting only femtoseconds [58]. Density functional theory (DFT) employs functionals based on the spatially30 dependent electron density to explore the electronic structure of atoms and molecules such as the interactions of single peptides. For proteins, the energy landscape, the phonon modes (e.g., to capture dipole changes and the polarization), and the distribution of charges within the molecule can be determined. The current computational limit for these quantum methods is in the order of a few thousand atoms. To capture the formation and interactions of secondary structures of polypeptides (e.g., P-sheets and a-helices) and solventmediated processes on the order of nanoseconds, molecular dynamics (MD) simulations use force fields trained by DFT results [59, 60]. Molecular dynamics was developed in the 1950's to describe multi-body motion and fluid dynamics using first-principles calculations [61, 62]. Today, MD is widely used to predict behavior and model various materials phenomena. Simulations of proteins have experienced a significant advance in complexity [59], from early reports of two-dimensional Monte Carlo Method folding predictions of small polypeptides [63], to fully 3D atomistic molecular dynamics simulations of a ribosome [64], the cellular component that reads RNA and synthesizes proteins. 2.1.1 Classical Molecular Dynamics In essence, molecular dynamics simulates systems in a thermodynamical ensemble by solving a set of second order differential equations describing the relative motion of particles in a system. For a pair interaction between atom i and j, the equation of motion is given by Newtons law dz dtz _ - dU(r 1 d ) d. 2.1 i,j=1..N, drij where mi is the mass of atom i, rij the distance between the atoms, U the potential, and N the number of particles in the system. This equation is solved by 31 update schemes, e.g. the Verlet scheme using a timestep At. The motion (atomic coordinates and velocity) is recorded and all other quantities are derived from it. The key to molecular dynamics simulations is the determination of the potential U(r). This quantity is determined from a prescribed force field. While reactive forces fields such as ReaxFF are capable of capturing covalent bond chemistry, molecular dynamics using non-reactive force fields such as CHARMM [65] can simulate noncovalent interactions (e.g., electrostatics (hydrogen bonding), and van der Waals forces) involved in secondary structure changes. Since molecular dynamics is directly linked to statistical physics, state variables in the simulated system can be set or calculated. For a thermodynamics ensemble it is sufficient to specify three state variables. These variables can be for example total energy E, number of particles N, volume V, temperature T, pressure p, chemical potential M, enthalpy H. Simulations of proteins without chemical reactions usually use the NVT (canonical), NPT (isothermal-isobaric) or NVE (microcanonical) ensemble. MD software packages such as CHARMM, NAMD, LAMMPS, or GROMACS are equipped with algorithms to calculate and control these state variables. 2.1.2 Quantities determined from Molecular Dynamics The statistical quantities that can be found from a molecular dynamics trajectory are [66]: * Average potential energy U of a system with N particles and M configurations M Ui U = (U) = 2.2 j=1 32 Average kinetic energy K of a system with N particles and M configurations M N Imi K = (K) = 2.3 - " j=1 -i=1 " Temperature T 2 T = yNk; (K) 2.4 " Pressure p N * (rij - p = NkBT 1 dU drj 2.5 Mean square displacement N (Ar 2 ) = 2.6 (r (t) - ri(to))2 i=1 Velocity autocorrelation function (v(to)v(t)) = 111 i=1 * 2.7 vi(t)] vitt.) - * j=1 Virial Stress ( - i7g 1 6 k1 N + 1 sil mivikv,1 0 2 N rirj dU(r) r =i,j,ii 33 dr 2.8 r rij 2.1.3 Steered Molecular Dynamics Steered molecular dynamics (SMD), complements experimental Atomic Force Microscopy (AFM) by using a virtual spring moving at a constant velocity or maintaining a constant force to deform the protein [67]. SMD is a nonequilibrium simulation technique to simulate unfolding. Changes in secondary structure, such as the strain-induced transitions to p-strands [68], and the rupture of non-covalent interactions are especially important in solvated single-molecule studies. In these studies energetic and chemical pathways are tracked while whole or partial domains of proteins are unfolded. This method often uncovers intermediate states during the unfolding process [69]. 2.1.4 Molecule Shape Calculation A common way to measure the shape of a molecule is the radius of gyration. This second order tensor is calculated as the (mass or charge weighted) root mean square distance of all the atoms from the center of mass of a molecule. It is used here to quantify the size of the ensemble of atoms in the molecular structure. The volume of the equivalent ellipsoid described by the gyration tensor is V = (4/3)7rAxly, where Ai is the eigenvalue of the gyration tensor associated with direction i. Usually, the tensor is calculated with respect to the principal components of the system. If one of the principal components coincides with the axis direction then the associated eigenvalue is the molecule size in that direction. By comparing the change of radius for the same structure in two different environmental conditions one can calculate the shape and volume change of the structure. Alternatively, the average end-to-end distance of the chains gives an estimate of the molecule size in axial direction. 34 2.1.5 Replica Exchange Molecular Dynamics Replica exchange molecular dynamics (REMD) is a parallel tempering method for improved conformational search introduced by Sugita et al. in 1999 [70]. Instead of performing a random walk in the multicanonical space, the system performs a random walk in the temperature space. This is achieved by simulating NREMD replicas of the same system at a range of initial temperatures (usually 300K - 600K) and then performing Monte Carlo like acceptance checks to exchange temperatures. This allows the system to escape from local energy minima. Usually, the analysis is focused on the trajectory of the sample at 300 K (most force fields are trained for this temperature) and the trajectories of the higher temperatures are neglected after the simulation. They serve only to overcome folding energy barriers. The optimum number of replicas (in a canonical ensemble) is dependent on temperature range [Tmmi, Tmax] and system size (number of atoms N) [71] NREMD,Opt = 1 + 1.789 NV In 2.9 (rx Usually, REMD is followed by an explicitly solvated equilibrium simulation and can thus be run in implicit solvent. This reduces simulation time and the number of required replica significantly. The output of REMD ois The analysis of the clusters is performed using the MMTSB Toolset [72]. The REMD protocol is a standard tool of most molecular dynamics software packages, such as GROMACS and NAMD [65]. 2.1.6 Entropy calculation It is possible to estimate the total as well as relative entropy (or entropy change) from the trajectory of a molecular dynamics simulation. The two most common approaches are the quasi-harmonic method and Schlitter's method [73, 74]. Both 35 rely on the determination of the covariance tensor C of atomic fluctuations (similar to a normal mode analysis) [75]. The components matrix of C is singular in most cases. Multiplying C with the system mass tensor M yields a symmetric semidefinite tensor that can be diagonalized. The quasiharmonic method approximates the fluctuations using a Gaussian probability distribution. First, the eigenfrequencies of the system are determined by solving the eigenvalue problem det M1/2 CM1 2 - 2.10 0. kBT These frequencies are then used to determine the entropy by calculating it as a sum of harmonic oscillators SQH = R 1 hwi/kBT In(hwi/kBT-1 - e(hwi/kBT) -_n(1 2.11 Schlitter showed that an estimate of the maximum system entropy is given by Ssch= 0.5 R ln det 1+ ( 2 )M1/2 C M1/2]. 2.12 Generally, Ssch > SQH. Both quantities can be determined by finding and diagonalizing the covariance matrix of the system fluctuations (after least square fitting the trajectory to the average structure to remove rigid motion of the molecule). For N atoms, the covariance matrix has the shape 3N x 3N and hence 3N eigenvalues and eigenvectors. The eigenvalues are decaying fast and for the normal mode analysis only a fraction of them are contributing to the entropy. This observation can speed up computation significantly when diagonalizing the 36 mass weighted covariance matrix and evaluating the entropy contributions of each mode in equation 2.11. 2.1.7 Free Energy of Solvation The free energy of solvation is the energy released or input when putting a solute in solvent, e.g. silk in water. It can be computationally estimated by switching off the interaction of the solute with the solvent and leaving only intra-molecular interactions. The problem is that the free energy difference has to be calculated from Monte Carlo probabilities of the transition from the solvated state to the unsolvated state. In MD such events are too rare to be sampled within a reasonable runtime. In order to estimate the free energy of solvation one can use the so called A-method using the Bennett Acceptance Ratio (BAR) [76]. The idea is to slowly decouple the solute from the solvent, using a parameter A e [0,1]. The difference in free energy of each A-step is calculated by the Monte Carlo transition probability P between state A and state B PA-B = ex 2.13 B-EA where E = K + U is the total system energy. If the increments in A are small, the transition energies are also small enough such that some of the phase space is shared between the 'adjacent' simulations. Therefore, one can calculate the Monte Carlo probabilities and the free energy difference between each state. For expediency reasons it is common to decouple Coulombic and van-der-Waals interactions simultaneously. In GROMACS, soft-core interactions prevent the overlap of charges (meaning, although solute and solvent do not interact, they cannot occupy the same space in the simulation box) without disturbing the freeenergy balance. Minimum increments of AA = 0.05 and a simulation time of 0.5 ns with a timestep of 0.5 fs and no holonomic constraints must be used. The cutoff scheme has to be changed to group cut-off, since GROMACS does not 37 support free energy calculations with the Verlet cut-off scheme. For very large molecules with multiple chains, the solvent-solute decoupling does not work. In this case, the free energy of solvation can be estimated directly from the solvent accessible surface area of the molecule's residues [76, 77]. The surface accessible area is calculated using the double cubic lattice method [78], and the relation to the free energy of solvation is (18 2 cal/mol/A 2 ) [77]. 2.1.8 Simulated Raman and Infrared Spectrum The following introduction to dipoles, dipole moments, and polarization is inspired by the excellent explanations in Reference [79]. In molecular dynamics force fields, each atom is assigned a partial charge, reflecting the unequal distribution of electrons within the molecule [80]. The phenomenon that strong electronegative atoms pull electrons from less electronegative atoms is called polarity and the resulting charge separation forms a so called dipole. The dipole moment of a molecule is determined by the magnitudes and the distances between the dipoles. For example, peptide bonds in the backbone of a protein have large dipole moments. A molecule with a permanent dipole moment creates a surrounding electrostatic field which induces a dipole moment in other molecules located nearby. This effect is called polarization. Polarizability refers to the ease with which electrons are shifted by this external electronic field. Molecules that lack electrons are easier polarizable than molecules that have many electrons. When molecules vibrate it causes fluctuating dipoles. Hence, the dipole moment fluctuates also. This coupling lets the oscillation of nearby molecules synchronize. The strength of the coupled fluctuation (dispersive interaction) is related to the polarizability of the two molecules or atoms. The vibrations of dipoles are experimentally directly determined by infrared (IR) spectroscopy. In a material, certain frequency ranges are absorbed when shining electromagnetic 38 waves on it. Absorbing infrared laser light leads to the oscillation of molecule bonds and are detectable in the reflected (not absorbed) spectrum. Each absorbed frequency is characteristic for a certain type of bond, such that infrared spectroscopy (IR) can detect material structures. A molecule is IR active only when the molecule has a changeable or inducible dipole moment. A molecule is IR inactive when the vibrations are symmetric to the center of the dipole. IR spectra range from 4000 cm~ 1 to 400 cm-'. Each molecule has a typical pattern in the spectrum, but at 1500 cm~' and below the origin of the spectrum cannot be determined any more (fingerprint region). For solvated spectroscopy systems, IR spectroscopy offers the possibility to cannot be investigate used. Here, Raman the crystallinity and composition of the material. This is especially important for biomaterials. Raman spectroscopy requires the molecule to exhibit changes in polarizability during its oscillations. A laser is shined at the material and the dispersed spectrum contains additional frequencies. This is due to the material's interaction with the light by transferring energy from or to the light. The additional lines in the spectrum are called Stokes-Lines, and the shift in the spectrum is called Raman spectrum. The quantity measured is therefore the fluctuation of the polarizability tensor. Since the selection criteria for IR and Raman spectroscopy are different, they work as complementary methods to characterize materials. The IR spectrum can be found directly from molecular dynamics simulations. Given the distribution of partial charges and the trajectory of all atoms, the vibrations of the dipoles can be determined. The Fourier transform of those vibrations generates the trajectory. This methodology is implemented as a subroutine in VMD [81]. The condition for accurate results is a high output frequency of atomic coordinates (to capture high frequency vibrations) and an accurate partial charge distribution. 39 The Raman spectrum cannot directly be found from classical molecular dynamics trajectories, since the partial charges are static on the atoms and polarizability is not modeled. Certain core-shell models take the polarizability into account and would allow the calculation of the Raman spectrum directly. For large protein molecules, like silk, the CHARMM force field is preferred which has fixed partial charges. An alternative route to the Raman spectrum is via the vibrational density of states (VDOS). The VDOS is also called phonon density of states or power spectrum. Under the Born-Oppenheimer assumption, the fluctuations of the polarizability depend linearly on the atomic displacements [82]. The VDOS is the Fourier transform of all molecule vibrations and its measured intensities g (w) are related to the Raman intensity IRaman(() gR( )[n(o ) ) (0 'Raman(w) by 2.14 + where gR (W) is the convolution of the VDOS with the light-to-excitation coupling factor C(w): g R (_') = f C(w C(w)g(w). - 2.15 C(w) is a priori unknown and can be determined with an independent probe. n(w) is the population of the vibrational mode at frequency w n(w) = (ehw/kBT - 2.16 )1 . 40 The intensities of the VDOS g (w) are found through the Fourier transform of the velocity autocorrelation function [83] g()= *(v(t _"o, i2wt 0 )v(t)) (v(t0 )v(t0 )) 2.17 dt. 2.1.9 Data Analysis Molecular dynamics simulations can produce large quantities of data when storing atom positions, velocities, energies, forces, stresses (pressure), and system temperature. The output frequency depends on the analysis type. For example, an equilibrium simulation requires a less frequent output than the calculation of vibrational spectra. While there is a range of data analysis toolsets available, such as GROMACS scripts, MMTSB [72], and VMD [81], most analyses have to be performed by scripts specific to the task. An efficient programming language for processing large data sets is Python. In this thesis, most analyses have been performed using self-coded Python and MATLAB scripts. Visualization of molecules is performed with VMD and data plotting with MATLAB and Origin. 2.2 Analytical approaches to materials failure/ fracture 2.2.1 Linear elastic fracture mechanics One of the main interests in materials science is the study of flaw influences on materials behavior and properties. The quantification of crack propagation is based on Griffith's and Irwin's work from the early 19th century [84]. It assumed that the energy needed to propagate a crack, the critical strain energy release rate 41 Gc, equals the surface energy of the two newly created surfaces along the crack path = Gc = _c-2 = 2ys ,2.18 9a E where Kc is the fracture toughness (mode III, III or mixed mode, a material constant), E' the equivalent Young's modulus, and ys can be determined from surface chemistry. Kc can be determined experimentally or analytically for specific boundary conditions [85]. In linear elastic fracture mechanics analyses the stress intensity K scales with aoV\F(), where uo is the stress applied on the specimen and F((a,Geometry)) a shape and boundary conditions dependent function. a is the crack length. Linear elastic fracture mechanics predicts that the stress singularity at the crack tip scales with 1/,fr and thus the yield (also known as process/plastic/damage/decohesion) zone 10 is given by 10 = F( ) 2 a 2.19 . For the classical Griffith crack problem this simplifies to ) 1= a 2.2.2 Polymer fracture mechanics 2.20 The fracture analysis of polymers is commonly addressed from two points of view: a statistical, micromechanical (e.g., using Bell theory or atomic potentials) [86, 87] or a continuum mechanical (e.g., using phase field theory or linear/nonlinear fracture mechanics based on Griffith's work) [88-91]. Both 42 approaches are well explored, but are yet to be unified in a comprehensive framework. In the nonlinear case including plasticity and irreversible damage before fracture, the Griffith equation can be rewritten as [92] Gc = Ys + f 1 EijklEiiEkl dy, 2.21 where oui is the stress tensor, Eij the strain tensor, Eijkl the elastic tensor, and 10 the yield zone size in direction y. Polymer and rubber materials are often modelled as homogeneous, nonlinear elastic materials, where the strain energy release rate is given by the J-integral [93] W nk - ti j = GC= where W = ui uijdEj = 2.22 dr E' EijklEidcij = EijklEijEkl is the internal strain energy density, t = n - a the surface traction und u the displacement. In this convention, the crack is oriented in the x1 direction. To account for heterogeneity, Eshelby's expression can be applied [94-96]. For heterogeneous, nonlinear elastic materials the strain energy release rate is given by [97] =FWnk -t where (Wk)expl )GcdI- (Wk)xdfl 2.23 is the derivative of the strain energy function in the nonhomegenous material. 43 For the elastic solution, the size of the decohesion or process zone lo is given by the Dugdale-Barenblatt yield-strip model, see Figure 8, 7r Gc 8 cr2 10 = -E 2.24 where Gc is the critical energy release rate, a material parameter that quantifies the amount of energy needed to drive a crack through a surface, E the elastic modulus of the bonds, and Ut the theoretical strength of a chemical bond [98, 99]. This solution is often applied to the fracture of ductile materials. a . .............. J-dominated zone Figure 8 1 Stress variation in the cohesive zone according to the Dugdale-Barenblatt model. Adapted from Reference [100]. In view of Figure 8, for a nonlinear material, Gc can be generalized to [92]. Gc= 2.25 oa(r)dr, where ro is the equilibrium lattice spacing of a cubic lattice on which the molecules are arranged. 44 3 Creation of Heterogeneity - Silk Assembly The research and review presented in this chapter will be published in: * T. Giesa, C.C. Perry, M. J. Buehler, Secondary Structure Transition and Critical Stress during the Assembly of Spider Silk Fibers. In submission, 2015. In this chapter, the following questions regarding silk assembly are answered: " How is the heterogeneity of silk fibers created on the smallest scale? " What is the importance of shear in the assembly process of silk fibers? " What is the smallest molecule size that might give rise to a 'silk-like' structure? " What shear stresses are required for the secondary structure transition from helical to sheet state? " Which components/ parts of the silk sequence are crucial for the secondary structure transition? " How can these results be used towards the design of artificial spinning devices? Research strategy: A model of the silk spidroin (solubilized silk before spinning) in equilibrium is developed and the assembly process of Nephila Clavipes silk sequences is computationally investigated. Using replica exchange molecular dynamics for a total of 1 microsecond, the equilibrium structure of the spidroin is determined. 45 Using steered molecular dynamics at natural pulling speeds a shear flow is modelled and the secondary structure transitions as well as shear stresses in the silk protein chains are determined. Intra-chain interactions are studied for structures having two, four or six poly-Alanine repeats. Additional inter-chain interactions are studied for structures built from structures having four polyAlanine regions in a parallel and antiparallel configuration and compared to the effect of a single molecule containing the same number of poly-Alanine regions. After shearing, the assembled chains of different lengths are equilibrated to test the stability. 3.1 Background There is little doubt about the importance of shear stress facilitating secondary structure transitions during the assembly process [101]. In fact, it has been observed that without shear stress, silk fibers do not form, but maintain globular shapes [102]. Surprisingly, given the current standing of the field, no one has performed a detailed investigation of shear stress in the assembly mechanism. The soluble spinning dopes stored in the glands of the spider is in a form that enables spinning when the need arises. The material present in the spun fibers is however, fundamentally different from the dope. While first attempts to synthesize silk fibers have been successful, and industry-scale production of recombinant silk has been made possible, the mechanical properties of the spun fibers either do not currently compare with natural fibers or the assembly was performed under non-ambient conditions [103, 104]. Furthermore, many synthesized fibers contain a substantial amount of regenerated silk, i.e. the spinning dope originates from natural spider silk [104]. Using genetically giLneered~ seqfILunc prtiLLs a way U ItUUc 11i1iatinLs1L!L UIL 101H suFpLy I11Im spiders, but yields weaker and more brittle fibers [105]. The inferior quality of 46 the engineered fibers is mainly related to the selection of assembly process parameters in industrial spinning devices, different rheological behavior and molecular weights [105]. Even though much is understood about the molecular structure of silk worm and spider silks, there is still ongoing debate about the molecular structure [106]. 3.1.1 Fiber formation- spinning from solution Even though much is understood about the molecular structure of silk worm and spider silks, there is still ongoing debate about the molecular structure [106]. The spidroin (Figure 9 left) is water soluble while stored in the spider's abdomen, but under certain conditions assembles into an insoluble fiber. The dragline silk spidroin consists mainly of two proteins, major ampullate spidroin 1 (MaSpi, studied here) and major ampullate spidroin 1 (MaSp2) [107]. Spider dragline silk after spinning could be described as comprising crystalline P-sheet crystals embedded in an amorphous or poorly structured matrix (Figure 9 right). The sequence is highly repetitive, where the number of repetitions a is in the order of 100's [35]. Recent studies using Raman microspectrophotometry have provided information on all the distinct spidroins present in the glands of Nephila Clavipes, an orb weaving spider that can be found in the south-east of the United States [108, 109]. These studies have shown that there are distinct types of spidroins, either totally disordered or globular-like and natively folded with a high content of a-helices [108]. Major Ampullate (MaSp dragline) and Minor Ampullate spinning dopes were suggested to have an intermediate structure that contains some helices. The presence of a-helices and left handed 3 10 -helices were both identified in MaSP silks, while none of the spinning dopes were found to possess silk in P-sheet conformation [109]. Interestingly, analysis of the silk sequences using prediction software optimized for unfolded segments 47 (PONDR) suggested that all sequences should be unfolded [108]. Likewise, secondary structure predictions using the Porter algorithm predicted the structure of MaSP silk to comprise loops and turns with helix being predicted for the (A)n region [108]. The MaSP sequences were also shown to have no natural propensity for aggregation to give -amyloid structures, clearly of importance to the use of silks in biomedical applications. Spinning Duct Spinning Dope FibrillFiber -water - + Ions Shear Figure 9 Molecular model of the silk fiber assembly process. Silk is processed from spidroin to a solid fiber in the spinning duct under ambient conditions. The dragline silk spidroin consists mainly of MaSpl (studied here) and MaSp2. The sequence is highly repetitive, a ~ 0(100). Shear stresses at the wall together with the removal of water from the protein lead to the formation of a nano-composite having an aligned, #-sheet rich crystalline phase. A change of ionic conditions during the spinning process is believed to lead to a conformational change in the terminal regions of the silk protein. The process for silk assembly and spinning into fibers is complex and there are a variety of theories for the work-flow of the natural assembly process [110]. Shear stresses at the wall together with the removal of water from the protein lead to the formation of a nano-composite having an aligned P-sheet rich crystalline phase. Spinning conditions including humidity, temperature, reeling speed and ionic conditions are believed to control the assembly of silks. For that to be possible, the protein needs to be protected from premature assembly which would clog the spider's spinneret or the spinning device. Premature self48 assembly poses one of the issues involved with the synthetic production of silk fibers. The importance of laminar flow in the spinning duct has been emphasized with two theories being proposed by which the silk spidroin transitions to the fiber form [111]. Firstly, the micelle theory states that in the spider's abdomen the silk dope is stored in micelle form, where the hydrophobic terminal regions of the sequence shield the interior a-helical structure from water [112]. This approach assumes a low concentration of the spinning dope. During the spinning process, the applied shear induces the formation of P-sheets, while a change in pKA and ionic conditions mainly affects the structure of the terminal region [113]. The Cterminal is believed to shield the protein from premature assembly [108]. Due to the pH-sensitivity of the terminal, the conformation changes in the spinning duct, where a pH drop is induced through ion pumps [112]. Simulation studies have confirmed that the terminal domains are critical in the initiation of multimerization [114]. Although their sequence is highly preserved between species, the N-terminal region is not present in all MaSpi silks [107]. Secondly, the liquid crystal theory for spider and silkworm silk states that silk proteins are stored in a nematic liquid-crystalline phase [41]. During assembly in the spinning duct, the proteins organize in bilayered disks that elongate along the fiber axis under shear flow. Under these conditions, random-coil and polyproline-II helixlike conformations transition to P-sheet-rich structures. This approach assumes a rather high concentration of the spinning dope (-50%) [41]. Experimental studies to mimic the spinning process using engineered and recombinantly produced spider dragline silk proteins from Araneus Diadematus in microfluidic devices have confirmed that shear flow is essential for fiber formation [115]. This experimental study generated results supporting the micelle hypothesis of Jin and Kaplan [113]. The experiments were performed at low-mid range concentrations and did not require liquid crystalline behavior for 49 fiber formation. Rather, high elongational flow rates of about 1000 s-1 in an in vitro microfluidic device were essential, though the shear forces were not measured. The high shear rates used were found to lead to significant increases in viscosity and this was used as a marker for fiber formation. Also, an increase in orientation of the crystallites in the fiber, shear thinning and an increase in force along the spinning duct has been observed [41]. Higher draw rates also lead to smaller crystals which increase stiffness and fracture resistance [49, 1001. 3.1.2 Silk structure after spinning Once the MaSP silk has been spun there is experimental evidence that all residual native-like structures are absent [116]. There is considerable information about silk materials as formed by spiders (and silk worms) where numerous experimental techniques including NMR (solution and solid state), X-ray crystallography and Infrared and Raman spectroscopy have been used to study the fibrous structures. As might be expected, there is more known about the crystalline regions than the disordered regions owing to a paucity of methods to study amorphous or poorly ordered materials. As examples of what is known for N. Clavipes drag-line silk, X-ray diffraction studies have shown that the fundamental crystalline unit has an orthogonal unit cell with the average size of crystallites calculated to be ca. 2 x 5 x 7 nm [45] and the inter-crystalline length along the fiber axis measured as 13 - 18nm [56]. The crystalline phase arises from the organization of poly-Alanine segments of the silk protein sequence into Psheet structures although the precise P-sheet content reported varies according to the measurement technique used. Analysis by Raman microspectroscopy suggests levels of 30 - 40%, [117, 118], as opposed to about 20% identified by XRD [119]. Computer simulations of the crystalline components of silk suggest that the poly-Alanine sections line up anti-parallel in the H-bonding direction 50 with parallel stacking in the side-chain direction. This arrangement also leads to stable P-sheets [48, 120]. The nanoconfined crystals are responsible for the high strength of silk materials, while the amorphous part contributes to the elasticity of the material [44, 47, 121]. The non-crystalline regions are alternately described as amorphous, poorly oriented or randomly coiled. They are not very well understood, although analysis of silk fibers by NMR has provided significant detail on the likely structures present in the less ordered domains [122, 123]. The presence of 3 1ohelical structures with GLGXQG motifs forming beta turns was proposed in the late 90's [122, 124]. Later studies using 13 C labeling of materials and multipulse NMR to measure backbone dihedral angles of Alanine and Glycine within silks suggest that these amino acids are 'ordered' on the NMR timescale of microseconds - with poly-Alanine and GGX motifs probably located in microcrystalline domains and the remainder of the silk in a 3 10 -helix arrangement [123]. Other structures including P-turn, (-spiral and helices have been suggested with different views as to the structure adopted by GGX; both less ordered helices/ distorted P-sheets have been proposed [109, 118, 123]. More recent studies have suggested that there are actually three distinct structure types in silk: crystalline P-sheet, weakly ordered (-sheet, and amorphous or disordered structures. The presence of the weakly ordered phase has been suggested based on H/D exchange and NMR/ IR analysis and is described variously as weakly oriented P-sheet [125, 126], or as oriented amorphous regions [127, 128]. Molecular dynamics simulation studies of model silk constructs built from bundles of silk fragments has shown that the density of Hbonds is lower in amorphous regions as compared to regions containing P-sheet crystallites [48]. 51 The observation that the primary sequence largely dictates which specific amino acids and amino acid motifs are associated with particular structures is particularly interesting in light of a long standing hypothesis from the Lewis group which states that there are no redundant components of the spider silk sequence [129]. They state that from a protein biochemistry standpoint, the highly repetitive nature of the proteins makes it highly unlikely for any random structure to be generated. 3.1.3 Simulation Setup 3.1.3.1 MaSpi Silk Model The wildtype Nephila Clavipes silk MaSP1 sequence: AAAAAAGGAGQGGYGGLGSQGAGRGGLGGQGAG (accession number UniProt KB P19837) is used to study the assembly process. Structures containing the 27 amino acids (Gly-rich semi-amorphous repeat, then the poly-Alanine region followed by two, four, or six multiples of the whole sequence to give the requisite number of poly-Alanine regions surrounded by 'amorphous/ less well ordered' material) are used as models, see Figure 9. The structures differ from those in the studies by Keten and Buehler [48] and Bratzel and Buehler [130] where bundles of polypeptide chains, each comprising a single poly-Alanine region with an amorphous region at each end were studied by REMD to replicate the P-sheet crystal dimensions and explore the length of the poly-Alanine region to obtain maximum theoretical strength and toughness. Multiple copies of the complete silk MaSP1 sequence are present within an individual structure. Structures are initially built using the Biopolymer module in Tripos Sybyl. Structures in random conformation with neutral ends are minimized in vacuum before being exported for replica exchange calculation (REMD) [70]. 52 3.1.3.2 Replica Exchange MolecularDynamics The REMD method is used to identify molecular structures. 48 replicas (see equation 2.9) of strands with initial configuration and length schematically shown in Figure 10a are simulated at temperatures between 300 and 800 K to ensure coverage of conformational space. The REMD simulations are carried out using the CHARMM force field [65] implemented in NAMD [131], using the Generalized Born implicit solvent model (a-cutoff 12.0, ion concentration 0.3). Solvent friction is added via a Langevin friction term (10 ps'). A cutoff of 15 A is used for long range interactions and a timestep of 2 fs. REMD is run for 20 ns /sample (total of 960 ns) and the temperatures exchanged every 0.2 ps. Statistical analysis is performed using the MMTSB toolbox, [72] with a k-means clustering algorithm (cluster distance 2 A). Representative final ensemble structures are chosen from the lowest temperature replica (300 K). The higher temperatures are used for fast conformational search and overcoming kinetic trapping in the REMD scheme. The simulations are run at Harvard's Odissey cluster with 64-256 cores. a REMD + MD a= 3 a =1 - ips b a =5 SMD - 1Ons (0.5 m/s) Inter-chain Intra-chain C MD 50ns Single-chain Layer GGQGGAGQGGYGGLGSQGAGRGGLGGQ (GAGAAAAAAGGAGQGGYGGLGSQGAGRGGL)a Figure 10 1 (a) Model of the silk spidroin in equilibrium. Using Replica Exchange Molecular Dynamics for a total of 1 microsecond the equilibrium structure of the protein is determined for different chain lengths (a = 1,3,5), here denoted as poly-Alanine regions (blue) and remainder of the intervening sequence (red). (b) Model of the silk assembly process. Using Steered Molecular Dynamics at natural pulling speeds a shear flow is modeled and the secondary structure transitions as well as shear stresses in the silk protein chains are determined. Intra- as well as inter-chain interactions are investigated. (c) Model of the final fiber structure. After shearing the assembled structure is simulated in solvent and vacuum to test the stability of the structure as a single-chain or as layered structure, i.e. multiple stacked sheets after shear. 53 The center of the largest cluster, meaning the structure with minimum RMSD (root mean squared deviation), is exported for analysis in explicit solvent. To obtain more realistic molecular conformation and secondary and tertiary protein structure, the structure is equilibrated for 20 ns using GROMACS (CHARMM27 forcefield) in a charge-neutralizing periodic waterbox of TIP3P containing 150 mM sodium chloride. To prevent image interactions, the waterbox pads the protein by at least 10 A. Equilibration is performed with Langevin dynamics in an isothermal-isobaric ensemble (NPT: 300K, ibar) using the Nose-Hoover thermostat and the Parrinello-Rahman barostat. Particle Mesh Ewald (PME) electrostatics with 1.4 nm cut-off is used to more accurately capture solvent interactions [132]. It is possible to use a 2 fs timestep because of holonomic constraints (LINCS algorithm). The secondary structure of each structure is determined using DSSP toolset and the STRIDE algorithm in VMD [81, 133]. 3.1.3.3 Shear Flow and Steered MolecularDynamics In order to simulate shear boundary conditions, SMD simulations are performed, see chapter 2.1.3 [69]. Soft springs (stiffness k) are fixed to the symmetry points of the equilibrated structures, as shown in Figure 11. The springs are set to provide stiffness only in the pulling direction so that the structure is able to relax in all other directions and hence no unnecessary constraints are enforced. An isothermal ensemble in an explicitly solvated periodic waterbox is set up. Isobaric constraints are enforced in the two directions that are perpendicular to the pulling direction. The box size in pulling direction is constant and periodic, but significantly larger than the molecule. A set of pulling speeds (0.5 - 5 m/s) and spring stiffness's (100-2000 kcal/mol/A ) is used to probe the robustness of the results. 54 (I) (ii) De Figure 11 | Two different boundary conditions (i) and (ii), both shear, are tested to investigate the trajectory of secondary structure and shear stress. The part of the sequence colored in blue is the Alanine rich region and the part of the sequence colored in red is the Glycine-rich region. The pulling force is determined with the steered molecular dynamics (SMD) algorithm. Two shear boundary conditions, Figure 11 (i) and (ii), are tested to investigate the trajectory of secondary structure and shear stress. The part of the sequence colored in red is the Glycine-rich region and the part of the sequence colored in blue is the Alanine-rich region. The symmetry positions of the sequence are held by soft springs that acted only in the pulling direction to ensure that the transition states could be captured during the simulation. The 'x' in Figure 11 (ii) marks a fix at the alpha-carbon of the terminal amino acid. The pulling force is determined and the diameter calculated at all points of the trajectory by the inplane radius of gyration of the structure and determining an equivalent molecule diameter Deq. Figure 12a shows the trajectory of secondary structure of the equilibrated structures during the pulling experiment. The graph shown is one out of four tests for a = 3 (with boundary condition (i)). All following graphs show averages over the results of the two shear boundary conditions. The pulling speed is set to 0.5 m/s, well within the range of experimental drawing speeds, leading to a simulation time of 10 ns. Higher pulling speeds have no significant influence on the results (see Figure 13). This is in agreement with NMR experimental results that state that the reeling speed has little effect on the secondary structure of the fiber [134]. Similarly, the spring constant used in the SMD experiment has little influence on the results. The average f-sheet content for each structure is 55 determined by averaging the P-sheet content for 1 ns around the maximum content as indicated in Figure 12a. The transition shear stress, an example shown in Figure 12b, is averaged in the same region as that taken to assess the P-sheet content. In this case, after approximately 8 ns the stress exceeds the strength of the material and the simulation results are valid only before this point. In all cases, the structure transition happens prior to reaching this limit. There is most likely an influence of the specific pulling residue positions on the behavior of the molecule. It is also highly likely that the specific residue involved in particular structures may be dependent on how the simulation is prepared. However, due to variety of different chain lengths and simulation setups (all simulation are repeated with different initial conditions) with different boundary conditions and long equilibration times, general insights into the transition mechanism can still be generated. b a 100 2500 max. P-sheet -a-Helix -P-sheet4 -Feq 2000 80 -Co _TUM rDeq 1500 60~ 500 200 0 2 6 4 Time (ns 8 0 10 2 6 4 Time (ns) 8 10 Figure 12 1 Secondary structure transition and shear stress during the assembly of silk. (a) Secondary structure trajectory during the pulling simulation in explicit solvent. The graph shown is one out of four tests for a = 3 (with boundary condition (i)). Starting from the spidroin, a high a-helical and coil content and no P-sheets can be found. During the shear induced assembly all helices and other structures transition into P-sheets or are destroyed in agreement with experimental observations of the processes in the spinning duct. The P-sheet content after pulling for each structure is determined by averaging the P-sheet content for 1 ns around the maximum content as indicated in the figure. (b) Shear stress associated with the secondary structure transition. The transition shear stress is averaged in the same region as that taken to assess the Psheet content. In all cases, the structure transition happens prior to reaching the strength limit of the material. 56 3.1.3.4 Post Shear Equilibration and Stability After exposing the structures to shear, a representative structure with a high amount of P-sheet (in the domain, that has not been pulled) is further equilibrated for 50 ns both in vacuum and in explicit (TiP3P) water containing salt at 150 mM to measure the stability of the secondary structures induced by shearing. In vacuum, an isothermal-isochoric ensemble is used with cut-off for Coulombic and van-der-Waals interactions (3 nm), velocity-rescale thermostat and dispersion correction (accounting for the cut-off van-der-Waals scheme). The secondary structure is tracked using the DSSP algorithm. In addition to the different length spidroins, stacked sandwich structures are analyzed for stability. 2-layer and 3-layer sandwiches are formed in parallel and anti-parallel stacking from chains with a = 3 taken from the result of the SMD simulation (within the barred area in Figure 12). These sandwich structures are then equilibrated in water with 150mM NaCl for 50 ns. 3.1.3.5 Transition Probabilities In order to obtain a better understanding of the secondary structure transitions from helix/turn to sheet, the probability of being either in a helical/turn (spidroin) or in a sheet structure (after pulling and after further equilibration) is calculated residue by residue. The probability for the initial equilibration, - defined as the percentage of frames each residue is part of a a-helix or turn/310 helix divided by the total number of frames, is calculated from the last 10 ns (out of 30 ns) of the equilibration trajectory. Similarly, during the pulling simulation the probability is calculated from the 1 ns that also defined the P-sheet content, see above. The probability from the equilibration after pulling is calculated from the last 10 ns (out of 50 ns) of the simulation. The joint probability for each residue is calculated as the product of the three probabilities. This allows us to track exactly which amino acid residues and which geometric relations are key to 57 the transition. The relative location of the residues appearing in all three structures is compared using the Sybyl Tripos Biopolymer module and VMD. 3.2 Size Dependence of Structural Transition The spider silk dope is stored in the abdomen of the spider in globular form to prevent premature assembly and clogging of the spinning duct. In the simulation, the general form of the simulated protein structure is largely independent of the size of the molecule (left hand column, Table 1). The spidroin remains unassembled and predominantly contains helical shapes (a-helices, turns and 3 10 -helices) as observed in the deconvolution of Raman spectra [109, 135]. Similar evidence has been found for S. Cynthia Ricini silk, where the polyAlanine regions form a helical structure when dried as a film but P-sheet in fibers [136]. The helices towards the beginning of the sequence are arranged at an angle to one another typically of 60 degrees. 58 is stored in the Table 1 1 Model structures of the stages during the assembly process. The spider silk dope peptide/ simulated the of abdomen of the spider in globular form. In the simulation, the general form remains It column). hand (left molecule the of protein structure is largely independent of the size a column) (second process shearing the During turns. and helices unassembled and predominantly contains in structure the of equilibration Subsequently, structures. 1-sheet into transitions structure large part of the column vacuum or explicit water in the presence of ions, the silk relaxes again (third and fourth In fl-sheets. their retain and stable are respectively). In vacuum, irrespective of simulation size, all structures original their to return 3) = a 1, = a stretches; poly-Alanine four water, the smaller silk structures (two and a = 5) largely disordered 'spidroin-like' state and only the larger structure (six poly-Alanine stretches; and hydrated fully between intermediate an be to assumed be can condition natural The remains stable. vacuum due to the removal of water and ions from the silk as it is spun into air. + 1. Equilibration Replica Exchange MD Water + Ions Ius 2. Shear Flow Steered Molecular Dynamics 1Ons 3a. Equilibration MD Vacuum 50ns 3b. Equilibration MD Water + Ions 50ns ~4-~- 1 Loop (a 3 Loops 5 Loops ~y f~i (a = 5) During the shear induced assembly (second column, Table 1) all helices and other structures transition into P-sheets or are destroyed, i.e., they transition into random coil, in agreement with experimental observations of the processes in the spinning duct [135]. Figure 12b and Figure 12c show the simulation results from the shear simulation for a chain with a = 3. Secondary structure transitions and associated shear stress results for other chain lengths can be found in the Appendix, section 7.1. The amount of P-sheet is insensitive to the pulling speed, Figure 13. This is in contrast to the results of experimental measurements which have shown that the faster silk is reeled, the higher the P-sheet content. In addition, at higher reeling speeds, smaller crystals are formed which positively 59 affects the toughness and ultimate strength of the fibers [56, 137]. Subsequently, after equilibration of the structure in vacuum or explicit water in the presence of ions, the silk relaxes again (third and fourth column respectively, Table 1). This is discussed in the next chapter. 500 __O0 Spring Stiffness [kca/mol/A 1500 1000 2500 2000 45 40 35 30 0 CO' 25 3 VA 20 15 1015 0 I 0.2 I 0.4 I 0.6 i i I 1.2 1 0.8 Pulling Speed [m s I I I 1.4 1.6 1.8 2 Figure 13 1 P-sheet content versus pulling speed for a set of spring stiffnesses. The observed P-sheet content after shear is insensitive to the simulation parameters. Figure 12a and Figure 12b provide further valuable insights into the assembly mechanism of silk structures. The initial unfolding of the spidroin solution takes place under very low shear stress. The stress drop (at -7 ns, see chapter 7.1) is a phenomenon also observed when measuring forces during in vivo silking [138]. For the first time, a link is established between the drop in stress to the point where a high P-sheet content in the structural transition is reached. This also 60 indicates that the creation of the crystal brings the molecule into a low energy state with the force providing the energy to overcome the barrier for state transition. The secondary structure transition from a-helical agglomerations to Psheet structures is ubiquitous and independent of the number of chain repeats a and also very robust with respect to the simulation setup, Figure 13. Figure 14a shows the average and maximum P-sheet content attained after the shear flow experiment (determined by the methodology described on the previous section, Figure 12a). The attained structure content is well within the range of experimental observation (20-25%, [33]; 36%, [117]). Figure 14b shows the associated transition shear stresses (determined by the methodology described in the previous section, Figure 12b). For all chain lengths, the stress is above the elastic limit of assembled silk fibers (150-300 MPa). This agrees with experimental observation that reorganization of silk requires significant shear stresses and the average force during silking is higher than the conventional yielding force [139]. The transition shear stress for silk assembly is determined to be between 300 and 700 MPa. Initial secondary structure transitions take place at lower stresses already, -20% of the breaking stress. In order to reach the highest P-sheet content, the stress needs to be increased to up to 50% of the breaking stress. Experimental observations made by measuring the in vivo silking force directly on the spun fiber, put the transition stress at 20-60% of the breaking stress (i.e., 300-850 MPa for N. Clavipes silk) [138]. Interestingly, predictions from flow simulations show that a much smaller pressure gradient is needed to push the solution through the gland [140, 141]. However, such studies neglect the need for secondary structure transition and investigate the process solely from a fluidmechanical point of view. 61 a b 50 max. 1400 #-sheet material stren-th 0ii'1200 40 21000 30 Experiment 10 ~20 0 -- Expednment I 20 __ 3 a a 5 Figure 14 1 (a) The P-sheet content attained after the shear flow experiment is independent of the chain length and well within the range of experimental observation. (b) The transition shear stress for the silk assembly is calculated to be between 300 and 700 MPa, whereas experimental observations put it between 20-60% of the breaking stress (300-850 MPa). This agrees with the observation that the reorganization of silk requires significant shear stresses. 3.3 Structural Stability of the Assembled Silk Molecules Structural stability, i.e. the stability of the equilibrium after the force-induced state transition, of the pulled protein is determined by equilibrating it for 50 nanoseconds in explicit water and also in vacuum, see third and fourth column in Table 1. The initial structure of this 'post-pulling' equilibration is a P-sheet rich structure after the transition. The structures are considered stable, if the P-sheet content does not change significantly during the simulation either in vacuum or in water. In vacuum, irrespective of the chain size a, all structures are stable and retain their P-sheets, though the amorphous ends of all of chains show some compaction during the equilibration stage. In solvent (water with ions present at 150mM), the smaller silk structures (two and four poly-Alanine stretches; a = 1, a = 3) essentially return to their original largely disordered 'spidroin-like' state and only the larger structure (six poly-Alanine stretches; a = 5) remain relatively stable, Figure 15. 62 Figure 16 summarizes the secondary structure content of the spidroin and the simulated structures after equilibration in explicit solvent (data taken from 40-50 nanoseconds of the simulation). While the shorter chains start to form 3 10-helices again and retain only a low percentage of P-sheets, the larger structure is stable and shows the secondary structure composition expected from the assembled dragline silk fibers [135]. Specifically, a low helical content (310- and a-helices) and a high P-sheet and P-turn content are expected. This suggests that the assembly mechanism proposed here may work similarly for a much larger number of repeats. Figure 17 shows the same analysis for the relaxation in vacuum, where all structures are stable. 50 -I -3 -5 40 LOOP Loops Loops 30 '20 10 20 30 40 50 Simulation Time [ns] Figure 15 1 Stability of the silk chains (P-sheet content) after shearing and a further equilibration in explicit solvent. The shorter chains (a = 1, a = 3) cannot retain the P-sheet crystal and the structure returns in its spidroin state; the larger structure a = 5 remains relatively stable. 63 45- I 35[ 30 T [ I 40- 3pidroin I Loop I Loops i Loops I 20[ [ 15 I 10 I T I 5 Y 01 L Beta-sheet Bend Turn Alpha-Helix 310 Helix Figure 16 | Stability in water. Summary of the secondary structure content of the spidroin and the simulated structures after equilibration in explicit solvent. While the shorter chains start to form 3 10 -helices and retain only a low percentage of P-sheets, the larger structure is stable and shows the secondary structure composition of final assembled dragline silk fibers. rn. I ;pidroin Loop 45 II Loops Loops 40[ 35[ I I 30- 25 T 2015- - 10 5 0 Beta-sheet Bend Turn Alpha-Helix 310 Helix Figure 17 1 Stability in vacuum. Summary of the secondary structure content of the spidroin and the simulated structures after equilibration in vacuum. All chains remain stable independent of the chain length. 64 In summary, a measure for the stress required to convert a largely disordered structure into a P-sheet structure, 300-850 MPa, has been obtained by simulation. Furthermore, the minimum size of structure that is required to stabilize the formed p-sheet structures has been determined. This minimum size for a stable structure is a = 5 for single chain and a = 3, for multi-chain. This result is supported by studies of genetically engineered silk repeat units without terminals, where the minimum number of repeats was a = 2 to form indistinguishable nanofibrils, and with higher a the P-sheet content increased and fibril growth was facilitated [142]. How this transition is effected, i.e. whether there is a rationale as to which residues undergo the transition and why structures remain, has remained one of the unanswered questions. Can the molecular details found here be taken forward into the design of other polymer materials? 3.4 Molecular Details of the Transition Mechanism To gain a more detailed understanding of the transition mechanism, the representative structures before pulling, after pulling and after relaxation in vacuum and in solvent are compared to investigate how a largely helical structure might transform into a stable P-sheet containing structure. The probability of these structures containing helical or sheet content is computed for each individual residue, see section 0, and the joint probability is plotted (blue line) for the residue to transition. Figure 18 shows the probability for the structure with a = 5 to transition from a-helix to P-sheet. Figure 19 shows the probability for the structure with a = 5 to transition from 3 10 -helix/turn to 1- sheet. The key residues of the transition are identified in Table 2. For comparison, this analysis has been performed for all structures (a = 1,3,5 as well as interchain and layered structures) and the results are compiled in the Appendix, chapter 7.2. 65 1.5 -REMD -SMVD a-helix- n-sheet - Equilibration -Joint Probability 1 .0 0.5 0 0 100 AA [#] 50 Figure 18 2Transition probabilities for a -5 from a-helix to each of the 206 residues of the structure with a = 150 200 P-sheet. The graph shows the probability for 5 to be in a-helical state after REMD, in 1-sheet state after SMD and in in 1-sheet state after Equilibration. The dark blue line is the joint probability defined as the product of the three probabilities and indicates the residues that transition from helical to sheet structure with high probability. 66 -REMD - SMD 3 10-helix/turn -> n-sheet - Equilibration -- Joint Probability 1i[ -. 0.5[ 0 AL I 50 150 100 AA [#J 200 Figure 19 1 Transition probabilities for a = 5 from turn/3 10 -helix to P-sheet. Probability for each of the 206 residues of the structure with a = 5 to be in 3 1 0-helical/tum state (after REMD) and in P-sheet state (after SMD/after Equilibration). Table 2 | Residues involved in the structural transition of silk chains with length a = 5 (206 residues). From the transition probabilities (a-helices - bold, 3 10 -helices/turns - italicized) the residue numbers are identified whose joint probability to transition is higher than 10%. These residue groups are the potential key players in the silk assembly mechanism of the core structure. I AO M mva. 1kT% I RAair vamp I A I - - I R Q.Q() I J(i1-14 I L3U I 172 I V 67 - - . . . I ......... .. .......... ........... ...... Figure 20 shows structure snapshots (a = 5) with highlighted residues identified from Table 2 of the four transition stages, depicted in Table 1. Only the sections with larger structural segments (> 4 residues) are shown. Blue represents Alanine, white represents Glycine, brown represents Glutamine and aqua-color represents Arginine. In spidroin state, Figure 20(i), the structure consists principally of a-helices (in this example, 93/206 residues), turns and a small number of residues in 3 10-helices. This 'exemplar' structure contains four long helices (10-17 amino acids in length) and nine shorter helical sections. During the shear simulation, the structure comprises three principal sections of P-sheet structure (55/206 residues, 13-16 amino acids in length) plus two smaller (four amino acids) regions as well as three extensive coiled/ open structured sections, Figure 20(ii). The P-sheet sections all align antiparallel to one another within a layer and also between layers. (III) (I) 9549 IO(v) ((IV) Figure 20 I Structure snapshots (a =5) with highlighted residues identified from Table 2 of the four transition stages (i - iv), in agreement with Table 1. Only the sections with larger structural segments (2 4 residues) are shown. Blue represents Alanine, white represents Glycine, brown represents Glutamine and aqua-color represents Arginine, compare to Table 2. Stability of the P-sheet structure (iii, iv) arises from a close proximity in space of the helices in the spidroin (i). Relaxation in solvent (iv), far from destroying the structures, leads to an apparent stabilisation of shorter strands of P-sheet that can be mostly poly-Alanine but may also arise from other parts of the sequence using a range of amino acids. 68 . .... .... .... ..... .. .. .. .. .. .. .... ............. ... .... .... In the vacuum equilibration after pulling, the structure comprises four sections of P-sheet (56/206 residues, six to seven amino acids in length) as well as four sections three to four amino acids in length and seven sections with two amino acids in length, Figure 20(iii). Additionally, the structure has a helix (six amino acids) and turn structures, particularly towards the beginning of the sequence. The structure now has two coiled/ open structured sections. The P-sheet sections all align antiparallel to one another within a layer and also between layers. In the fully solvated structure after equilibration (post-pulling) there are numerous short sections of P-sheet (48/206 residues), though with some residues in turns/ 3 10 -structures, Figure 20(iv). A helix has formed from residues towards the start of the sequence and the 'loops' between the 'pulled' regions have compressed somewhat, this leading to development of some of the additional Psheet structuring that is observed. The structure now comprises P-sheet that is largely formed by the poly-Alanine regions (although usually not the whole length of the (Ala)6 regions) and also has several extensive regions of 'new/ induced' P-sheet that forms after the relaxation involving Glycine, Alanine and Glutamine residues. There are still unfolded regions that sit on the outside of the structure. The P-sheet region is not as centered (symmetrical in terms of the surrounding amorphous structure) as in the vacuum relaxed structures though the P-sheet sections still all align anti-parallel to one another within a layer and also between layers. Some H-bonding can be observed between poly-Alanine Psheet regions for those sections that lie virtually parallel to one another. Note, whereas Tyrosine can be found in either helical or turn structures, it shows no propensity for involvement in the generation of P-sheet under conditions where the resultant structure is able to relax as in 'post-equilibration' in solvent case. This seems to be the sole domain of the Alanine, Glycine, Leucine and Glutamine with only a single Arginine involved. 69 Considering the probability data, Figure 18/Figure 19, in conjunction with the specific amino acids involved, it can be concluded that the structures that are constantly as either 'helices' or 'p-sheet' in the initial equilibrated ('post-REMD') structures, the 'post-pulled' structures and the structures equilibrated in solvent (joint probability, blue line) largely arise from the more compact part of the initial structure. The sequences (four amino acids or longer) represented in the probability data, i.e. parts of the sequence that are most likely to transition, are the second half of the first large helix, the second half of the second large helix and the first half of the third large helix. The additional 'smaller common' segments are either close in sequence position to the longer helices/ P-sheet sections or are physically close in space when the 3D structure of the protein is considered. An exception to this is Arginine, which seems to sit 'between' several helices in the initial structure but is then close in space to the P-sheet structured elements after pulling and relaxation. This is the point in the sequence from which regions of P-sheet develop in the relaxed solvated structure that were not present in the post-pulled structure. In conclusion, stability of the P-sheet structure seems to arise from a close proximity in space of the helices in the first place. Interestingly, relaxation in solvent, far from destroying the structures completely leads to an apparent stabilisation of shorter strands of P-sheet that can be mostly poly-Alanine but may also arise from other parts of the sequence using a range of amino acids. 3.5 Layered Structures To understand the formation of larger crystals in silk, composite structures with two or three P-sheet layers (see Figure 10c and description in section 0), are LrtdLU Uy staLcing two or three chains of the a = 3 sLrULLUr. Fig6 U sMYWs U M P-sheet content of the layered structures. 10-15% of the structure stabilizes as P- 70 sheets in solvent, 25-30% in vacuum, almost independent of the number of layers or the orientation. The transition probabilities and associated residues are presented in the appendix, chapter 7.2. Again, predominantly poly-Alanine regions form the stable crystals. The anti-parallel stacked crystal forms hydrogen bonds within the sheet and also in between sheets. The parallel oriented layers ('2 layer parallel') form in-plane anti-parallel sheets, but initially out of plane parallel sheets that are not connected by hydrogen bonds. This suggests that although the structure is stable, it would not transmit significant shear stresses under loading. Therefore, reorganization to anti-parallel P-sheets is required. 4U I I Water Vacuum 35 0 4.' 25- 0) 4-' C 0 0 200) 0) U) 15 10 5 0 InterChain 2Layer piradel 2Layer 3 Layer Figure 21 | P-sheet content of sandwich structures after equilibration in vacuum and solvent. Layered 10-15% structures are formed from chains with a = 3. Independent of the amount of layers or the orientation vacuum. in 20-30% and solvent, in of the structure stabilize as P-sheets 71 3.6 Conclusion The shear induced secondary structure transition process of the core sequence during the spider dragline silk assembly is investigated through molecular dynamics analysis. Robust results are found where a shear stress of the order of 20-50% of the failure stress induced an a - P-transition in the poly-Alanine region. The results are in agreement with the experimentally determined secondary structure and pulling forces of spider dragline silk. While the transition stress is independent of the chain length, the crystal is stable only in larger configurations. This minimum size for a stable structure is shown to be six poly-Alanine regions for a single chain and four poly-Alanine regions for multiple chains. This marks the smallest molecule size that gives rise to a 'silklike' structure. While the poly-Alanine region plays a key role in the transition from helix to sheet, other parts of the sequence (Glutamine, Arginine) may also be involved in the stabilization of the molecules. In general, the stability of the Psheet structure seems to arise from a close proximity in space of the helices in the spidroin state. This study emphasizes the role of shear in the assembly process of silk and can guide the design of microfluidic devices that attempt to mimic the natural spinning process. Establishing the molecular details of the assembly process can guide the synthesis of bioinspired protein materials by designing sequences that transition with high probabilities to stable P-sheet structures. 72 4 Heterogeneity Nanoconfinement and of Fibrils to Increase Strength and Toughness The research and review presented in this chapter have been published in: " T. Giesa, M.J. Buehler, Nanoconfinement and the Strength of Biopolymers. Annual Reviews of Biophysics, 42, pp. 651-673, 2013 " T. Giesa, N.M. Pugno, J.Y. Wong, D.L. Kaplan, M.J. Buehler, Mhat's Inside the Box? - Length Scales that Govern Fracture Processes of Polymer Fibers. Advanced Materials, 26, 3, pp. 412-417, 2014 In this chapter, the following questions regarding silk's fracture mechanical properties are answered: " How is the heterogeneity of silk fibers related to its fracture mechanical properties: strength and toughness? " How does nanoconfinement relate to the process zone size of fibers and ultimately increase their strength? " How is flaw tolerance achieved in silk and other biopolymer fibers? " What other strengthening mechanisms can be observed in biopolymer and bio-composites? " How is the nanoconfinement related to the silk assembly? 73 Research strategy: The first part of this chapter is a review of size effects observed in the mechanical strength of biopolymers, specifically spider silk, that are organized in microstructures such as fibrils, layered composites or particle nanocomposites. Two natural strategies are emphasized: confined mineral platelets that transfer load through a biopolymer interface in nanocomposites and confined fibrils as part of fibers. The application of confinement as a mechanism to tailor specific material properties in biological systems is discussed. Furthermore, it is shown how nanoconfinement of basic material constituents at critical length scales relates to the mechanical performance of the entire material system (elastic modulus, strength, extensibility, and robustness). Analytical fracture mechanical arguments are presented to illustrate the relation between fracture strength and heterogeneity. The concept of flaw tolerance on multiple hierarchical levels is connected to nanoconfinement. It is shown that the considerations of interatomic interactions alone cannot explain the fracture strength observed in biological fibers. Instead, the fracture strength of a fiber depends strongly on the lengthscale of observation, and structures at multiple length-scales must be considered to explain their remarkable mechanical performance and resilience, including a fiber's sensitivity with respect to cracks (and other flaws). 4.1 Background Many manmade materials and structures bear a conflict between strength and robustness; strong materials are often fragile whereas robust materials tend to be soft [4, 24]. In nature, this conflict is resolved by the arrangement of structures at multiple scales and the confinement of the constituents (building blocks) at critical length scales. One of the remarkable u.'etes natural material is thr ability to combine organic and inorganic components at the nanoscale [20]. In 74 materials like bone, universal patterns form a nanoscale composite of strong but brittle nanoscale minerals (hydroxyapatite) and softer but more ductile biopolymers (collagen) through assembly into complex shapes on seven or more levels of hierarchy, giving rise to dramatically enhanced mechanical properties compared with those anticipated from the basic constituents. The exploitation of nanoconfinement design principles reviewed here may pave the way for a wide range of tailored properties in biomaterials, outperforming the design space of classical engineering materials and providing a powerful paradigm to scale up molecular properties to the macroscopic world. 4.1.1 Nanoconfinement Strategy Nanoconfinement is one of the most important mechanisms that control the properties of biopolymers and polymer composites. Its exploitation defines a potent mechanism to create specific material properties in biological systems. When the characteristic dimensions of a polymer-based material approach molecular sizes, properties such as chain structure, chemical reactivity, and thermodynamic behavior change significantly [143]. This effect arises in biological systems with limited space, such as cells, shell nanocomposites, and thin fibers. The thorough understanding of fundamental confinement paradigms is especially important in view of the vast range of applications of such polymer structures as sensors, actuators, transistors, surface layers, filler matrices, and many others [144]. The thermodynamics of confined polymers and the strength of biopolymers are two intensively studied fields, but have never been thoroughly connected. The quantities that determine the strength of a material are elastic modulus, strain at failure, and failure strength (stress), as well as fracture toughness. Fracture toughness is a material parameter that indicates the ability of a material to withstand fracture. The influence of confinement on these quantities is not well 75 understood. Figure 22 shows the effect of the system dimension D (in the case of a fiber, its diameter; in the case of a thin film, its thickness) on the elastic modulus E and on the fracture strength a for a selection of materials. The quantities are normalized by their respective bulk value (bulk modulus Eb and bulk strength ab), see Table 3. 76 a 4.5 # Silk-elastin nanofibers (exp.) a Okra nanofibers (exp./sm.) a Polystyrene thin film (exp.) thin fim (sim.) * Polyethylene nanofiber (sim.) V 4 3.5 F V o o a Polystyrene 3 e + * A 2.5 2 1.5 F 0 ?I a ....... I I age 0.5 F "0 0.5 1.5 D/D 1 2 2.5 3 b - W * Nylon 6.Ar anofibers (exp.) A Polypyrrole nanotubes (exp.) v PAMPS nan ofbers (exp.) A 3 A Bombyx Mod silk (exp.) v Polycaprolactone nanofiber (exp.) 2.5 - 2 - 1.5 4 * Okra nanofiber (explsim.) - -sbeet nanocrystal (sim.) * MaSpI silk nanofiber (sim.) * Polystyrene thin film (exp.) YNA .4 ......#....................... ----------- 1 00 0.5 0 1 2 3 D/D 4 5 6 Figure 22 | Relative elastic modulus and strength of polymer materials as a function of size. (a) In most nanofibers, the relative elastic modulus, as compared with the bulk modulus Eb, increases dramatically at a critical diameter D*. This behavior is explained by the role of spatial confinement on entropy and the dominance of intermolecular interactions in thin nanofibers. Thin polymer films with free surfaces tend to display a decrease in the relative modulus due to the formation of highly mobile surface layers. (b) The alignment of crystalites and the degree of crystallinity in the fiber also improve with smaller diameter, leading to greater strength and toughness, sometimes even exceeding the bulk properties. The data for the fracture strength of thin films show a decrease of the material strength. (The references, critical length scales, and bulk values are summarized in Table 3.) Figure reprinted from Reference [121]. 77 Table 3 1 Bulk modulus, bulk strength, and critical length scales for polymer materials. Taken from Reference [1211. Eb [MPa] D* [nm] References 70 [146, 147] 20 35,000 [149] 160 18 6b [MPa] Polypyrrole nanofiber 500 Silk-elastin fiber 1,500 Polyethylene nanofiber 800-2,750 MaSpl silk fiber -11,400 Polystyrene thin film 5,200 1[1511 120-80 50 20 [47, 48, 153] 1[155, 156] The most important aspects that relate nanoconfinement of biopolymers to their performance as materials are: " Change in conformational behavior. Confinement influences the equilibrium shape and can lead to increased alignment of the polymer chains. On free surfaces, thin films display increased mobility due to a smaller interaction strength. In turn, this decreases the modulus of thin polymer films. " Increase in effective surface area. Splitting a fiber into n fibrils yields an increase in surface area by a factor of V, giving rise to surface effects such as increased chain mobility, less packing frustration, and increased diffusion. Furthermore, the effective adhesion area increases (van der Waals interactions). This is relevant in the context of critical length scales for nanosized fibrillar structures (e.g., gecko adhesion). " Change in elastic properties. Depending on the shape and environment of the polymer, the modulus and glass transition temperature can increase or decrease dramatically. 78 * Change in failure mode. Below a critical size, beam-like systems tend to fail in a shear rather than in a beam bending deflection. This has been shown to be advantageous for systems that are sensitive against tensile stresses. For instance, P-sheet nanocrystals are confined below 2-4 nm to avoid a tensile stress-induced bending, which facilitates the intrusion of water into the system, resulting in the breakdown of the crystal. * Additional fracture mechanisms. Structure-splitting induces fracture mechanisms such as crack deflection on interfaces, fiber bridging, crack branching, and others. Paired with a composite structure of a weak, ductile phase and a strong, brittle phase-as found in many natural systems (e.g., bone, nacre, silk)-this design dramatically increases the fracture toughness, robustness, and damage tolerance of the whole system. An extreme case is the repeating patterns of brittle phases with structural holes, i.e., gas. This arrangement, found, e.g., in the silica structure of diatom algae, gives rise to improved strength, extensibility, and robustness. * Homogeneous deformation. Confinement below a critical size forms a potent design strategy to map nanoscale properties to the macroscale despite system imperfections such as cracks and tears. Confinement is not the only mechanism that contributes to the strength and toughness of biopolymers and polymer-mineral composites. On various scales, additional extrinsic mechanisms such as fiber bridging and grain bridging that shield cracks, as well as plasticity-like intrinsic mechanisms [24], friction on the scale of hundreds of nanometers [158], crack branching and microcracking in composite structures [159], and many others increase the toughness of most biomaterials. Here, the discussion is limited to two generic natural strategies (universally observed in nature and representing critical mechanisms that play an important 79 role in developing a comprehensive understanding of biological materials): confinement of the biopolymer interface in composites [quantitatively equal to thin film structures [160]] and confinement of fibrils in fibers. Both these mechanisms are ultimately important to understand the importance of spider silk's nanostructure. 4.1.2 Confinement of Polymers To understand the effects of confined system dimensions on the material properties of a polymer, the basic thermodynamics and kinematics of the polypeptide chains must be analyzed through experimentation and simulation. The statistics of confined polymer materials can be studied by molecular dynamics or Monte Carlo simulation [161-164]. The most important properties of a spatially confined solution are: the nature of the confining surface, geometry of the confining object [165, 166], and chain flexibility [167, 168]. Recent research, e.g., by Zhou et al. [169], has explored the influence of spatial confinement or macromolecular crowding on (re)folding, equilibria, shape, and entropy of biomolecules in cells. The presence of a physical boundary reduces the accessible volume of the chain and thus the entropic free energy by SG = -kBT(ln(fconf) - n(fth)) , 4.1 where T is the absolute temperature, kB is the Boltzmann constant, fconf is the accessible volume fraction, and ft is the theoretically allowed fraction of configurations (with respect to the total system volume) [170, 171]. Hence, a confinement provides a stabilization of the native state, decreases the folding free energy, and increases the melting temperature. A mechanism similar to confinement is macromolecular crowding [172, 173]. Close to thin film boundaries in liquid-crystal polymers, the interfacial perturbations create various types of new local organizations, such as an alignment along the surface (called 80 weak nematic ordering) [174-176]. The shape of the confined zone can critically influence the rates of protein folding [166]. Therefore, confinement is a complex problem, highly dependent on boundary conditions such as substrate, environment, and shape. It is thus of utmost importance to include the ramifications of these effects into molecular studies [177]. Much work has been done on the thermodynamics of confined polymer and liquid-crystal materials. The influence of confinement on the glass transition temperature has been intensively studied and comprehensively reviewed [178]. The behavior is complex and dependent on many variables, such that decades of research still cannot provide consistent answers. A general review on the effect of confinement on thermodynamic properties of glassy polymers by molecular dynamics simulation can be found in Reference [179]. Recent studies include experimental [180-183] and computational [184] investigations, also in the context of nanocomposites [185, 186], many of which examine polystyrene as it displays strong size effects [187]. Confinement dramatically changes the elastic properties of polymer materials. Arinstein & Zussmann [146] reviewed the effect of reduced sizes and dimensions on the mechanical and thermodynamical behavior of electrospun polymer fibers. They identified critical length (nanometer) scales at which the material behavior deviates significantly from the bulk behavior. This includes an increase in elastic modulus with smaller diameters, and the temperature dependence of the elastic modulus [188]. For larger fibers, the modulus increases with elevating temperatures, as predicted by rubber elasticity, whereas it decreases with rising temperature for smaller-diameter fibers, similar to solid and crystalline materials. This behavior is explained by the role of spatial confinement on entropy and the dominance of intermolecular interactions in thin nanofibers. Also, the degree of crystallinity and orientation of the crystallites increases with smaller diameters [188], similar to thin films [189]. The crystallization of polymers in thin films 81 plays an important role in defining their mechanical response. Crystal orientation is enforced in polymers with larger Kuhn lengths, and the polymer can be modeled as a system of self-avoiding rigid rods (Onsager model). It can be shown via the correlation length of a crystalline region in an amorphous domain that the size of the oriented region is in the same order as the fiber diameter at which the dramatic increase in elastic modulus is observed. The experimental results for elastic modulus of confined thin films are ambiguous as the measurements are often highly influenced by the stiff substrate [190, 191]. The controversial discussion as to what extent surface effects and supramolecular microstructure control the behavior of thin films and nanofibers is still not settled. In materials for which surface effects dominate (amorphous thin films), the modulus tends to decrease with confinement (in comparison with the bulk behavior), whereas in materials with significant crystalline phases (polymer fibers), the modulus can increase by up to a factor of 20. The confining substrate and the molecular mass of the polymer influence the confinement behavior, which is of great relevance to the analysis of composite materials [192]. 4.1.3 Strength of Polymers 4.1.3.1 Strength of Hydrogen Bonds The failure of hierarchically organized protein domains is governed by the competing mechanisms of H-bond breaking and bond reformation. Such a process can be modeled by statistical models based on Bell's theory or Kramer's diffusion model [193]. In these frameworks the existence of a H-bond between a donor and acceptor atom is treated as a random variable, where the number of attempts to break or reform the bond per second is prescribed by the natural vibration frequency wo ~ 10" s-'. Hence, w0 represents the attempts of each bond to overcome an energy barrier Eb. An externally applied force f reduces 82 this energy barrier. The energy associated with the force is fxb cos(8), where Xb is the distance between the equilibrated state and the transition state, and 0 is the angle between the direction of the reaction pathway of bond breaking and that of the applied load. The off-rate X is the product of the vibration frequency and the probability to overcome the energy barrier and is given by: Eb - fxbcos(6) kBT 4.2 ) = ( X woexp ( The off-rate is the reciprocal of the bond lifetime. Note that this equation is related to the probability of state transition, see chapter 2.1.7. For complex chemical reactions, a similar model can be developed that includes higher-order correction terms [194]. 4.1.3.2 Strength of Hydrogen Bond Clusters The common basic secondary structures found in protein materials are a-helices, -sheets, and P-coils. Recently, their mechanical performance has been elucidated by atomistic simulation [195]. These structures can be understood as nanoscale 'brick-and-mortar' - a strong constituent (e.g., the polypeptide chain) is stabilized by a weak glue-like matrix (e.g., clusters of H-bonds) [196]. To study the unfolding of these structures, the hierarchical architecture that enhances H-bond cooperativity must be correctly captured. Bond cooperativity is the ability of hydrogen bond clusters to act synergistically to withstand failure; usually enforced under shear deformation rather than bending deformation. The rupture strength of H-bonded protein domains can be determined based on the energetic analysis discussed above [197]. In an array of H-bonds, the optimal number Ncr of cooperating bonds that fail simultaneously in a cluster is given by: 83 Ncr = kT Eb (4(p [(1 acr)2 + 4acr - 1] - In (UXO)I 4.3 where acr is the ratio between end-to-end length and contour length of the polymer chain at failure and 4, is the persistence length of the polypeptide. It was found that for H-bond clusters Ncr ~ 4, in accordance with cluster sizes found in nature (Figure 23) [157]. Protein materials such as spider silk, muscle tissue, and amyloid fibers consist of arrangement of ordered and unordered (amorphous) structures. An example of an ordered structure is the P-sheet, composed of hierarchical assemblies of weak H-bonds (100-1,000 times weaker than covalent bonds). Despite the intrinsic strength limitation of H-bonds, these materials gain exceptional strength and robustness by clustering weak bonds in groups of three to four that then break concurrently. For hierarchical assemblies of H-bonded clusters Ackbarow et al. [195] devised a model to capture the force distribution in a hierarchically organized network. The off-rate for such a hierarchical network Xh is given by: Xh = wo exp ( Nc~rEb + kBT In (Ncr) - fxbcos(6) kN 4.4 where Ncr is given by equation and N is the total number of bonds in parallel. They also defined a robustness index r(Ncr) = 1 + (kbT In Ncr - Eb)/(EbNcr). Again, Nr 4 optimizes the robustness in the Pareto sense. 84 .... ............. . . ........ .. -2-F" -- I I-- I 10 12 200 -sheet 9150 100 50 0 2 6 4 8 #of residues 14 16 Predicted critical size -sheet Dimensions found in several natural protein secondary structures P-helix -* a-helix' Figure 23 I Rupture strength of H-bond clusters and critical sizes of some protein secondary structures. Top: The intrinsic strength limit of H-bonds can be overcome by clusters of three to four H-bonds that then interact synergistically to resist deformation and failure. The H-bond assemblies are loaded in parallel in ahelices, P-sheets, and P-helices. Note that the shear strength curve is derived for a single n-sheet in a pull-out test. The natural load condition for an a-helix is tension, for which an unzipping effect is easily achieved. Bottom: This result explains the cluster size found in natural protein secondary structures (a-helix: N = 3.5; 0-helix: N = 5; and 1-sheet: N = 2.5 - 8). Figure adapted with permission from [157] Copyright 2008 American Chemical Society. 4.1.3.3 Strength of Single Molecules Many biopolymers, e.g., amyloids, spider dragline silk, and titin as part of the muscle sarcomere, gain their strength through the arrangement of p-sheet rich crystalline regions and non-crystalline (amorphous) regions [198, 199]. Fundamental insights into their mechanical performance is gained by single molecule experiments. The strength of single molecules can be tested experimentally and computationally. Comprehensive reviews of single-molecule experiments in biological physics can be found in References [69, 200]. Atomic force microscopy thereby provides a tool for a variety of characterization tasks, 85 such as determining the surface morphology, the elastic modulus, and the yield strength of small-scale samples [201-204]. On a similar length scale, the strength of polymer complexes can be determined by a computational method called steered molecular dynamics (see chapter 2.1.3) that simultaneously sheds light on the (un)folding pathways of proteins under force, as well as their corresponding intermediate states [205-207]. 4.1.3.4 Nanoscale Surface Strength - Weak Interactions Nanoscale biological surfaces feature properties of self-similar and self-affine fractals, that can be treated by mathematical analysis. Self-similarity is the property whereby an object's structure at a larger scale is equal to that at a smaller scale and self-affinity is the property whereby a substantially different amount of anisotropic scaling does not affect the morphology of an object's surface. Splitting a fiber into n fibrils yields an increase in surface area by a factor of Ni. Similarly, fractality increases the effective surface area dramatically, which enhances adhesion and diffusion. On the other hand, increased chain mobility at interfaces can decrease the interfacial strength [208]. For general fractal surfaces, the effective adhesion yeff can be derived as a function of measurable surface geometry properties and the van der Waals surface energy difference [209]. Hence, it is possible to calculate the interface strength of various materials, including biological systems such as the gecko foot [209]. In hard solids, the area of contact within the effective cutoff is extremely small, and additionally, they store only limited elastic energy. Although gecko skin is primarily composed of stiff keratin, the branching fibers and nanometer-sized spatulae lead to the formation of a soft, elastic layer that can follow the profile of the substrate at all length scales [210]. The gecko adhesion effect has been successfully mimicked, but the efficiency of natural systems, i.e., the transmitted load per projected area, still cannot be reached [211]. 86 4.1.4 The Relation between Confinement and the Strength of Polymers Most biological materials are hierarchically organized, i.e., they consist of hundreds of nanoscale features that are copied many times, e.g., in a self-similar way [1]. The protein constituents are often of molecular size, hence the abovediscussed changes in properties are relevant in these materials. The following chapter outlines how the specific effects of confinement come to work in biopolymers and biopolymer composites. The focus of this discussion is mainly on the confinement of the final structure, as this is crucial to the actual performance of the system, and not the effect of confinement during the growth process. The question of how the confined systems are actually self-assembled is of a different nature, as a detailed analysis of the growth process is needed. Specifically, the growth of the crystals in polymer-mineral nanocomposites is likely controlled by the free surface energy provided by the surrounding polymer matrix (i.e., a nucleation problem), whereas the size of silk fibers, for example, is controlled by the conditions during the spinning process (e.g., channel size, shear rate, and pH). This is discussed in chapter 3. 4.1.4.1 Thin Films Nanoconfined thin films (on a substrate, but with a free surface) tend to decrease their fracture strength relative to their bulk strength, which has been confirmed experimentally [1561 and computationally [2121, see Figure 22. Furthermore, it has been reported that their strain at failure increases with confinement [213]. This can be explained by the decrease in intermolecular entanglement density near the free surface leading to increased segmental mobility but reduced strength of the intermolecular forces [156, 208]. Thin biopolymer films with freestanding surfaces are relevant in many applications, for example, as sensors and coatings. Thus, their low mechanical performance at small scales is not necessarily a performance issue. 87 4.1.4.2 Biopolymer and Mineral Composites Composite biomaterials consist of multiple distinct materials (e.g., organic and inorganic as in brick-and-mortar) or multiple phases (e.g., crystalline and random-coil) and feature enhanced characteristics constituents alone 215]. [214, Their applications with respect to their and properties in nanotechnology have been reviewed in Reference [216, 217]. For example, polymer-layered silicate nanocomposites are much lighter than a bulk polymer material of equivalent strength and stiffness, are easier to recycle, and do not suffer from the 'one-dimensionality' of fiber-reinforced materials [218]. The mechanisms that play a key role in these materials rely on the transfer of shear stresses across the polymer-mineral interface. The stiff boundaries of the mineral phase diminish the potentially negative surface effects (e.g., high chain mobility) and rather enhance the strengthening effects (e.g., an increased shear transfer capability). In most applications, catastrophic failure can be avoided by using strong and tough (fracture-resistant) materials that fail gradually. Hard materials tend to be brittle because the high strain energy due to stresses at the crack tip cannot be dissipated, whereas soft materials dissipate the energy by plastic deformation. Composites rely on a synergistic interplay of crack arrest, stress delocalization, and dissipation, see Figure 24. Specifically, natural composite materials rely on intrinsic and extrinsic toughening mechanisms [24]. Intrinsic mechanisms (based on plastic deformation, e.g., platelet sliding) usually originate from smaller length scales (akin to dislocations in metal), whereas extrinsic toughening mechanisms (e.g., fiber bridging) usually take place on micrometer scales. Nacre, for example, features a brick-and-mortar structure, where the soft phase allows small movements between the strong aragonite platelets. As the movement is limited to -1 pm due to a friction mechanism, this constitutes a toughening mechanism. Crack deviation and platelet pullout provide further contributions to 88 a fracture toughness of an order of magnitude higher than either of its constituent phases [24]. These mechanisms have been applied to synthetic bioinspired ceramic materials made of aluminum bricks and a polymeric lubricant with toughness values exceeding 30 MPa Vi (for comparison, the value for aluminum is -20 MPa Viii, and for the polymer it is -1 MPa V). Furthermore, the repeating patterns in a hierarchical material contributed to an improved redundancy; i.e., a partially flawed system maintains its mechanical functionality through a load bypass. b a 40 Intrinsic toughening Brick-mortar A Extrinsic toughening E 30-p **1. zone Grain bridging vage fracture L6 ijrovoidcoalescence I C 10 Nacr e Ahead of crack tip Behind crack tip A Lamellar 00 03 1.2 Fibre bridging 20- K3 =0.012 A1203 0.2 0.8 0.6 0.4 Crack extension, Aa (mm) 1 Figure 24 | Intrinsic (plasticity) versus extrinsic (shielding) toughening mechanisms associated with crack extension and R-curve. (a) The illustration shows mutual competition between intrinsic damage mechanisms, which act ahead of the crack tip to promote crack advance and extrinsic crack-tip-shielding mechanisms, which act primarily behind the tip to impede crack advance [24]. Intrinsic toughening results essentially from plasticity and enhances a material's inherent damage resistance; as such it increases both the crack-initiation and crack-growth toughnesses. (b) Toughness behavior of various materials. In many natural materials, it is an order of magnitude tougher than its constituent phases. Figure adapted from [24], copyright @ 2011, with permission from the Nature Publishing Group. In their fundamental studies on the fracture strength of nanocomposites, Gao et al. [219] derived a fracture mechanics argument to determine the optimum platelet aspect ratio and platelet size. Their analysis is based on the concept of 89 nanoconfinement and flaw tolerance (scaling of fracture strength). It is often argued, at least for crystalline materials, that the increasing strength of confined materials is related to the decreasing occurrence of defects on smaller scales (see, e.g., Reference [220]). In contrast, Mishra et al. [221] discovered that the defect density actually dramatically increases with decreasing film size. Their analysis is valid for block-copolymer nanofilms with low Flory-Higgins parameter. If a dimension (in the case of a composite, the height of the platelet size h; in the case of a fiber, its diameter D) becomes confined to the size of the zone of large deformation around the crack tip, the stress concentration vanishes and the composite/fiber is homogeneously deformed. This defines the flaw-tolerant state, as none of the material strength is lost [215, 219, 222]. The flaw tolerance mechanism, Figure 25, has been experimentally confirmed for thin, steel sheets and for notched, nanoscale thin films [223] and recently also via molecular dynamics simulation for graphene [224] and hydroxyapatite [225]. In each case, no stress concentration could be observed and failure initiated far away from the crack tip. For many materials including nanotubes, the critical size tends to be close to the size of the unit cell of a crystal [226], but depends on the boundary conditions and other material properties [47, 219, 227]. Gao et al. [219] argue that during the biomineralization process, crack-like flaws in the form of soft protein inclusions are trapped within the mineral crystals (Figure 26). They derive a scaling law for the optimum flaw-tolerant platelet size h* h = F2 ysEm 4.5 U2 Below h* the fracture strength of a perfect crystal is reached despite the presence of a crack. Therein, F is a geometry-dependent parameter of order unity (see chapter 2.2.1), y, is the surface energy, Em the elastic modulus of the mineral phase, and Ut the strength of the perfect crystal. The results from molecular dynamics simulation of the coarse-grained (CG) model of bone are shown in 90 Figure 26. Coarse graining is a molecular dynamics technique that reduces degrees of freedom and is often used for protein materials simulation. a 1.73pm 0.17pm 0.05pm b I 10 nm Flaw tolerant state 1 d C Strength scaling of a flawed crystal (classical) 1 W 1h , D Figure 25 1 Confinement and flaw tolerance. The graph shows the concept of flaw tolerance. According to the classical prediction, the strength of a material scales with Jiili or, Ji73 respectively. In the case where the strength of the perfect crystal is reached at h = h* (D = D*), the flawed system exhibits no loss of strength. This concept has been experimentally and computationally verified, e.g., for (a) spider dragline silk [47], (b) hydroxyapatite nanocrystals [225], (c) thin metal strips [223], and (d) nanocrystalline graphene [224]. Images in panel (a) reprinted from Reference [47]; images in panel (b) from Reference [225]. Images in panel (c) adapted with permission from Reference [223]. Copyright @ 2009, American Institute of Physics. Image in panel (d) adapted with permission from Reference [224]. Copyright @ 2012 American Chemical Society. Figure adapted with permission from Reference [121]. 91 Y 1 b a I OA 0 0.8 Ih!I 0.6. 0.40.2- At 1 - Hydroxyapatite Surface Adhesion Spider Silk - A 2 Wh, IIColagnma aCG Bone Model * 3 C Flaw due to surface 'ouC1"" 4 rD/D* Figure 26 1 Critical size of a composite and an adhesion system. Bone-like materials typically consist of fragile, brittle mineral platelets (hydroxyapatite) embedded in protein matrix materials (collagen). (a) The mineral platelets carry a tensile load and the protein transfers carry loads between the platelets via shear. (b) These platelets are confined and optimally arranged to maximize the strength and toughness of the material. (c) Similarly, the adhesion of a spatula on a rigid surface is optimized for a critical diameter D*, e.g., for gecko adhesion [211]. The critical sizes h* and D* determine the point at which the system becomes flawtolerant. The solid line on the left corresponds to the classical fracture mechanics prediction, which breaks down at the length scales at approximately the critical size (on the order of a few nanometers). Figure adapted with permission from Reference [121]. Wei et al. [2281 derived a closed-form solution for two critical overlap lengths L* and L** to which the polymer/mineral interface is confined. They found that L* = 2.318 Ehb 4.6 - in the elastic limit using the shear-lag model to maximize the elastic strain energy density and the shear transfer efficiency. The respective parameters are shown in Figure 27a. For the plastic, limit they found that 4.7 L** = U b, Tf at which the strong and the weak phase fail simultaneously. Here, Tf is the shear strength of the interface (Figure 27b). Wei et al. also showed that in the case where L* ~ L**, the toughness and the strength of the composite are both 92 optimized. Comparing their model predictions with abalone shell, collagen, and spider silk nanocomposites, they find a remarkable agreement with the overlap length scales (Figure 27b). De Gennes & Okumura [229] developed a scaling argument for the strength of biocomposites with stratified structure based on a CG elasticity field that neglects plastic flow and viscoelasticity where 0f th lEs(hs + a og 8 h4.8 Ehds with the subscript h for the stiff mineral phase and subscript s for the weak polymer phase, h as the respective layer thicknesses, a as the crack size, and 6 as a cutoff length scale in the continuum (-h). In the thin layer limit and by neglecting bending of the platelets, they approximate the energy functional for a notched, stratified structure where the organic layers are much thinner than the platelet size. Based on their analysis, De Gennes & Okumura find that nacre enhances its fracture toughness by a factor of 1,000 and the fracture strength by a factor of 30 in comparison with pure aragonite. Notably, this behavior could be controlled by the stiffness ratio Es/Eh. This idea can be used for the design of composite materials, to increase the fracture toughness [230, 231]. 93 b 25 --- Fracture Toughness Elastic Energy Density L 1600 1400 23 21 _ -1200 ~21 - *1:2 10005 8000 LM 1600 1400 100. U2 1 200 0 C '- 15 0.5 0 2b 0 1 LA 1.5 2 107 103 0 Experimeental Data 17 10rs aModel Prediction 10 5 LU 100 10' 10, 101 Size (nm) composite (e.g., nacre). (a) Hierarchical structure scales length Figure 27 Optimized 2012 merian Soie0y Cemiclin a ineral-polymer of nacre and a schematic of a 2-dimensional continuum model for the composite architecture to predict the critical sizes that maximize the strength of the whole material. (b) Elastic and fracture toughness varying with overlap length normalized byL. Total elastic strain energy density (squares) maximizes at L = V, and fracture toughness (circles) exhibits a sudden drop when L > L*. (c) Comparison of overlap lengths for basic building blocks of three natural materials (nacre, tendon, and spider silk) from experimental observation (circles) and model prediction (squares). Figure adapted with permission from Reference [228]. Copyright@ 2012 American Chemical Society. 4.1.5 Nanoconfinement in Silk Silk heterogeneous nanostructure features aspects of composites materials (namely, the composite of 1-sheet crystal and semi-amorphous phase) and fibrous materials (fibrillation). As explained in the previous chapter, many fibrous biomnaterials achieve their strength by confining their basic building blocks at critical length scales. This can occur at several hierarchical levels, spanning from nanometers to micrometers, i.e., confined nanocrystals and confined fibrils. 94 The breaking of H-bonds represents a fundamental unit mechanism of failure in protein materials, see chapter 4.1.3.1 [157, 197, 232]. The studies of simultaneous failure of H-bond clusters (chapter 4.1.3.2) reveal critical dimensions for structural constituents such as helices and nanocrystals. For example, it has been shown that networks of a-helices are flaw-tolerant [233]. In a recent study, Keten et al. [49] determined the critical size for the most stable P-sheet nanocystal to be H = 2 - 4 nm. Figure 28 shows the robustness, a measure of its stability against failure, of a single P-sheet crystal as a function of its height H. Larger crystals do not fail in a shear deformation (with optimized cooperativity of the H-bond clusters with Ncr = 4, see chapter 4.1.3.2) but rather in a bending deformation. In this unfavorable deformation mode, water molecules enter in the crystal regions under tension and split up the crystal, leading to an early and catastrophic (brittle) failure. In accordance with the simulation, the crystal dimensions were experimentally determined to be 2-4 nm [234]. The P-sheet nanocrystal size may be tuned through genetic modification and control of the self-assembly process. For example, by controlling the poly-Alanine length in MaSpi protein aggregation in dragline silk, the resulting stability of the nanostructure can be tuned through the p-sheet crystal size [130]. Computational atomistic structure predictions with REMD (see chapter 2.1.5) predicted a critical poly-Alanine length of four to six residues for the amyloidization of a defined and stable p-sheet nanocrystal. This result has been experimentally verified through the observation that MaSp dragline silk typically has poly-Alanine repeats of six to eight residues in length, whereas the minor ampullate spidroin protein of viscoelastic capture silk (less strong) has poly-Alanine repeats of only two to four residues [235]. This study revealed an in-depth understanding of how nature controls material self-asssembly processes and creates its highperformance materials. Furthermore, it is an important engineering opportunity for tailored polymer materials. 95 35 -, 30 H 25 E 20 15- 0 0 Crystal size 2 4 H [nm] 6 8 Figure 28 | Robustness of f-sheet nanocrystals as a function of their height. P-Sheet nanocrystals are especially strong and robust if their height is confined to 2-4 nm. This critical dimension is in agreement with experimental results. Figure adapted from Reference [49]. Similar to the elastic modulus, the dependence of the fiber strength on the fiber diameter is ambiguous in the literature. Almost no effect of diameter has been reported in silk-elastin-like protein polymer fibers [149], whereas a dramatic increase is seen, e.g., in native and regenerated Bombyx Mori silk [152], natural plant fibers [150, 236], and polycaprolactone nanofibers [154, 237]. In fibrillar nanoscale adhesion systems, nanoconfinement has been shown to be the dominant strategy to ensure a robust design, thus giving an explanation for the widespread occurrence of hairy attachments in nature (gecko, fly, beetle, spider) [238]. Most biopolymer materials are flawed, meaning that preexisting cracks, inclusions, and tears compromise the mechanical integrity under load through the existence of stress concentrations. The mechanical strength and robustness of flawed fibers can also be maintained by use of a confinement strategy [219]. 96 Nanoconfinement of P-sheet crystals has been determined to be the driving mechanism in squid sucker ring teeth [239]. Recent work has quantified a flaw tolerance mechanism for spider silk fibers [47]. The result for the critical silk fibril size D*(MaSpi) = 20 - 80 nm, determined in Reference [47] via simulation, agrees remarkably well with experimental observations in the range of 20-150 nm [42]. This critical size can be similarly determined by Equation 4.5 using the specific geometric boundary conditions of a fiber. The results from MD simulation with different loading conditions are shown in Figure 29. Silk fibers consist of nanoconfined fibrils of 20-150 nm that are capable of maintaining their full strength and robustness despite possible flaws in the fiber material. The robustness can be for example measured by the toughness modulus, the area under the stress-strain curve of a material after fracture. The relation to the fracture toughness is discussed elsewhere [230]. A similar mechanism to increase the structural robustness could be at play in many biological fibers that feature a fibrillar nanostructure. This idea will be explored in the following chapters. 97 * Loading Condition (1) I 0.8 F =* a Loading Condition (2) - Loading Condition (3) A Loading Condition (4) N. Pilipes (exp.) C. Darwini (exp.) A. Argentata (exp.) Atomistic Simulation 2 0O- 80~ f/l 11 (1) (2) (3) (4) 0.9 0.8 E3 0.7 0.6 0.5 0.4 20.4 D> 1000nm 0 * 0.2F * A DA1 -0.2 OI 0.4 Relative Strain 0.8 I 10 Figure 29 1 Critical size of spider dragline silk major ampullate spidroin 1 (MaSpi). The graph shows the dependence of the failure strain and failure stress on the fibril size D under various loading conditions (1-4) as well as a direct comparison with experimental results (under the tensile loading condition 1 and the mechanical behavior of a defect-free silk fiber. For decreasing fibril sizes, the perfect material behavior (i.e., = 1,400 MPa failure stress and 68% failure strain) is approached and reached at D = D* = 50 30 nm. D* is denoted the critical flaw-tolerant size of the fiber. The results show that the high strength and extensibility observed in experimental studies can only be reached by nanoconfinement of fibrils close to D*. Figure adapted from Reference [471 with permission. 4.2 Fracture Mechanics Analysis In this chapter, the characteristic length scales in hierarchical biological fibrous materials are related to their fracture strength. The fracture strength of natural fibers like silk remains a topic of intense debate, as it has not yet been possible to predict their fracture strength directly from an atomistic point of view [240, 2411. The impact of such stress concentrators on a material's performance was first quantified by Griffith and Weibull, see chapter 2.2. They developed strong theoretical arguments to explain the experimentally observed scaling of strength with specimen size, where smaller tends to be stronger, suggesting that the probability of having a critical flaw decreases with diameter. 98 4.2.1 Importance of the Process Zone A polymer fiber of diameter D with no intrinsic flaws is considered (Figure 30a). Under applied tension uO, such a fiber's failure strength af would reach the theoretical limit of the interatomic bonds it consists of, denoted by ath. Similarly, one could consider a fiber with a distribution of intrinsic flaws. Such a fiber would fail at a finite 'macroscopic' strength. In contrast, a fiber made of the same bonds but containing a large flaw will significantly decrease its strength according to Griffith's size scaling [219, 222], where the crack size scales with the diameter, such that af ~ 1/T. The process zone, also depicted schematically in Figure 30b, characterizes the amount of material that contributes to resisting fracture, and is also referred to as the 'cavitation box' [240]. It can be understood as the region surrounding a crack that is damaged during crack propagation. In the limit of a very small process zone (Figure 30b), on the order of atomic bonds, the material is very brittle like glass and bonds simply snap to create new surfaces. In the limit of a very large process zone (Figure 30c), driving a crack through the material leads to widespread damage that is not limited to the crack surface. An important aspect of a large process zone is that stress concentrations at the crack tip are diminished, further reducing the threat imposed catastrophic damage to a material. 99 by cracks of causing C b a oo D af=oa at r D D 0 unflawed no nanostructure with nanostructure Figure 30 I (a) A fiber of diameter D that has no intrinsic flaws. Under tension ao, such a fiber's failure strength af would reach the theoretical strength of the interatomic bonds it consists of, ot. (b) A fiber made of the same bonds without internal structure but containing a flaw of length a would decrease its strength according to Griffith's size scaling as the ratio 1 0/D and the strength of the fiber become smaller. (c) A possible strategy to maintain the strength of the fiber at the macroscale is to increase the size of the process zone, such that 1o = D. Then, the strength of the fiber will approach the theoretical strength of the internal structure, af Cath. Figure adapted from Reference [100]. 4.2.2 Derivation of the Process Zone Size The process zone size, characterized by the length-scale to, is indicative of the fracture toughness K of the fiber [242, 243], since K-,fl0 [244]. This makes intuitive sense because dissipative processes that increase the size of the process zone will thereby lead to a larger fracture resistance of polymeric fibers. For example, a self-healing processes (via H-bond stick-slip or hidden length in molecular domains) can increase the toughness during fracture. In addition to molecular processes, distributed failure processes can encompass micro-cracking, crack bridging, crack-deflection, as well as interfacial sliding (e.g., shearing of interfaces and unfolding of organic matrices) [1, 20, 24, 159, 241]. How can the process zone size 10 be identified? In most cases, it can be measured by experimental analysis only. Some progress has been made from a continuum mechanical point of view [93, 96, 245]. One can estimate the process zone size 10 100 from the ratio between a materials fracture strength ug and its yield strength -y (defined as the stress where the material undergoes irreversible, plastic deformation), 10 = fl UY 4.9 , where fl is proportional to the crack size, depends (for example) on the specimen geometry, and can include nonlinear geometric effects [92]. Equation 4.9 is related to Equation 2.19, chapter 2.2.1. The scaling law in Equation 4.9 reflects the importance of the nonlinear nature of the stress-strain relationship as a means to decouple the failure stress from the yield stress. However, this does not provide an immediate route to predict 10, because the ratio of failure to yield stress is not generally known. 4.2.2.1 Experimental observations For many biomaterials, the process zone size 10 has been experimentally determined to be on the order of several micrometers, e.g. for bone [20, 246], polymers [247-249], and quasibrittle materials [250, 251]. The process zone has also been directly measured in concrete, graphite, and wood via acoustic emission, digital imaging, and SEM [252-255]. In these studies, the damage zone surrounding the crack tip was visually identified as process zone. In a recent paper, empirical relations were used to estimate the process zone size 10 in spider silk and other polymeric fibers, and found to be on the order of one micrometer [240, 256]. To obtain this result, the authors invoked an empirical scaling relation that relates the yield stress to the elastic properties via -y = 0.028 E. As other analyses, Reference [240] did not describe the fracture properties of fibers from a fundamental, interatomic potential point of view but rather lumped a variety of mechanisms into an empirical equation that ultimately relates the yield strength to the fracture strength. 101 4.2.2.2 Atomistic Derivation The process zone size can be directly estimated from an interatomic potential. This fundamental approach accounts for the fact that fracture involves the rupture of atomic bonds. The generalized Lennard-Jones-n, m-potential (J) is often used as an approximation for the behavior of interatomic interactions in polymers and crystalline materials (e.g., Reference [257]), and therefore provides a good basis for the study of the fracture mechanisms in natural fibers [258, 259]. The generalized LJ-(n, m)-potential formulation is given by n m n -L) L(rnm)=-) r- 4.10 , n-mr Here, ro is the equilibrium bond distance, r the current bond distance, and W, n, m potential parameters. Similar to the derivation in Reference [258], one finds for this potential the strain energy density W =4.11 the stress-displacement function with a = dW/&E, where E = (r - ro)/ro is the bond strain, n+1 nmw a(r, n, m) = (n - ) + m+1 0 4.12 - , and the elastic modulus E = d2 W/gE 2 nmw E(r, n, m) = nm- (ro n+2 _ (+1)rom+2] [(n + 1) - (m + 1) - ], 4.13 with fi as the bond volume. The theoretical attainable bond stress is given by setting E = 0, _Uth "nt, Mi) = , nmw (m m+1 + _1n-m 414 102 Furthermore, the cohesive energy, which equals the fracture surface energy of the bond, is given by 4.15 dr = - Ys = fu(r) These definitions have been reported in Reference [258]. The process zone 10 is given by the Dugdale-Barenblatt yield-strip model (see chapter 2.2.2), 1 = 8 E 4.16 2' where GC is the critical energy release rate, a material parameter that quantifies the amount of energy needed to drive a crack through a surface. Usually, derivations for nonlinear constitutive behavior entail a linearization of the elastic modulus around E = 0 (i.e., r = r0 ) [258], nmw EO(n, m) = 4.17 ,2' The modulus as it appears in Equation 4.16 is the modulus of the stress far field, to account for the change in strain energy in the specimen upon crack propagation. In a very large specimen, it is reasonable to set E = E0 . However, if the specimen size is small (in the order of the plastic zone), then E # E0 . This is the case for nanoconfined constituents in many biological materials, e.g., the fibrils in spider dragline silk or cellulose fibrils in wood, as discussed in chapter 4.1. In most nonlinear elastic material models, the stress behind the crack tip degrades with 1/VT, similar to the linear elastic case. This implies that the material within a zone of 3-5 times the crack size (which can be estimated from fracture mechanics, here without proof), has a much lower modulus than the far field. In a nano-sized material, this zone can contain the entire specimen. To 103 account for the highly nonlinear behavior in the cohesive zone, it seems more appropriate to calculate an average elastic modulus that averages from the crack . tip, E = 0, to the linearized modulus E = E0 nmwo nm+ = oth(n, m) m+1 . E(n,m) (( m f2(n + 1) n ++1-1 4.18 Most materials have a modulus that is higher than the theoretical strength, so the estimate E = P will serve as a lower bound for the process zone size 10 in Equation 4.16. Due to the elastic nature of the potential, G, as given by Equation 2.21, chapter 2.2.2, can be simplified to G= Ys + dy 0 (fuijdEi ys. 4.19 Now, an expression for the plastic zone size only in terms of the bond potential parameters using Equation 4.16 can be written. The lower bound for 10 is given by the averaged modulus in Equation 4.18, M+1 n(n + 1) (n+1) n10'0, = 4.20 r. 8nm Alternatively, using the linearized expression of the elastic modulus, Equation 4.17, an upper bound for 10 is given by, 2 + 1) 8nm w(n + 1)2 (n 10,E= m+1 n- 4.21 r The values for 10 are only a function of (n, m) and the bond distance ro. This assumes a 1D arrangement of bonds, for example present in P-sheets. For a lattice-like arrangement one would need to sum over the contributions from several bonds, which would lead to an increase in 10. 104 Equation 4.20 and 4.21 give the ratio of process zone size to the characteristic length scale ro, with 10/ro = f(n, m). Here, f(n, m) is a function of the parameters of a generalized Lennard-Jones-potential. Table 4 summarizes the results for typical combinations of (n, m) and it can be seen l0 /ro is relatively insensitive to the parameters (n, m). Estimating an upper and lower bound for this ratio yields an interval of l 0 /ro ~ [0.5,16], or [0.25 nm, 8 nm] for most real materials. An important observation from this derivation is the existence of a relation between characteristic length scale ro and process zone size 10, which is in turn directly connected to the fracture toughness of the material. On the level of chemical bonds, one can think of ro as the 'lattice' spacing in the strong and often ordered polymer domains. In strong and tough protein or polymer fibers this refers to the inter-chain distance in highly aligned P-sheets (e.g., nylon, amyloids, spider dragline MaSpi and silkworm silks, strained collagen, titin rich myofibrils, and some synthetic polymers) [260, 261], a-helices (e.g., intermediate filaments, actin fibers), and P-turns (e.g. silk MaSp2, elastin and some synthesized polymer fibers) [118]. The ratio ro/1o depends on the actual lattice structure present in the material and on the specific make-up of its substructure, as well as other effects such as orientation and temperature. In view of the generality of this analysis and the purpose of obtaining an order of magnitude estimate, this is considered negligible. Table 4 I Process zone (cohesive zone) for the generalized LJ potential. n m I r 12 8 0.4-9.2 105 The predictions are in agreement with several earlier reports. For soft elastic materials, Hui et al. estimated the size of the process zone to be approximately 1 nm [262]. Similarly, Porter et al. derived the process zone size 10 for silk materials to be around 4 nm [263]. Keten et al. investigated the fracture behavior of hydrogen bond clusters, found in many secondary structural elements of (semi)crystalline polymer fibers, leading to a process zone size of approximately 1 nm [197]. Generally, this analysis holds not only for polymeric fibers but also for metals or ceramics. In sum, the calculation based on the LJ potential results in exceedingly small sizes on the order of a few nanometers. Consequently, this suggests a low fracture toughness (as K-1fi), in contrast to the experimental results, which will be addressed in the next section. 4.2.3 Fracture Length Scales in Silk Fibers How is it possible that polymer materials, like spider silk, show dramatically higher fracture toughnesses than predicted by the preceding atomistic analysis, with a remarkable mismatch by a factor of 1,000? This contradiction can be explained by Equation 4.9. Considering the purely elastic nature of the potential up to failure, the ratio between yield strength ay and fracture strength af = oU of a LJ material is approximately one, i.e. af ~ ay (or close to it) [264]. In order to achieve a very large process zone size the yield stress and fracture strength must be decoupled. As shown in chapter 4.2.2.2, this cannot be achieved in a homogeneous material and therefore the existence of a heterogeneous material microstructure is critical. Spider silk's heterogeneity is caused by a composite arrangement of Alanine-rich nanocrystals within a Glycine-rich semiamorphous phase, see chapter 3. Other well-known examples that incorporate such comnlec microstructural architectures are wood, bone, glassy sponges, nacre, and tendons, to name a few [1]. 106 In such structures, the characteristic length scale ro is not a material constant, but depends on the length scale of observation [265]. Figure 31 depicts this concept in the context of spider dragline silk. On the lowest hierarchical level, the scale of the atomic bonds, ro and l are small. Locally, the maximum stress is then the theoretical stress of the perfect crystal ot. Therefore, a strong and tough fiber requires a nanoscale substructure that has dimensions of the process zone size in order to be robust at larger scales. In spider silk, nanocrystals constitute this nanoconfined substructure with dimensions of 2-4 nanometers, see chapter 4.1.5. On the next hierarchical level, the P-sheet crystals form a structure that can be understood as a lattice with spacing of ro ~ 10 nm, the distance between the crystals (ro changes with the length scale of observation). This composite is also confined to the size of the fibril, such that ro and 10 are again of the same order of magnitude. 107 Process Zone Size 10 50nm 1nm M ET c ro 1pm ro Stress transfer through lattice of confined size Purely elastic r Fibril sliding and delocalization (soft and hard phases) Nanoconfined Fibril Hydrogen Bonds Fiber Semi-amorphous Phase and Nanoconfined Crystals 0.2 nm 10 nm Characteristic Length Scale ro 150 nm Figure 31 1 Length scales and toughening mechanisms in spider dragline silk. At the lowest hierarchical level (the scale of the atomic bonds) the maximum stress is the theoretical bond stress O-th, and ro as well as 1o are small (in the order of nanometers). The nanocrystal is extremely robust because it is geometrically confined to the size of the plastic zone. At the next hierarchical level, the beta-sheet crystals form a structure that can be understood as a lattice with spacing of ro ~ 10 nm (the distance between the crystals). The intrinsic strength of the lower scale feature - here the crystal and amorphous phase - is scaled up to the next scale. Paired with the unfolding of the semi-amorphous protein domains, the process zone size is then on the order of 20 -150 nm, the size of the fibrils. Through hierarchical assembly, i.e., the weak binding of many layers of flaw-tolerant fibrils to fibers, the material induces further toughening mechanisms (fibril sliding and delocalization, inducing a process zone of 1 pm) and maintains its toughness at micrometer dimensions. Figure adapted from Reference [1001. The intrinsic strength of the lower scale feature - here the crystal and amorphous phase - is scaled up to the next larger scale. Paired with the unfolding of the semi-amorphous protein domains, the process zone size is then of the order of 20-150 nm, the size of the fibrils [56]. This is in agreement with the estimate of the process zone size 10 /ro ~ [0.5,16]. 4.2.4 Continuum Fracture Mechanics Analysis In order to estimate the length scales of the process zone size in a continuum sense, appropriate boundary conditions to the crack problem in a fiber must be selected [85]. Notably, the Griffith crack condition is not an appropriate model for a flawed fiber, because of the confined boundary. Figure 32 shows three 108 boundary conditions that reflect the influence of intrinsic defects such as cavities and surface cracks, e.g. on a polymer fiber. (ii)(i) (i) D D a 2a Figure 32 | Three typical boundary conditions for a fiber under tension. (i) Cylinder with circumferential crack. (ii) Cylinder with inclusion. (iii) 2D tensile specimen with surface crack. Figure adapted from Reference [100]. cracks with a/D = 0.05 are given by K,(i) = 1.144 aTV, K,(ji) = 0.6368 aV , For the three cases the stress intensity factors for a typical value of macroscale and KJ,(ii) = 1.1473 cirV. Generally, case (i) and case (ii) display a very similar stress intensification behavior (for a/D < 0.3). To find the process zone size for this continuum analysis, Equation 2.19 from chapter 2.2.1, which is a particular form of Equation 4.9 in chapter 4.2.2, is used. Here, it is important to know that Equation 2.19 and Equation 4.9 differ only by a factor for a linear elastic material. Equation 2.19 can be rewritten in a more general form that does not entail the crack length as a specific parameter. Rather, it is written in terms of the material parameter K, (the stress intensification factor). The size of the plastic zone for a linear elastic material under mode I loading is then more generally given by 10 2mI~r \Uy) =4.22 ' *0 109 where m is a parameter between 1 (plain stress) and 3 (plane strain). From experimental data for spider silk fibrils, the relation between the 'yield stress', i.e. the stress at which the unfolding of the protein domains begins (a sort of plasticity mechanism), and the maximum tensile strength is acf/Cy ~ 4 [37]. At the failure point of a fiber this yields for plane strain conditions and lo,a,c = 0.22 (uf/cy) 2 a ~ 0.17 D - 700 nm and lo, = 0.068 (f /y) 2 a 0.04 D ~ 200 nm. It is well known that most polymer fibers cannot be considered linear elastic. Either they feature a non-linear 'post-yield' behavior (e.g., spider dragline silk with a stiffening behavior [50] or silkworm silk with a softening behavior [29]) or they are generally nonlinear elastic (e.g., elastin with a pronounced stiffening behavior [266]). Furthermore, fibrous materials often are anisotropic, deform hyperelastically and/or plastically and are subjected to geometric confinement. Accounting for these nonlinearities is possible, but does not influence this orderof-magnitude estimate significantly. Therefore, an estimate using linear elastic fracture mechanics is sufficient. Finally, the process zone size expected for biomaterials is in the order of 10 300 - 1000 nm, 4.23 in agreement with experimental estimates of the process zone sizes for many polymeric materials. 4.3 The Importance of Heterogeneity in Silk Fibers The importance of spider silks heterogeneity in connection with these length scales has been intensively studied using SEM, X-Ray and neutron scattering techniques [41, 55, 267-269]. Through hierarchical assembly (chapter 3), i.e. the weak binding of many layers of flaw-tolerant fibrils to fibers, the material induces further toughening mechanisms (fibril sliding and delocalization, 110 chapter 4.1.2) and maintains its toughness at micrometer dimensions. This shows that the resilience of materials is greatly enhanced through hierarchical structure originating at the nanoscale, as deformation and damage processes become translated to larger scales [230, 231]. A similar setup can be found in the cellulose fibers in wood, where an arrangement of nanocrystals forms nanometer-sized microfibrils that are densely packed into a lattice like structure [270]. Also collagen (e.g., in bone and mussel threads), keratin based materials, chitinprotein fibers and their derivatives contain highly repetitive patterns on several length scales that can be interpreted as a lattice with spacing ro [271-273]. The concept is schematically shown in Figure 33 for a general case. The fiber becomes robust to flaws at all length scales and does not fail in a brittle manner, when part of a larger-scale hierarchical structure. This is achieved through the confinement of each microstructure to the length scale of the process zone size (called r* in Figure 33), governed by 10/ro ~ [0.5,16], where ro is the characteristic size of the length scale under observation. Hierarchy Level 0 Level 1 Leve1 2 Building Block Assembly M (m MWe ri= rono:r* seeen son 0 r2 =rlnlr2* 0 000 .. r3 =r 2 n 2 <r3* Figure 33 1 Schematic picture of the hierarchical build-up of materials, where at each level the building blocks are repeated n times such that the total length is confined to r*. At each hierarchical scale, the stress concentrations become delocalized. This is achieved through the confinement of each microstructure to the length scale of the process zone size lo/ro = [0.5,16], where ro is the characteristic size of the building block. The fiber becomes robust to flaws at all these length scales and does not fail in a brittle manner, when part of a larger-scale hierarchical structure. The resilience of materials is greatly enhanced through hierarchical structuring from the nanoscale upwards, as deformation and damage processes are translated to larger scales. Figure adapted from Reference [100]. 111 In spider silk, the onset of failure in the early deformation stages of the weakly bonded semi-amorphous phase controls the nonlinear softening after the elastic regime (chapter 1.2.2). This constitutes a yield mechanism where c-y ~ (1/4)uf [37, 50]. A general observation of great importance is that spider silk features a very small yield stress and at the same time, a very large fracture stress. This leads to a very large process zone size on the order of 300 nm to 1,000 nm (see chapter 4.2.4). The key to these considerations is that the particular nonlinear stress-strain relationship in silk fibers originates from a hierarchical arrangement of distinct components, starting at the nanoscale. An overview of various mechanisms at different hierarchical levels of silk fibers is shown in Table 5. 112 larger Table 5 1 Summary of key structures and associated mechanisms of upscaling from the atomistic to scales. Table adapted from Reference [47]. 1p-sheet I Sci nanocrystals I I Critical P-sheet crystal size between 2-4 nm allows for robust and shear-dominated deformation, enabled by cooperative action of clusters of nm 1-Lhnndc Fibers Bundling of several fibrils into fibers where each fibril is in homogeneous deformation state; akin to concept of structure splitting mechanism where bundles of fibrils are assembled into fibers to enhance the overall mechanical pm 4.4 Conclusion In this chapter, the existence of an intrinsic mismatch between the length-scales involved in the fracture mechanics of biological materials is shown and a simple model to connect the interatomic potential to the fracture toughness of 113 hierarchical materials is derived. Without considering heterogeneous structures, the size of the process zone derived from the atomistic scale is only a few nanometers, constituting a mismatch by up to factor 1,000. By incorporating a hierarchy of structures, each confined to a certain critical length-scale, one can explain the process zone size observed in experimental observations and link it directly to an interatomic potential. Beyond the specific case of silk, the strategy to achieve large process zones is ubiquitous in many natural materials, where strong nanoconfined constituents (in fibers: crystallized fibrils, in composites: platelets) are bound by a weak matrix (in fibers: weak chain interactions, in composites: a weak polymer phase). The scaling law (Equation 4.9) for the strength of a material and also the related experimental and theoretical analysis are universal throughout all length scales. What matters is the interpretation of the parameters, which differs between the nano- and the macroscale. The nanostructure is fundamental to the mechanisms that transfer the strength from atomic bonds to the macroscopic fiber. This paradigm could provide an answer to the longstanding question how natural fibers scale up the nanoscale strength to the experimentally observed high strength, extensibility, and robustness. By controlling parameters such as the stiffness at the scale of molecular bonds, it is possible to map nanoscale features to the macroscale, enabling a synergistic interplay of effects at different length scales. Through the regulation of deformation mechanisms of biopolymer fibers at the nanoscale by chemical engineering it will become possible to tune the mechanical performance, and specifically the failure characteristics, to engineering requirements [274-277]. This insight provides a path towards new material designs by embracing heterogeneous structures. 114 5 Supercontraction - Silk's Interaction with Water The research and review presented in this chapter will be published in: * T. Giesa, R. Schuetz, A. Masic, P. Fratzl, M. J. Buehler, Molecular Origin of Supercontraction in Spider Dragline Silk Revealed by Simulation and Experiment. In submission, 2015. All experiments presented in this chapter were performed by Roman Schuetz and Dr. Admir Masic at the Max Planck Institute for Colloids and Interfaces in Potsdam, Germany. In this chapter, the following questions regarding silk supercontraction are answered: " What is silk supercontraction? * Which parts of silk's heterogeneous nanostructure is responsible for the supercontraction mechanism? " Are there specific residues in the silk sequence that can be identified as key players? " How can the supercontraction mechanism be suppressed or modified? Research strategy: Supercontraction is the phenomenon where dragline silk shrinks by up to 50% when immersed in water, and if the fiber is constrained this will generate a tensile stress called supercontraction stress. The molecular origin of dragline silk supercontraction is explored using a full-atomistic model and molecular 115 dynamics combined with in situ Raman spectroscopy and mechanical testing in a humidity controlled chamber. The experimental platform can monitor the extent of supercontraction and molecular interactions simultaneously, whereas molecular dynamics simulations provide a detailed view on the thermodynamics of the material and the behavior of individual residues. A genetic engineering strategy to alter silk's behavior to industrial requirements is proposed. The most important parts of the silk amorphous structure that control supercontraction are identified using a hydrogen bond analysis and a rotamer analysis. Informed mutations to the core sequence of N. Clavipes dragline silk are tested that reduce or even reverse the supercontraction mechanism. This study demonstrates the importance of a combined experimental and computational approach for genetic engineering and innovative materials design, not only for silk. 5.1 Background Spider dragline silk has evolved over millions of years to develop finely tuned mechanical properties to serve specific functions, including the ability to change its material properties upon external signals [28, 278, 279]. The structure of Nephila Clavipes MaSpi dragline silk is described in detail in chapter 3.1.2. Water has the ability to fundamentally reorganize silk's molecular structure and can cause dramatic changes in mechanical properties and physical characteristics [280-284]. Immersion in water typically results in the reduction of the fiber's stiffness by up to an order of magnitude, and noticeable improvement in breaking elongation [282, 285-289]. At high humidity, some spider dragline silks will shrink by up to 50%, a phenomenon known as supercontraction [290, 291]. N. Clavipes dragline silk reversibly shrinks 15-20%, and if the fiber is constrained it will generate a tensile stress [41, 282, 292-296]. There is an ongoing discussion on whether supercontraction is an evolutionary advantage [296, 297], or a 116 constraint [298]. It is an essential feature of the spinning process, since the wet elastomeric silk can be processed easier [291]. 5.1.1 Mechanism of Supercontraction While there are numerous studies on supercontraction, the exact mechanism behind it has not yet been revealed [299]. It has been suggested that since the Psheet crystals are hydrophobic, they do not undergo significant structural changes when hydrated, so the origin of the supercontraction phenomenon is likely to be located in the semi-amorphous phase only [288, 300]. Above a critical hydration level (-70%), water molecules intrude the H-bond network between strands in the amorphous structure and allow them to reorganize into a less ordered, more coiled, lower energy state [294, 296, 298]. Also, their orientation relative to the bulk fiber decreases [296]. Concurrently, the orientation of the disordered and Glycine-rich linker regions (GGX motif) decreases [298]. The response of silk to water indicates that the dry fiber is frozen into a glassy state that is partially extended. Exposed to water releases the glassy state and the wetted silk turns into an elastomer [300-303]. Using nuclear magnetic resonance Yang et al. linked the supercontraction process to the highly conserved YGGLGSQGAGR block in the silk sequence [284]. They identified Leucine (Leu, L) as potential key residue of the supercontraction effect, while noting the proximity of Tyrosine (Tyr, Y) and Arginine (Arg, R). 5.1.2 Combined Simulation and Experimental Approach Figure 34 shows the heterogeneous hierarchical structure of dragline silk (chapter 3.1.2) together with the experimental setup (Figure 34e) used to measure the in situ supercontraction process of N. Clavipes silk. After increasing the humidity, the strain in the fiber decreases under isostatic conditions. The molecular dynamics study consists of two experiments, in both silk is 117 represented by 15 repeats of the main silk sequence, Figure 34a. This constitutes the representative unit of silk, with a stable P-sheet crystal and two independent amorphous phases. The first experiment considers a model of spider dragline silk protein MaSpi, equilibrated in a water box by Replica Exchange Molecular dynamics and explicit water simulation [304]. To measure supercontraction, the effect of removing the solvent on the molecular structure and shape, hydrogen bonding, and entropy of the P-sheet crystal and semi-amorphous phase of MaSpl is evaluated. By comparison to Raman spectroscopic results of wet and dry silk, the residues most active in the supercontraction process are identified. In the second experiment, point mutations on the core sequence shown in Figure 34 are performed, where the identified residues are replaced by residues with shorter side chains. The effect on the level of supercontraction and stability of the structure is investigated (Figure 34f). 118 0 a I CO .0 's- 0r ynm 0 2 '0 Wildypedr 9gIne silk sequence GGAGQGGYGGLGSQGAGRGGLGGQGAG GGAGQGGYGGLGSQGAGRGGLGGQ Sequence modification to suppress supercontraction 4 GGAGQGGFGGLGAQGAGLGGLGGQGAG GGAGQGGFGGLGAQGAGLGGLGGQ Figure 34 1 Nephila Clavipes dragline silk nanostructure and supercontraction mechanism- Bridging from experiments to modeling. Supercontraction is the shrinking of silk in water in comparison to its dry state (by up to 50% depending on the silk, around 15% in Nephila Clavipes silk). (a) In the full-atomistic molecular dynamics simulation silk is represented by a unit of silk, with a stable 1-sheet crystal and two amorphous phases. The amorphous phase is believed to be responsible for the supercontraction mechanism. (b)-(d) Silk assembles in nanofibrils of size 20-150 nanometers. Hundreds of fibrils form dragline fibers of micrometer size. The spider spins the strong dragline silk as structural support for its webs and as lifeline for escape. (e) Measurement of the supercontraction process on dragline silk fibers in a humidity chamber using tensile testing and in situ Raman spectroscopy. (f) In this multiscale approach the macroscale supercontraction effect is linked to nanoscale changes in the structure. Mutations to the core sequence can be proposed to suppress the supercontraction effect. (d) courtesy of Charles J. Sharp. Figure composition, courtesy of Dr. James Weaver, Harvard University. 5.1.3 Molecular Dynamics Setup MD simulations are performed using a model of N. clavipes MaSpi dragline silk, predicted from REMD and equilibration in explicit solvent [279, 304, 305]. The 15-strand sample of MaSpi including a crystal and two amorphous phases is further equilibrated in an explicit water box for 30 ns without holonomic constraints and a 0.5 fs timestep. The GROMACS software package with CHARMM27 force field and the Tip3P water model is used for the explicit water simulations of this complex biological molecule. This force field is able to capture electrostatic interactions without chemical reactions [306]. Isobaric-isothermal 119 conditions (1 bar, 300 K) are modeled with charge-neutralizing solvent and 15 mmol sodium chloride. Equilibration is performed with Particle Mesh Ewald (PME) electrostatics, velocity-rescale thermostat and Nose-Hoover barostat. The vacuum model is also simulated for 30 ns in a canonical ensemble. To prevent image interactions, the periodic box wraps the protein by at least 10 A distance. VMD including the STRIDE secondary structure algorithm is used for visualization trajectories and analysis [307]. Hydrogen of protein molecules bonding and and their equilibration hydrogen bond energies are determined by geometric proximity of hydrogen donor and acceptor, using DSSP [133]. For the H-bonds, a 3.0 A cutoff distance and a 300 cutoff angle is used. 5.2 Supercontraction of the silk wildtype 5.2.1 Silk Supercontraction in Simulation and Experiment The graph in Figure 35 shows the typical (macroscopic) in vitro supercontraction process of N. Clavipes dragline silk using the experimental setup schematically shown (also in Figure 34e). After increasing the humidity from 10% to 90% (blue line), the strain in the fiber decreases under isostatic conditions (red line). 120 Humidity Generator To spectrograph Motor |Laser Silk fiber Load cel 0 -- - 90 75 -T560 C 45 ~30 -15 15 20 60 40 Time (min) 80 the humidity (blue Figure 35 I In vitro supercontraction process and experimental setup. After increasing A. Masic. Data credit: Image line). (red line), the strain in the fiber decreases under isostatic conditions Masic. A. courtesy collected by R. Schuetz and A. Masic. Figure Dry silk, at a humidity around 10%, is very stiff and not extensible, and wet silk, at 100% humidity, is much more compliant and extensible [282, 286-289]. Figure 36 shows a tensile experiment sequence of a single N. Clavipes fiber in the elastic region at three different humidity conditions: at 25%RH (black line), 50%RH (blue line) and after the supercontraction at 90%RH (red line). 121 3,0 -- 2,5 - 25% rh 50% rh 90% rh 2,0- E 0 LL ,1,51,0 0,5 0,0 I 10500 ---- ------------- 11000 11500 ----------- 12000 I 12500 I 13000 Length (pm) Figure 36 I Tensile experiment sequence of a single fiber in the elastic region at three different humidity conditions: at 25%RH (black line), 50%RH (blue line) and after the supercontraction at 90%RH (red line). Data collected by R. Schuetz and A. Masic. Figure courtesy A. Masic. In the simulation, supercontraction is measured by the change in the average end-to-end length of the molecule chains as well as the radius of gyration from the molecular dynamics equilibrium trajectory of the silk vacuum and hydrated model, shown schematically in Figure 37. The radius of gyration (weight averaged ellipsoid) reflects the shape of a 3D molecule and is indicated in Figure 37 together with an overlay of snapshots of the hydrated and the vacuum structure. For details, see chapter 2.1.4. In the wildtype, a contraction from dry to wet state of 13.2 + 5% (radius of gyration) and a contraction of 8.6 + 1.7% (average end-to-end length) is determined. For the contraction in the axial direction of N. Clavipes dragline silk fibers, agreement between simulation and experiment (13.2 0.2%) is found. The values also reflect other literature results for dragline silk fibers [282]. The contraction does not lead to a significant change in volume, as determined from the radius of gyration in the three axis directions of the fiber, Figure 37. 122 f Simulation Radius Simulation Length Experiment 20 15 0- 10 - -5 Sz LO -20 AL 2 DRY 15- z (axis) -y is Figure 37 | Supercontraction measured from Simulation and Experiment. The strand silk model well as length end-to-end average by measured is supercontraction equilibrated in water and vacuum and as radius of gyration of vacuum versus hydrated model. Results of the simulation in three axis directions are compared with experimental results and good agreement is found for the contraction in the axial direction. 5.2.2 Secondary Structure Change during Supercontraction Figure 38 shows the secondary structure of the N. Clavipes MaSpi silk wildtype sequence determined from Molecular Dynamics Simulation in dry and wet conditions. The -sheet content increases from wet to dry conditions, while there is more disordered (coiled) structure in wet conditions, in agreement with experimental results reported elsewhere [116]. The secondary structure is averaged over 30 ns of the simulation time and determined using the STRIDE algorithm. Raman spectroscopy performed on supercontracted silk fibers suggested a change in secondary structure, specifically the loss of ordered structures including turns, helices, and to a lesser extent, r-sheets, and the gain of disordered random coils [308]. 123 80- iWater iVacuum 70. 60 50S403020' 10 0 Betasheet I Coil Helix Turn Figure 38 1 Secondary Structure of the MaSpl dragline silk wildtype determined from molecular dynamics simulation using the STRIDE algorithm. 5.3 Molecular Origin of Supercontraction 5.3.1 Raman Spectroscopy and Hydrogen Bonding In Figure 39 the Raman spectra of N. Clavipes dragline silk in wet (85% RH) and dry (15% RH) conditions are reported. While slight differences can be easily detected in the analyzed spectral range, the most striking change is observed in the 830 - 860 cm' region associated with vibrations in the Tyrosine (Tyr) side chain (Fermi resonance between the in-plane breathing mode of the phenol ring and an overtone of the out-of-plane deformation mode). The relative intensity of the two bands depends sensitively on the extent of mixing of the two modes, and thus on the hydrogen bonding condition of Tyr's phenol side-chain. The relative intensity ratio of two peaks 1860/1830 is up to 2.5 when the OH-group of Tyrosine serves as an acceptor (A) of a strong hydrogen bond (A/D >> 1) and is down to 0.3 when the OH-group serves as a donor (D) of a strong hydrogen bond (A/D << 1) [3091. Figure 39 (left) shows values for 1860/1830 peak ratios 124 determined from the deconvoluted Raman spectra. The line at peak ratio 1.5 is not established and serves only to illustrate a comparison to the simulation. On average, the Tyr residues tend to be both donor and acceptor of hydrogen bonds in dry conditions and turn into strong acceptors in supercontracted wet state. This suggests a specific involvement of Tyr and specifically of the OH-group in the folding and supercontraction of silk. While this phenomenon has been observed [310], the precise implications for the supercontraction process have not yet been investigated. 3 DRY wet D 2.5 -WET Tyr U) 0.5 800 900 1000 1100 Raman shift (1/cm) 0 Raman Spectroscopy Figure 39 I Polarized Raman scattering of N. Clavipes dragline silk in wet (85% RH) and dry (15% RH) 1 conditions. Significant changes are observed in the 830 - 860 cm- region which can be associated with vibrations in the Tyrosine OH-group. The relative intensity ratio of two peaks 1860/1833 suggests a role of Tyrosine and specifically of the OH-group in protein folding and the supercontraction of silk. Experimental data collected by R. Schuetz and A. Masic. Figure (left) courtesy A. Masic. 125 The hydrogen bonding in the 20 nanoseconds molecular dynamics trajectory in fully solvated and vacuum conditions is analyzed. The change in Acceptor/Donor (A/D) ratio, Figure 40, follows the same trend as in Figure 39. In agreement with the Raman experimental results, Tyr changes its donor/acceptor nature of hydrogen bonds when passing form dry to wet conditions. This change is mainly associated with the OH-group, as seen in the subplot of Figure 40. Interestingly, a similar change is prominent in other polar and/or charged side chains such as Arginine (Arg) and Serine (Ser), suggesting a possible contribution to the macroscopic contraction of silk also from these residues. Note, that all these residues are located in the amorphous part of the silk sequence. 1.5 15 0.5 Molecular Dynamics Figure 40 1 The acceptor/Donor (A/D) ratio determined from the hydrogen bonding analysis of the molecular dynamics trajectory is in agreement with the Raman experimental results. Tyrosine tends to be a donor in dry conditions and a donor/acceptor in wet conditions. In the simulation, other polar and/or charged residues in the amorphous part of silk such as Arginine and Serine display a similar behavior. 126 5.3.2 Simulated Infrared Spectrum and Vibrational Density of States When shined on with monochromatic light, silk features characteristic peaks in its spectrum. Using the infrared (IR) and Raman spectrum, the composition, crystallinity and bonding state can be determined. Both spectra are established experimental tools, see chapter 2.1.8. Although quantities like crystallinity and H-bonding can be determined from molecular dynamics trajectories, a direct comparison to the experimental spectra is useful. Figure 41 and Figure 42 show the simulated IR spectrum for the MaSP1 silk nanostructure. The characteristic peaks around 1650 cm-' indicate the amount of P-sheet present. Silk in wet conditions cannot be analyzed using IR spectroscopy. However, Raman spectroscopy can be applied in dry and wet conditions, but is not easily simulated. Figure 43 shows the vibrational density of states of dry and wet silk versus the Raman spectrum (dry silk). While many similarities, especially in the high frequency regime, can be observed, the spectra do not match perfectly. This is due to the light-to-excitation coupling factor C(o>) that has to be determined. More fundamental studies are necessary to connect the Raman spectrum and the VDOS for biomolecules, as to date no literature is available. Especially in the low frequency spectrum, e.g. in the 850 cm-' regime of Tyr's breathing mode, the spectrum is not detailed enough. Longer simulations are necessary to capture those low frequency vibrations. 127 0.0140.012[ Z. 0.01 0.008 h C 0.0060.004 0.002 0 2000 1 uuU Shift [cm~ 1] Figure 41 1 Simulated infrared (IR) spectrum for silk in vacuum (no filter). LI I~ 0. r~4'~. 6.. I--Water 0.01[ I 0.00W1 C 0.006F a I 0.0040.002- 100 I 1. RmR .I 1000 Shift [cm- 1 1500 Figure 42 1 Simulated infrared (IR) spectrum for silk in water (no filter). 128 2000 45 -VDOS -VDOS DRY WET -Raman DRY 40 35 30 20 15 10 5 800 1000 1200 1400 1600 Shift [cm 1] Figure 43 | Simulated VDOS spectrum for silk in water and vacuum versus Raman spectrum. 5.3.3 Energy Balance and Supercontraction Stress To gain a more fundamental understanding of the supercontraction process, the energetics of supercontraction and how a stress can be generated from it is discussed in this chapter. The supercontraction strain esc = AL/LO is the strain generated in an unconstrained fiber when immersed in a humid environment. The supercontraction stress -scis the stress needed to retain a contracting fiber at its original uncontracted length. It can be determined from tensile tests (experiment and simulation) or a free energy balance (simulation). In the following, the energy balance of supercontraction is derived from the Helmholtz Free Energy. To estimate the supercontraction stress, the change in free energy of the system during supercontraction has to be determined. The thermodynamic ensemble during the simulation is NVT, since the vacuum simulation requires a constant volume (otherwise the box collapses). The simulation in full solvation is first equilibrated in NPT, to set the pressure at ibar. 129 For clarity, all described changes are from dry to wet state. For the protein, the change in entropy AS is expected to be positive (since the wet system is more 'disordered'). The change in enthalpy (or in this case internal energy AU) is related to the free energy of solvation, and changes due to the formation of new hydrogen bonds. It consists of an endothermic part (breaking protein-protein and solvent-solvent H-bonds) and a competing exothermic part (forming new solvent-protein H-bonds). Hence, the change in free energy can be decomposed as AA = Awet - Ady ~ AAH- TAS. 5.1 It is assumed that the change in free energy is directly converted into the 1D supercontraction stress. The change in free energy is then related to the supercontraction force by F ~ AA A Al' 5.2 where Al is the change in length due to supercontraction. The stress needed to pull a supercontracted silk fiber into its original (dry) state is given by 4AA irD2 Al - 4(AAH - TAS) D 2A 5.3 ' sc 4F D2 with D as the molecule diameter. 5.3.3.1 Entropic Contribution The entropic term TAS is determined using the methodology described in chapter 2.1.6. In order to reduce the memory required to compute the eigenvalues of the covariance matrix, the system is divided in three parts, the left and right amorphous parts and the crystalline part separating them. This is possible due to the stability of the crystal, since the number of representations of 130 the total system can be computed as the product of the number representations of the subsystems, M = of M 1M 2 M3 . This holds true as long as the particles in the three systems remain distinguishable (which they are in a molecular dynamics simulation). Therefore, the entropy S = k ln(M) can be calculated as the sum of the entropies of the subsystems, S = S1 + S2 + S3. By splitting the calculation into three parts using the different subsets of atoms of the same trajectory, the computation becomes feasible. The entropy is determined from a 5 ns unconstrained simulation in a canonical ensemble (NVT) with 0.5 fs timestep. In the vacuum simulation the entropy is calculated from the trajectory of the aligned molecule. In the solvated simulation the entropy is calculated by removing the solvent from the final trajectory and determining the entropy of the silk molecule. While this neglects the entropic contribution of the water in the structure, the silk molecule before and after desolvation in a canonical ensemble can be compared (since N, V, and T are equal in both systems). The thermostat in the solvated simulation separated water and solute, such that the temperature in the protein was indeed constant at 300 K. Table 6 summarizes the entropies for the silk molecule. The change in entropy, from dry to wet, is AS ~ 2.1 MJ/mol at T = 300 K, a 12.3% increase of absolute entropy. . Table 6 1 Entropy in J/molK for vacuum and solvated structure split in three independent parts S 1 , S 2 , S 3 The change in entropy is calculated for the amorphous phase only and the value in brackets gives the change for the entire structure. Entropy [J/molK] S, Vacuum S2 S3 S, Water S2 S3 Difference Quasiharmonic Schlitter 28,272 7,060 30,696 7,920 29,931 30,739 8,361 34,299 7,035 (7,476) 27,477 31,041 7,495 32,008 7,300 (7,735) 131 5.3.3.2 Enthalpic Contribution The enthalpic term AAH is approximately the change in internal energy, in this case the solvation energy. This can be shown by the following approximation. In - the present isothermal system the change in Helmholtz Free Energy dA = dU TdS and the Gibbs Energy dG = dH - TdS = dU + d(pV) - TdS Free are approximately identical: 0(d(pV)) dp dV : 10 N/mm 2 (150 nm 45 nm 45 nm) 3 - 10-30 kJ 1.83 10-6 kj/mol «TdS. Here, dp is approximated with the change in pressure from vacuum to atmospheric conditions (i.e., 1 bar). The change in volume is approximated as the system volume of the silk unit cell. This yields an upper bound and shows that the term d(pV) is several orders of magnitude smaller than the other terms. Therefore, in this case, dG ; dA holds. AAH is determined with the methodology described in chapter 2.1.7. The silk molecule in this study contains 15 chains and approximately 10,000 atoms. With the water molecules this amounts to approximately 150,000 atoms. Therefore, it is not possible to decouple the degrees of freedom, which is needed for the Amethod. The solvent accessible surface area has AAH = 1.25 + 0.05 MJ/mol free energy of solvation associated with it (GROMACS). The core sequence of silk is a slightly hydrophobic material [311], therefore the enthalpic term is endothermic (of opposite sign as the entropic term, but smaller) and reduces the stress needed to reverse supercontraction. The internal energy part that actively contributes to the supercontraction effect is the change in H-bonding within the silk molecule, which then alters the shape of the molecule. For comparison, the number of Hbonds within the protein (solute-solute) in dry and wet state is calculated. On average, 90 to 100 additional H-bonds are formed in the dry structure (with a 132 typical bonding energy between 4-20 kJ/mol). This strongly contributes to the free energy of solvation. 5.3.3.3 Supercontraction Stress The entropic part of the supercontraction stress cent needed to transition from the supercontracted to the un-contracted state is then determined by cent ~ 4 TAS/(wr Ar D 2 ), where Ar is the change in length (-13%), and D is the diameter of the molecule (- 4 nm). This yields an entropic supercontraction stress of Uent = 64.5 + 10.9 MPa. With TAS ~ 2.1 MI/mol and AAH ~ 1.25 MJ/mol follows usc = 37.4 + 12.4 MPa, 5.4 where Al ~ 3 nm and D ~ 4 nm. Figure 44 shows the supercontraction strain versus stress of silk dragline silk fibers from simulation and experiment. Specifically, the simulation data points in Figure 44 are determined in two different ways. The data point at zero stress is obtained directly from the change of shape, as equilibrium simulations are used to determine the size of the silk molecule. The zero-strain stress is the supercontraction stress usc (computed by decomposing the free energy as described above). The change in entropy and enthalpy is calculated, not absolute values. This change in internal energy generates the supercontraction stress. Alternatively, the supercontraction stress is directly estimated from simulation with a tensile test, using the stress at supercontraction strain, the blue dots in Figure 44. This stress is determined as asim ~ 70 MPa (blue line in Figure 44), very close to the entropic stress of cent = 64.5 + 10.9 MPa (red shaded area in Figure 44). This is intuitive, since the tensile test is performed in solvent and the energy related to the desolvation has not yet been considered. 133 - 0.2- o o v 0.15 Experiment (this study) Experiment (Literature) Pulling Simulation Entropic 9 Entropic + U. Z9 C 0 0.1 0.05 0 60 40 20 Supercontraction Stress [MPa 80 Figure 44 I Supercontraction strain versus stress for dragline silk fibers determined from experiment and in simulation. The simulation data points are found by determining the entropy of spider silk protein in similarity The supercontraction. reverse to needed stress the deriving and hydrated and dry conditions, a provides difference size in magnitude of orders the despite simulation and shape of the experimental amorphous the in strong indication that the supercontraction process is mainly driven by entropic effects structure as well as changes in H-bonding. Experimental data collected by R. Schuetz and A. Masic. The shaded area in between the zero-stress and zero-strain value connects the lower and upper values of the standard deviation and hence indicates only a possible pathway between those two states. This pathway is not necessarily linear. Figure 44 also presents experimental data for silk fibers that were generated from tensile tests of dragline silk fibers. The experimental supercontraction stress and strain of N. Clavipes (orb-weaving) is osc ~ 30 MPa and Esc ~ 13 %. This is in agreement with literature results for N. Clavipes dragline silk (a = 38 + 5 MPa) and A. Aurantial A. Diadematus (same family as Nephilae, a = 40 + 4 MPa) [312]. It is notable that these stresses are well below the yield point of spider dragline silk (ca. 150 MPa). The results presented in Figure 44 have a range of important implications. Agreement is found between the experimental results and the theoretical 134 predictions, when including the enthalpic contribution in the free energy balance. 5.3.3.4 Ratio of Enthalpicand EntropicContribution Among the different types of silks of orbweaving spiders, the ratio between supercontraction stress and supercontraction strain is of similar magnitude, 2 - 3 MPa/%. This suggests that the free energy scales directly with the supercontracted length. Therfore, the mechanism behind supercontraction is similar for different types of dragline silk. This claim is supported with supercontraction experiments of C. Salei (non-orbweaving, asc ~ 150 MPa, esc = 0.23), shown in Figure 45 (in similar style as Figure 44). Literature results for a range of orbiculariae(Lsc = 66 + 10 MPa, esc = 0.32 + 0.06) are also plotted [313]. 0 .4 rT' O1 - T 0 N. Clavipes (this study) v N. Edulis (this study) 0.35 a N Edulis (Literature) 0j 0 C. Salei (this study) 0 Orbiculariae (Literature) N. Clavipes (Simulation) 1 S0.25 .0 - -- 0.21 W0.15j . o 0.054 0 0 100 50 Supercontraction Stress [MPaj 150 - .n among species. Figure 45 I The ratio between supercontraction strain and supercontraction stress is similar the mechanism Therefore, This suggests that the free energy scales directly with the supercontracted length. collected by R. data Experimental behind supercontraction is similar for different types of dragline silk. Schuetz and A. Masic. 135 In conclusion, the supercontraction process is predominantly driven by entropic effects in the rearrangements), amorphous part of whereas the the structure (and H-bond related P-sheet crystals remain mainly unaffected. Therefore, it should be possible to control the supercontraction effect by performing a targeted mutation on the key residues that mostly contribute to the change in entropy and internal energy. 5.3.4 Conformational Changes From the hydrogen bonding analysis and the Raman spectrum it can be inferred that polar and/or charged amino acid residues with large side chains such as Tyr and Arg are potential key players in the supercontraction process. This is reasonable considering the polar nature of water interacting with the protein. However, it remains unclear through which type of mechanisms these residues are able to affect the protein structure. Conformational changes in proteins are often linked to changes in dihedral angles, the torsion angles in the residue backbone and the side-chain. Probability distribution for the backbone dihedrals (P and 'T in form of a Ramachandran plot are plotted [314]. They are calculated from the trajectory of the molecular dynamics simulation in wet and dry state. Figure 54 shows the Ramachandran for the entire silk molecule (excluding Glycine) in dry (left) and wet (middle) state as well as the change in dihedral angle distribution from dry to wet state (right). The distributions are normalized as probabilities. The P-sheets present in the dry state turn into mostly random and helical structures in the wet state. 136 ALL No GLY 0.015 150 150 100 100 IOGLY 5050 -L 0 -. 0.05 150 0045 0-04 .035 100 0 0 000 -50 -150O -10 005 so P0 0 -. -50 0 001-.0 o[oz -150 - 0.01 -100 -100-100 -100 n0LY 0 0 100 150 -150 -100 -50 100 0 150 0 -15o-00 -1 -100 -so0 50 100 ISO -0.015 wet (middle) Figure 46 I Ramachandran plot of the entire silk molecule (excluding Glycine) in dry (left) and are distributions The (right). state wet state as well as the change in dihedral angle distribution from dry to helical and random mostly into turn state dry the in normalized as probabilities. The P-sheets present structures in the wet state. A more detailed analysis is possible when the side-chain dihedrals Xi of all residues are analyzed individually. The following graphs and tables show the change in backbone and side-chain dihedrals of the amino acids in the semiamorphous phase of silk. The distribution of the side-chain dihedrals Xi is also plotted. The allowed configurations of an amino acid are called rotamers. They are tabulated in libraries [315, 316] and are associated with specific secondary structures. Therefore, one can infer conformational changes and associated secondary structure changes from the dihedral shifts of individual amino acids. Secondary structures (letter symbols, see Table 7) are assigned to the rotamers present in the silk structure and the configurational shifts are identified. In the tables, the rotamer type for each - combination of dihedrals is shown: + stands for gauche-positive rotamer type, stands for gauche-negativerotamer type, t stands for trans rotamer type. 137 Table 7 | Nomenclature for Secondary Structure Assignments Letter o, g, n e,m Associated Secondary Structure Right-handed helix, 3 10-helix I P-sheet 5.3.4.1 Alanine Figure 47 shows the structure, dihedrals and associated structure shift for Alanine (Ala, A). Agreement with the experimental dihedral value (-135,150) is found [123]. Small changes are observed from dry to wet. ALA 0.015 15' ).09 10( 0.01 D.07 0.005 D.06 54 .05 0 D.04 -5' D.03 D.02 -101 -0.01 D.01 -15 -0.015 Figure 47 1 Changes in backbone dihedrals of Alanine. 138 5.3.4.2 Glycine Figure 48 shows the structure, dihedrals and associated structure shift for Glycine (Gly, G). Agreement with the experimental dihedral value is found [123]. Glycine is extremely flexible, hence the presence of additional dihedral conformations is not unexpected. Changes are observed from dry to wet, but no clear assignment is possible. GLY 0015 0.03 r iso5 GY 1et ISO4LY 0.01 0.026 100 100 so 50 0-O D IDO 0.0150- 0 0,01 -Wo -50 - 000.O5 0.02 0 0 -so-000 -a0 _01 00010.006 -150015 015 -10-100- 5 0 so 100 - -100 -so 0 so 100 So 0 -iso -1 -so 0 so I i NO 0so Figure 48 1 Changes in backbone dihedrals of Glycine. 5.3.4.3 Serine Figure 49 and Table 8 show the structure, dihedrals and associated structure shift for Serine (Ser, S). Serine has a rather small but polar and hydrophilic side-chain that is not very flexible. While Serine features a shift from rotamer 3 to rotamer 1 (from dry to wet), there is no clear association with the changes in secondary structure. Serine is not expected to be important for supercontraction. 139 SER ,-weI a 2 X1 3 -150 -100 -50 0 X, 50 100 150 1 0.1 0.09 0.00 19 D.05 Is D.04 UC 0.03 0,07 0.02 0.06 0.01 0.05 -0.01 -002 0.04 0.03 -10 -15 0.02 -10( 0.01 -15( 0 4p Figure 49 | Changes in backbone and side-chain dihedrals of Serine. 140 -0.03 -0.04 -.005 Table 8 | Rotamers and associated secondary structures of Serine. Rotamer Xi 2 -67' (t) 0, gn x x e,m p e x 5.3.4.4 Tyrosine Figure 50 and Table 9 show the structure, dihedrals and associated structure shift for Tyrosine (Tyr,Y). Tyrosine has a larger polar and hydrophobic side-chain with two degrees of freedom. xjis rather unflexible and does not shift from dry to wet state. X2 (movement of the phenol ring) is associated with shifts from rotamer 3 and rotamer 4 to rotamer 2 (from dry to wet). In the dry state Tyrosine is involved in structures (sheets and helices), whereas in the wet state it tends to form random structures. Due to its hydrophobicity, Tyrosine tends to interact with the solute itself. Here, a clear shift can be observed. Tyrosine could be important for supercontraction. 141 TYR + 2 2,4 1,3 C1 3 3 X2 4 1 Xii A 4) -50 -100 -150 -- 0 X, M*) [ 50 150 100 -100 -150 -50 0 z 2 I' 50 100 [003 I 0.03 005 W.yi 150 0.02 0.045 10.04 -001 0.035 50 0.025 0.02 so -50 -001 .015 -IC -i 150 -100 0.005 A - ! -1500 10 50.0 0 4. 4, Figure 50 1 Changes in backbone and side-chain dihedrals of Tyrosine. Table 9 1 Rotamers and associated secondary structures of Tyrosine. Rotamer Xi 2 65 (+) 98* (-) 4 740 (+) 167'(t) I i X2 Io,g,nI e,m I X x x I I e p Ix I 5.3.4.5 Leucine Figure 51 and Table 10 show the structure, dihedrals and associated structure shift for Leucine (Leu,L). Leucine, like Tyrosine, has a hydrophobic side-chain with two degrees of freedom, but is non-polar. xjis rather unflexible and does not shift from dry to wet state. X2 is associated with a small shift from rotamer 4 142 to rotamer 1 and more importantly from rotamer 6 to rotamer 2,3,4 (from dry to wet). In the dry state Leucine is involved in structures (sheets and helices), whereas in the wet state it tends to form random structures. Since most of these rotamers are involved in the formation of various structures, there is no clear assignment possible. -wet LEU - 1,3,6 2 2,3,4,5 46 5.'V XX a. 6 4 5 10 -50 -100 150 0 x, [M 50 100 150 -150 LEU ciy -100 0 -50 x 2[1 so 150 100 0.03 15 005 50 002 10 0.04 0,01 60 003 0 9. 0 0.025 005 0.026 -0 -5 -. 01 0.0 -10 -0.02 0.006 -is M -100 -150 -0.03 10 00 Figure 51 1Changes in backbone and side-chain dihedrals of Leucine. 143 Table 10 | Rotamers and associated secondary structures of Leucine. o,g, n e, m Rotamer X1 X2 2 -177" (t) 65" (+) 4 -143* (t) -148"/-1720 (t 6 -770(- 5.3.4.6 x X e X xx x -54"(-) P x Glutamine Figure 52 and Table 11 show the structure, dihedrals and associated structure shift for Glutamine (Gln, Q). Glutamine is polar and hydrophilic and has a large side-chain with three degrees of freedom. Xjis rather unflexible and does not shift from dry to wet state. X2 is associated with a small shift from rotamer 1,2,5 to rotamer 4,6 (from dry to wet). X3 is associated with a small shift from rotamer 1,2,3,4,5 to rotamer 1,5,7 (from dry to wet). Due to the flexibility of Glutamine, especially in X4, no clear assignments are possible. 144 GLN 2 q X 5 5 -2,4,7 1,2,3,4 -1,2,-+ -- 0 50 1,5,7 1,2,5,6 3,7 ISO -100 -50 ll 1,2,3,4,5 +-1,5 1,3,6 100 150 -150 X, M1 -100 -50 4,6 0 Z211 50 -150 150 100 -100-50 0 50 100 10 Y31 0.015 0.05 151 0.045 0,04 15 0.0t 101 0.035 51 0.03 S 0.005 0 0025 0,02 -5 -0m 0.015 -1CH -10 001 005 -0.01 0 -0015 Figure 52 1 Changes in backbone and side-chain dihedrals of Glutamine. 145 Table 11 1 Rotamers and associated secondary structures of Glutamine. Rotamer X1 X2 2 -1740 (t) 177*0t o, , n X3 -110* (Ng9) e,m P xX 65' (Og -) O*(Nt), 62*(0g+) 4 -176* (t) 69- (+) -76- (0g-) 110 (Nt) 6 -73' (-) 81*-(+) 760 (Ng +) x x x X 5.3.4.7 Arginine Figure 53 and Table 12 show the structure, dihedrals and associated structure shift for Arginine (Arg, R). Arginine is polar and the only (positively) charged residue in the core sequence of MaSpi and has a large side-chain with four degrees of freedom. X1 is rather unflexible and does not shift from dry to wet state. X2 is associated with a significant shift from rotamer 1,2,4 to rotamer 5,6 (from dry to wet). X3 is associated with a small shift from rotamer 2,3,4 to rotamer 5 (from dry to wet). X4 is associated with a significant shift from rotamer 1,2,3 to rotamer 6 (from dry to wet). Similar to Tyrosine, Arginine tends to shift towards more random structures. This is important for the supercontraction mechanism since the random structures increase the entropy which in turn leads to the fiber contraction. 146 ARG / 2,5 X3 -I w,3 4,6 X1 1 X2 -150 50 -100 0 50 100 150 xM - -- Wet DT 34 2,3,4,5 + 6 1,5,6 5,6 3 3,4 1,2,3,5 , 1,2,4 3,4,5 5 53,4 53,4 50 -100 -50 0 50 X,f1*1X, 100 150 150 -100 -0 0 50 100 -150 150 -100 -50 0 7. ll 50 100 150 0.05 0.05 0.04 0.045 0.03 004 002 0.035 0.03 I. 0025 0.02 0.01 9. 0.015 0.01 0005 -1 0 Figure 53 | Changes in backbone and side-chain dihedrals of Arginine. 147 0 -am2 .,ot -003 -is( -0.04 -0.05 Table 12 | Rotamers and associated secondary structures of Arginine. Rotamer Xi X3 X2 X4 IiIo,g,n e,ml p e x 68' (+) 4 -1770 (t) 6 1 460 (+) j 80 (+) 1760 (t), -820(-), 1800 (t), - x 68-(-) 700 (+) lxI 850(+) -150 0 (t) x I x x 5.3.5 Hydrogen Bonding of Tyrosine and Arginine Prominent changes are found in the X 2 -angle of Tyr, shown in Figure 54. The X2angle describes the torsion angle of the aromatic ring that is coplanar with the hydroxyl group of Tyr. From wet to dry conditions, a peak shift from -90* to -10' as well as a symmetric shift from 90* to 1700 is observed. As seen in the previous chapter, Arg's and Tyr's shift in the side-chain dihedral angles is specifically associated with a secondary structure transition from sheet-like in the dry state to coiled or helical structure in the wet state leading to a contraction in the wet state. 148 -- Wet Dry -- 2 *X21 -150 -100 -50 0 50 100 150 X2 [" Figure 54 I Dihedral side chain angle distribution of Tyr determined from simulation. The X2 -angle describes the torsion angle of the aromatic ring that is coplanar with the hydroxyl group of Tyr. From dry to wet conditions, a peak shift from -90* to -10* as well as a shift from 90 to 170* is observed. Figure 55 illustrates these transitions with two detailed snapshots from the molecular dynamics simulation showing the local environment of Tyr associated with the peaks (1) at -10' and the peaks (2) at ~170* peaks in Figure 54. The identical residues at a similar simulation time in both dry and wet conditions are shown. Three of the 15 polypeptide chains are visible (blue, red and green), while Tyr and its hydrogen bond partners are highlighted. Peak (1) in dry conditions is associated with the intermolecular hydrogen bonding of Tyr's OH-group with the Gly residue of an adjacent chain. The H-bond is formed between hydrogen of Tyr and the C=O oxygen of Gly that also forms the protein backbone. In this case, Tyr acts as a hydrogen bond donor. In the wet state, Tyr both accepts hydrogen bonds from mobile water within the structure as well as donates Hbonds through intramolecular interactions with the hydrophobic Gly, associated with peaks (2). 149 H-bond with water 1: Intermolecular H-bond 2: TYR TYR H2 SGLY GLY DRY WET IntramolecularR H-bond Figure 55 1 Two detailed snapshots from the molecular dynamics simulation showing the Tyr local environment. The same residues at a similar simulation time are shown in dry and wet conditions. Three (out of 15) polypeptide chains are visible (blue, red and green), while Tyr and its hydrogen bond partner are highlighted. The peak at -10* and 170* (1) in the dry state is associated with the intermolecular hydrogen bonding of Tyrosine's OH group with the Gly of the adjacent chain. The hydrogen bond is formed through hydrogen of the Tyr and oxygen of the Gly involved also in forming protein backbone. In this case, Tyr acts as a hydrogen bonds donor. In the wet state, Tyr both accepts hydrogen bond from mobile water within the structure as well as donates H-bonds through intramolecular interactions with the hydrophobic Glycine (2). Similar results are found for the peaks in the X 4 -angle of Arg, Figure 56, where intra-molecular interaction is replaced by interaction with water. In the dry state, Arginine is a strong H-bond donor, mainly donating through its NH 2 groups to Glycines that are part of adjacent chains, peak (1) at -130*. In water, Arginine becomes both donor and acceptor, where H-bonds are formed with mobile water, peak (2) at 180*. This is illustrated in Figure 57. 150 Wet --Dry 2 2 1 -150 -100 0 -50 X 50 100 150 [01 Figure 56 1 Dihedral side chain angle distribution of Arg determined from simulation. The X4 -angle is the last torsion angle in the long side chain of Arginine where three possible H-bonding sites are present. From dry to wet state, a peak shift from -130* to 180* is observed. 1: Intermolecular H-bond 2: H-bond with water H2 0 ARG ARG GDLY WET DRY - Figure 57 | Two detailed snapshots from the molecular dynamics simulation showing the Arg local environment. The same residues at a similar simulation time are shown in dry and wet state. Two (out of 15) polypeptide chains are visible (blue and red), while Arg and its hydrogen bond partner are highlighted. The peak at -130* (1) in the dry state is associated with the intermolecular hydrogen bonding of Arginine's NH 2 group with Glycine of an adjacent chain. In the wet state, Arg donates hydrogen bonds to mobile water within the structure (2). 151 5.3.6 Key Residues for Supercontraction The findings presented in the previous chapter take the understanding of supercontraction to a new stage. On one hand, the large mobile side chains of Arg and Tyr lead to a significant increase in entropy during the supercontraction process, thus shrinking the molecule. On the other hand, the changes in Hbonding (and associated rotamer configuration indicated by the side-chain dihedrals) relate to the finding that the supercontraction process is additionally controlled by enthalpic effects. The entropic term is so dominant that even at the freezing point of water, MaSp 1 silk still uptakes water, although it contains many hydrophobic components. In dry conditions, the chains are stabilized through intra-molecular interactions and the fiber compacts in radial direction, while the chains elongate in axial direction. The formation of short P-sheets in the non-crystalline part, especially in the GGX motifs (X = Arg, Ala, and Leu) increases this effect. Interestingly, many of the newly formed P-sheets start or end with a Tyr residue. In the wet state, mobile water disturbs these interactions and the chains, especially the long side-chains of Tyr and Arg fold onto themselves. The abundance of hydrophobic Gly in the amorphous silk phase drives the large side-chains of Tyr and Arg to form H-bonds. This marks the fundamental trigger for supercontraction in silk and highlights the importance of the sequence of specific amino acid residues in the amorphous phase. In summary, the potential key players for supercontraction are: Tyrosine (polar, hydrophobic, uncharged), Glutamine (polar, hydrophilic, uncharged), and Arginine (polar, hydrophilic, charged). As a polar and charged amino acid, it can be assumed that Arginine is most active in the supercontraction mechanism. As a polar and hydrophobic (structure building) amino acid, it can be assumed that Tyrosine is also active in the supercontraction mechanism. Furthermore, there is evidence from the Raman spectrum supercontraction mechanism. 152 that Tyrosine is involved in the The following structure mutations are proposed: Tyrosine is substituted with its counterpart without hydroxyl group, PhenylAlanine (Phe, F). Arginine is substituted with an uncharged smaller amino acid, e.g. Leucine (Leu, L). As control experiment, Serine is substituted with Alanine (Ala, A) and no change in supercontraction is expected. 5.4 Controlling Supercontraction Control of the dynamics in the presence or absence of solvent is crucial for the design of polymer materials [284]. In this context, genetic engineering and synthetic chemistry are offering pathways to design materials on demand if appropriate modifications to the sequence can be proposed [317]. In the previous chapters, the crucial role of polar/charged amino acid residues in the supercontraction mechanism was demonstrated. Through simulations, the effect of the substitution of these key residues with their apolar equivalents can now be evaluated. The sequence of N. Clavipes dragline silk is altered and residue mutations on Tyr, Arg, and Ser are performed, as illustrated in Figure 34f. This is especially of interest in view of the technical application of spider silk, where contraction may not be a desirable effect. Three additional sequences are equilibrated using molecular dynamics simulations in solvent and in vacuum. In the first mutation experiment, Tyr is replaced by Phe (similar molecular structure, but no hydroxyl group on the aromatic ring). In the second mutation experiment, Arg (positively charged amino acid with long side chain) is substituted with Leu and in a third mutation experiment, Tyr is replaced by Phe, Arg by Leu and Ser by Ala. As before, the contraction/expansion (by radius of gyration and average end-to-end length) is measured. Furthermore, the P-sheet content of the mutated sequences is evaluated (using equilibrium simulations) and compared to the wildtype silk. 153 While the wildtype silk shows a strong contraction (-15%), a mutation of Tyr indeed leads almost to the suppression of the contraction, Figure 58. The effect of the Arg mutation is even more significant, leading to an expansion in the wet state. Similarly, mutation of all three residues yields a strong expansion in the wet state (and hence a predicted axial expansion of the fiber). The effect of the Serine mutation is small, which can be explained with its comparatively small side-chain and small changes in side-chain dihedral angles. These results provide further evidence for the fundamental role of Tyr and Arg in the supercontraction mechanism through polar and charged side-chain group that can be related to entropic effects (large side chains), but also to internal energy contributions (less hydrogen bonding in the wet state). In all of the mutated sequences, the P-sheet crystals remain intact (and the P-sheet content approximately constant, Figure 59) suggesting that the mechanical properties of the entire structure is unaffected by the point mutations. A computational tensile experiment, shown in the supplemental material, confirms this hypothesis. Figure 60 shows the stress-strain plots determined from MD simulation using steered molecular dynamics in full solvation, with boundary conditions as described in Reference [48]. In the mutated sequence Tyr has been replaced by Phe, Arg by Leu and Ser by Ala. While the stress strain behavior changes slightly, the characteristics are very similar. Thus, it can be concluded, that changing the molecular structure of the amorphous phase, at least concerning the residues investigated, does not alter the high strength and toughness behavior of silk. 154 0 0 Expansion ---------- -------- ------- - -- -- . 0) 10 1 1 1 0 z-Radius V End-to-End Length - 20 .: CD) -10 Contraction -20 Wildtype Tyr -+ Phe Arg - Leu Tyr/Arg/Ser -uPhe/Leu/A4 Figure 58 1 Molecular dynamics simulations of point-mutations of the spider dragline sequence and its effect on supercontraction. Three additional sequences are equilibrated, where polar and/or charged amino acids are substituted by their apolar/uncharged counterparts. The contraction/expansion of the structure is measured by radius of gyration and average end-to-end length. In the first mutation experiment, Tyrosine is replaced by PhenylAlanine (similar molecular structure, but no hydroxyl group on the aromatic ring). In the second mutation experiment, Arginine (positively charged long side chain) is replaced by Leucine. In a third mutation experiment Tyrosine and Arginine are substituted with additional replacement of Serine with Alanine. While the wildtype shows significant contraction (-15%) in the wet state in comparison to the dry state, a replacement of Tyr leads to suppression of the contraction. Mutation with Arginine even leads to an expansion in the wet state. The effect of the Serine replacement is quite small, which can be explained with its comparatively small side-chain. From these results it becomes clear that Tyrosine and Arginine play a crucial role in the supercontraction mechanism through polar and charged side-chain group that can be related to mostly entropic effects. 155 4( 3X Dry * Wet * 5C CL Cn 1OF Cii Tyr Wildtype -+ Arg Phe - Leu T.yr/Arg/Ser *Phe/Leu/Ald P-sheet content Figure 59 1 In all of the mutated structures the P-sheet crystal remains intact (and the unaffected by is structure entire the of properties approximately constant). This suggests that the mechanical the point mutations. I - 160C - 140( - - Wildtype Dry Wildtype Wet Mutated Wet T 1200 1000 ,.- CO 800 600 400 200 0 0 20 40 Strain [%] 60 80 Figure 60 1 Stress-strain curve of wildtype and mutated silk determined with SMD. 156 100 5.5 Conclusion Full-atomistic simulation combined with in situ spectroscopic and tensile experiments provides an excellent experimental platform to monitor the extent of supercontraction and molecular interactions simultaneously. It is a powerful approach to study supercontraction in N. Clavipes dragline silk fibers. Molecular dynamics simulations allow a detailed view on the thermodynamics of the material and the behavior of specific residues or regions within the molecule. Parallelization has enabled computation to investigate environmental changes and mutations with high precision, and has made molecular dynamics a feasible tool for materials engineering. For the first time, these strategies are applied in a synergetic effort to understand the supercontraction mechanism at the molecular level, followed by targeted adjustment of the sequence to tailor properties of the biomaterial. To this end, simulations offer the possibility to explore many options in the material's design space effectively and in a relatively short time. Tyrosine and Arginine are identified as key residues involved in the supercontraction process. They are then mutated with their apolar/uncharged equivalents (PhenylAlanine and Leucine). This results not only in suppression of the supercontraction effect, but even its reversion. By tuning the protein sequence of the amorphous silk phase, a silk material lacking supercontraction while maintaining its extraordinary mechanical properties could be engineered. 157 158 6 Summary and Outlook Silk is a hierarchically structured protein fiber with a high tensile strength and great extensibility, making it one of the toughest materials known. Nephila Clavipes MaSpi, the protein in dragline silk studied in this thesis, contains a sequence of Alanine- and Glycine-rich repeats leading to distinct higher-level structures. Its heterogeneous structure comprises P-sheet crystals embedded in an amorphous matrix. The Alanine-rich region makes up the hydrophobic Psheet crystals, while the semi-amorphous phase associated with the Glycine rich region features a significantly poorer orientation of the strands. The crystals provide strength to the material while the amorphous phase is responsible for elasticity and dissipative mechanisms. Dragline silk has a hierarchical structure where the silk unit cell assembles into nanofibrils of size 20-150 nanometers. Hundreds of fibrils are spun through the spider's spinneret and form a dragline fiber of micrometer size. In chapter 3, the origin of the nanoscale heterogeneity during the Nephila Clavipes dragline silk assembly is investigated. Using molecular dynamics simulations, a shear flow at natural pulling speeds is modelled and the secondary structure transitions as well as the shear stresses in the silk protein chains are determined. Robust results are found where a shear stress of the order of 20-50% of the failure stress induced an c - P-transition in the poly-Alanine region. The results are in agreement with the experimentally determined secondary structure and pulling forces of spider dragline silk. While the transition stress is independent of the chain length, the crystal is stable only in larger configurations. The stability of the assembled p-sheet structure seems to arise from a close proximity of the a-helices in the silk solution. The smallest molecule size that might give rise to a silk-like structure is determined to comprise four to six repeats of the silk sequence. The 159 results emphasize the role of shear in the assembly process of silk and other biopolymers. The determined critical shear values can inform the design of microfluidic devices that attempt to mimic the natural spinning process. Establishing the molecular details of the assembly process can guide the synthesis of bioinspired protein materials. In chapter 4, the heterogeneity of silk fibers on the nanoscale is related to the fracture mechanical properties of the entire fiber. Analytical fracture mechanical arguments are presented to illustrate the relation between fracture strength and toughness and heterogeneity in silk as well as other biopolymers. Nanoconfinement and flaw tolerance are presented as natural strategies to increase the mechanical performance of the entire material system. Confinement refers to the splitting of a macrostructure into small-scale micro- or nanostructures, as observed in brick-and-mortar structures (e.g., bone, nacre) and fibrillation (e.g., silk, collagen). In natural composites and fibers, the hierarchical architecture of the structural components together with a confinement strategy leads to improved load transfer and robustness against failure. Fibrils that are confined to only a few nanometers and then bundled to form fibers maintain their mechanical performance despite the presence of stress concentrations such as cracks, tears, and other flaws. It is shown that the considerations of interatomic interactions alone cannot explain the fracture strength observed in biological fibers. Instead, structures at multiple length-scales must be considered to explain the remarkable mechanical performance and resilience of silk. In chapter 5, the interaction of water with silk's heterogeneous nanostructure is investigated. At high humidity, some spider dragline silks will shrink up to 50%, a phenomenon known as supercontraction. The molecular origin of dragline silk supercontraction is explored using a full-atomistic model and molecular dynamics simulation supported by in situ Raman spectroscopy and mechanical testing performed at the Max Planck Institute in Potsdam, Germany. Tyrosine 160 and Arginine are identified as the key residues in the Nephila Clavipes silk sequence that control supercontraction. They are then substituted with their apolar/uncharged equivalents (PhenylAlanine and Leucine). This results not - only in suppression of the supercontraction effect, but even in the reverse effect expansion in the wet state. By tuning the protein sequence of the amorphous silk phase, a silk material lacking supercontraction while maintaining its extraordinary mechanical properties could be engineered. From the H-bond clustering at the lowest scale to the self-assembly of nanocomposites and subsequent fibrillation, there are still many lessons to be learned before manmade fibers will be able to compete with nature's versatile architectures. Despite the attention and research efforts dedicated to polymers and biocomposites in the past years, there is no consistent explanation for the mechanisms that govern the confined constituents or even for the reason why the microstructures confine during the material's assembly process. Dissection of the constituents of the highly complex natural systems and the determination of the single phase contribution to the overall performance remain major challenges. There is significant evidence showing that the interplay of mechanisms in fibrous biomaterials at all length scales is responsible for their remarkable mechanical properties. It is in the control of this interplay that recent manufacturing techniques cannot compete with natural biopolymers and biocomposites. Interesting novel pathways to manufacture composites have been revealed, e.g., by mixing polymer matrices with highly functional materials such as singlewalled carbon nanotubes. In these materials, the confinement effect due to alignment increases modulus and strength significantly. Furthermore, recent studies have led to the discovery of P-sheet crystals as an important structural component not only in silk, but also in less studied biomaterials such as squid sucker ring teeth. 161 Computational studies (through statistical thermodynamics and mechanics, e.g., molecular dynamics simulations) have made important contributions to the understanding of natural assembly and deformation processes. With increasing computing power, it will be possible to model and analyze larger systems at atomic or quantum resolution to link biology, chemistry, and mechanics. One of the future research opportunities relates to the connection of Raman spectroscopy and the simulation of vibrational properties, briefly addressed in chapter 2.1.8 and 5.3.2. Further research is necessary to identify the isolated (confinement, deformation, supercontraction) mechanisms in a real-world material environment. Experimentally validated modeling and simulation will enable the control of material properties from a bottom-up perspective. The combination of synthesis, controlled processing, and modeling of biopolymer fibers within a unified framework will result in tailored materials with prespecified properties. The research in this thesis presents a step towards the industrial manufacturing of low-cost and environmentally benign functional fibrous materials processed from abundantly available resources. 162 7 Appendix 7.1 Secondary Structure and Shear Stress Trajectories Figure Al-A4 show the secondary structure transition and shear stress during the 10 ns pulling simulation for a = 1 (Figure Al), a = 3 (Figure A2), a = 5 (Figure A3), inter-chain (Figure A4). The upper panel (a) refers to shear boundary condition (i), as seen in Figure 11, chapter 0, and the lower panel (b) refers to shear boundary condition (ii). a 14 n 2500 90 -Heix Betasheet 80 -Coil Turn 80 70 2000 60 1500 1000 40 500 6 8 10 0? 0 2 4 2 4 6 8 6 8 10 Time Ins) Time [ns) b 100 90 -Betnsheet -coil Tur 80 2000 70 1500 60 1000 40 500 2010 '' 2 4 Timne (ns] 6 8 10 0 0 Time [ns] (b) Figure S1 I Secondary Structure Transition and Shear Stress for a = 1 in (a) shear boundary condition (i) shear boundary condition (ii). 163 10 a 2500 100 go. .. 80. -Helixci -- Tum Betasheet 2000 70 1500 60 1000 30 500 2010 00 A et- 2 - S 6 4 Time [ns] 10 2 4 2 4 6 8 10 6 8 10 Time [ns] b 100 2500 -- -90 - Helix Betasheet 80 -- Tr Tum 2000 70 60 100 0~ 40 30 500 20 10 n 4[n Tm2 Timne [ns] 8 10 0 Time [ns] Figure S2 I Secondary Structure Transition and Shear Stress for a = 3 in (a) shear boundary condition (i) (b) shear boundary condition (ii). 164 a 2500 1uu -Helix 90 Betasheet -- Coil Tum 70- 2000 a 1500 60 50 ~1000 I 40 30 500 20 10 o0 2 4 Time (ns] 6 8 10 0 2 4 2 4 Time [ns 6 8 10 6 8 10 b 100 90. -Beteet -Helix coil o -Tum 2000 70 a1500 60 d 50 1000 40 500 30 20 10 00 n 2 6 4 8 10 Time ins) 0 Time [ns] Figure S3 I Secondary Structure Transition and Shear Stress for a = 5 in (a) shear boundary condition (i) (b) shear boundary condition (ii). 165 a 2500 irvi -Helix 90 -Betsheet -Coil 80 -Tum 70 t 2000 1500 LI 60 50 1000 500 6 4 8 10 0 6L 2 4 2 4 Time [ns] Time (ns] 6 8 10 6 8 10 b krivi 100 90 --- Helix - Betashe so-coil 2000 20 ~1500~ 50 40 1000 30 500 10 00 - 2- 4 6 8 10 Time [ns) 0 0 Time [ns] Figure S4 I Secondary Structure Transition and Shear Stress for inter-chain in (a) shear boundary condition (i) (b) shear boundary condition (ii). 166 7.2 Probability for the a-0-Transition Figure B1-B5 show the probabilities associated with the secondary structure transitions for a = 1 (Figure Bi), a = 3 (Figure B2), a = 5 (Figure B3), 2 layers antiparallel (Figure B4), 2 layers parallel (Figure B5). The left part (a) refers to transition from a-helix to -sheet and the right part (b) refers to the transition from 3 1 0-helix/turn to P-sheet. See methods section in main manuscript for details. Tables B1 -B6 show the residues that have a joint probability higher than 10% (for a = 5, 2 layers anti-parallel and 2 layers parallel) 167 a 1.5 a bb 1.5 -REMD 1loop -SMD I loop Equilibration 1loop -Joint Probability I oop -REMID I oop 310 SMD Ioop 310 Equilibration 1 loop 310 -Joint Probability Iloop 310 1 1 e 0.5 JA 20 '0 60 40 MAA# 0.5 "0 80 Figure B1 I Transition probability for a = 1 from (a) a-helix to a 1. 5 - - 20 40 AA[# i 60 A ' 0. 80 P-sheet (b) 3 10 -helix/tum to P-sheet. b 1.51 ,--~- 3loops 310 -REMD SMD 3loops 310 Equilibration 3loops 310 REMID 311oops __SMD 3loops -- - Equilibration 3loops -Joint -Joint Probability 3s Probability 3loops 310 1 Z . 0- 0. 5 I. 0 Figure B2 50 100 150 AA [#] 0.5 _'0 100 50 AA [#] I Transition probability for a = 3 from (a) a-helix to P-sheet (b) 3 10 -helix/tum to P-sheet. 168 150 a 1.5 b -_____ 1.5 -REMD 5loops SMD 5loops Equilibration Sloops -Joint Probeblt 5op Equilibration Sloops 310 -Joint Probability Sloops 310 1- .1 0. 0.5 0 50 100 AA [#1 150 Figure B3 I Transition probability for a = Table B1 I Residues with joint probability I Residue Number 161-162 Table B2 I Residues with joint probability REMD 5loops 310 SMD 5loops 310 -Post 0.5 i 200 5 from (a) a-helix to I 50 100 AA [#] 150 200 P-sheet (b) 3 10 -helix/turn to P-sheet. 10% for a = 5 from a-helix to P-sheet. I Residue Name I I L. OGAGAAAA I I AA I 10% for a = 5 from 3 10 -helix/turn to 1-sheet. Residue Number Residue Name 65 A 103-104 GQ 169 b1.5 a - - REMD 2layers 310 SMD 2iayers 310 Equilibration 2layers 310 Probabilit ayers 310 -Joint REMD 2ayers SMD 2layers Equilibration 21ayers -Joint Probability 2iayers 1 110 0.5 0.5[ 0 50 100 150 200 AA [#] 250 00 300 50 100 150 200 250 Figure B4 I Transition probability for 2 layered structure (anti-parallel) from (a) a-helix to -sheet (b) 31ohelix/turn to 1-sheet. Table B3 I Residues with joint probability 10% for 2 layers anti-parallel from a-helix to 1-sheet. I Residue Name I AAAA 213.216-222 I QGAAAAAA Table B4 I Residues with joint probability > 10% for 2 layers anti-parallel from 3 10 -helix/turn to 1-sheet. I Residue Name 236 170 300 a b 1.5 REMD 2layers parallel SMD 2layers parallel Equilibration 2layers parallel -Joint Probability 2layers parallel, 1.5 R EMD 2layers parallel 310 SMD 2layers parallel 310 Equilibration 2layers parallel 310 -Joint Probability 2layers parallel 310 - 0. 0.5 0.5 00 50 100 150 AA [#J 200 250 00 50 300 100 Figure B5 I Transition probability for 2 layered structure (parallel) from (a) a-helix to helix/turn to P-sheet. Table B5 I Residues with joint probability 10% for 2 layers parallel from a-helix to 94-97 I AAAA 216-221 I GAAAAA 283 IA Table B6 200 150 250 AA [# P-sheet (b) 310- P-sheet. I Residues with joint probability > 10% for 2 layers parallel from 3 10 -helix/ turn to 1-sheet. Residue Number I Residue Name 92. 97-98 I A, AA 248 13% IA NU 171 300 7.3 Nomenclature Symbol Name Unit a Crack length m a Number of silk core sequence repeats A Helmholtz Free Energy J/mol AAH Change in enthalpy J/mol b Platelet width m C(&w) Light-to-excitation coupling factor C Covariance tensor of atomic fluctuations Deq Equivalent radius of a m molecule D Diameter m e Electronic charge C E Young's modulus N/m2 Eb Bulk modulus N/m2 Eb Height of energy barrier J Em Elastic modulus of mineral N/m 2 phase E Elastic tensor N/M 2 EO Initial modulus N/m 2 R Average modulus N/m2 E' Equivalent Young's modulus N/m2 fconf Accessible volume fraction ftahTheoretically allowed fraction of configurations 172 g R(O) Convoluted VDOS intensity G Gibbs Free Energy G Critical strain energy release J/m2 rate h* Platelet height m h* Critical height m h Planck constant Js H Fibril width m H* Enthalpy J/mol H** Critical size, homogeneous m complete m deformation H** Critical size, unfolding IRaman Raman intensity kB Boltzmann constant J/K k Spring stiffness in SMD kcal/mol/ A K Kinetic energy J Kc Stress intensity factor, mode Viii N/m2 Iji,1II 10 Process zone size Al Length m during change m supercontraction L Length/ fiber length m L*, L** Critical overlap length m m Lennard-Jones potential parameter Mi Mass of atom i M Number of configurations 173 kg M Mass tensor n(w) Population kg of vibrational mode ae frequency w n potential Lennard-Jones parameter n Number of fibrils in a fiber n Surface normal vector N Number of atoms in a system Ncr Critical number of bonds NREMD Number of replicas in REMD simulation N/M 3 p System pressure r Distance from crack tip ro Equilibrium bond length m ri Bond length between atom i m and j R Gas constant J/molK S Entropy J/molK t Surface traction vector N/m 2 T Temperature K u Displacement m U Potential energy J Vi Velocity vector of atom i m/s V Volume M3 W Strain energy density J/m 3 Xb Distance between equilibrated m state and transition state acr Ratio between 174 end-to-end length and persistence length / Geometry parameter, related m to F(() Ys Adhesion energy S Cutoff length scale N/m in the m continuum E Strain Emax Failure strain ESC Supercontraction strain Ratio of crack size to system size (a/H) Coupling parameter (BAR method) Aj Eigenvalues of the radius of m gyration tensor x Off-rate 1/s Xi Side-chain dihedral angle rad 6 Angle rad between reaction pathway and applied load 4, Persistence length v Poisson's ratio Ub Bulk strength N/m 2 07ent Entropic stress N/M 2 af Failure stress N/m 2 UsC Supercontraction stress N/m 2 Uth Failure stress of the perfect N/M 2 m crystal UY N/m 2 Yield stress 175 Far-field stress N/m 2 Stress tensor N/m2 Shear stress N/M 2 Interface shear strength N/m 2 > Frequency 1/s > Lennard-Jones CO, U00 Tf potential parameter (AO Eigenfrequency 1/s Frquency of hydrogen bond 1/s vibrations ' Backbone dihedral angle rad Backbone dihedral angle rad Bond volume M3 176 7.4 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. References Fratzl, P. and R. Weinkamer, Nature's hierarchical materials. Progress in Materials Science, 2007. 52(8): p. 1263-1334. Sanchez, C., H. Arribart, and M.M.G. Guille, Biomimetism and bioinspiration as tools for the design of innovative materials and systems. Nature Materials, 2005. 4(4): p. 277-288. nature based innovation. Biomimetics Bar-Cohen, Y., Biomimetics series2012, Boca Raton: CRC Press. p. Ackbarow, T. and M.J. Buehler, Hierarchical Coexistence of Universality and Diversity Controls Robustness and Multi-Functionality in Protein Materials. Journal of Computational and Theoretical Nanoscience, 2008. 5: p. 11931204. Buehler, M.J., Tu(r)ning weakness to strength. Nano Today, 2010. 5(5): p. 379-383. Garcia, A., D. Sen, and M. Buehler, Hierarchical Silica Nanostructures Inspired by Diatom Algae Yield Superior Deformability, Toughness, and Strength. Metallurgical and Materials Transactions A: p. 1-9. Cranford, S.W., et al., Nonlinear material behaviour of spider silk yields robust webs. Nature, 2012. 482(7383): p. 72-76. Aizenberg, J., et al., Skeleton of Euplectella sp.: Structural hierarchy from the nanoscale to the macroscale. Science, 2005. 309(5732): p. 275-278. Currey, J.D., Materials science - Hierarchiesin biomineral structures. Science, 2005. 309(5732): p. 253-254. Fritsch, A. and C. Hellmich, 'Universal' microstructural patterns in cortical and trabecular, extracellular and extravascular bone materials: Micromechanicsbased prediction of anisotropicelasticity. Journal of Theoretical Biology, 2007. 244(4): p. 597-620. Hofstetter, K., C. Hellmich, and J. Eberhardsteiner, Development and experimental validation of a continuum micromechanics model for the elasticity of Wood. European Journal of Mechanics a-Solids, 2005. 24(6): p. 1030-1053. Kroger, N., Prescribing diatom morphology: toward genetic engineering of biological nanomaterials. Current Opinion in Chemical Biology, 2007. 11(6): p. 662-669. Eisenberg, D., et al., Proteinfunction in the post-genomic era. Nature, 2000. 405(6788): p. 823-826. Marcotte, E.M., et al., A combined algorithm for genome-wide prediction of proteinfunction. Nature, 1999. 402: p. 83-86. Marcotte, E.M., Detecting protein function and protein-protein interactions from genome sequences. Science, 1999. 285: p. 751-753. Strohman, R.C., The coming Kuhnian revolution in biology - Commentary. Nature Biotechnology, 1997. 15(3): p. 194-&. 177 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. Nurse, P., Reductionism and explanation in cell biology. Limits Of Reductionism In Biology1998: John Wiley and Sons. Buehler, M.J. and Y.C. Yung, Deformation and failure of protein materials in physiologically extreme conditions and disease. Nature Materials, 2009. 8(3): p. 175-188. Cranford, S.W. and M.J. Buehler, Materiomics: biological protein materials, from nano to macro. Nanotechnology, Science and Applications, 2010. 3: p. 127-148. Launey, M.E., M.J. Buehler, and R.O. Ritchie, On the Mechanistic Origins of Toughness in Bone. Annual Review of Materials Research, 2010. 40: p. 2553. Allison, J.R., et al., Current Computer Modeling Cannot Explain Why Two Highly Similar Sequences Fold Into Different Structures. Biochemistry, 2011. 50(47 DOI: 10.1021/bi2015663). Ghosh, S.K., ed. Self-healing Materials. 2009, Wiley- VCH Verlag: Weinheim. Chen, P.Y., et al., Structure and mechanical properties of selected biological materials. Journal of the Mechanical Behavior of Biomedical Materials, 2008. 1(3): p. 208-226. Ritchie, R.O., The conflicts between strength and toughness. Nature Materials, 2011. 10: p. 817-822. Denny, M., Physical-Propertiesof Spiders Silk and Their Role in Design of OrbWebs. Journal of Experimental Biology, 1976. 65(2): p. 483-506. Hayashi, C.Y. and R.V. Lewis, Evidence from flagelliform silk cDNA for the structural basis of elasticity and modular nature of spider silks. Journal of Molecular Biology, 1998. 275(5): p. 773-784. Hu, X., et al., Molecular mechanisms of spider silk. Cellular and Molecular Life Sciences, 2006. 63(17): p. 1986-1999. Kaplan, D.L. and e. al, Silk polymers : materials science and biotechnology, Chapter 1. ACS symposium series1994, Washington, DC: American Chemical Society. 370. Omenetto, F.G. and D.L. Kaplan, New Opportunitiesfor an Ancient Material. Science, 2010. 329(5991): p. 528-531. Vollrath, F., Silk Evolution Untangled. Nature, 2010. 466(7304): p. 319-319. Huemmerich, D., et al., Novel assembly properties of recombinant spider dragline silk proteins. Current Biology, 2004. 14(22): p. 2070-2074. Foo, C.W.P., et al., Solution behavior of synthetic silk peptides and modified recombinantsilk proteins. Applied Physics a-Materials Science & Processing, 2006. 82(2): p. 193-203. Gosline, J.M., et al., The mechanical design of spider silks: From fibroin sequence to mechanical function. Journal of Experimental Biology, 1999. 202(23): p. 3295-3303. 178 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. Wegst, U.G.K. and M.F. Ashby, The mechanical efficiency of natural materials. Philosophical Magazine, 2004. 84(21): p. 2167-2181. Lewis, R.V., Spider silk: Ancient ideas for neu biomaterials. Chemical Reviews, 2006. 106(9): p. 3762-3774. Dong, Z.Y., R.V. Lewis, and C.R. Middaugh, Molecular Mechanism of Spider Silk Elasticity. Archives of Biochemistry and Biophysics, 1991. 284(1): p. 5357. Agnarsson, I., M. Kuntner, and T.A. Blackledge, Bioprospecting Finds the Toughest Biological Material: Extraordinary Silk from a Giant Riverine Orb Spider. PloS One, 2010. 5(9). Swanson, B.O., et al., The evolution of complex biomaterial performance: The case of spider silk. Integrative and Comparative Biology, 2009. 49(1): p. 2131. Swanson, B.O., et al., Variation in the material properties of spider dragline silk across species. Applied Physics a-Materials Science & Processing, 2006. 82(2): p. 213-218. Gosline, J.M., M.W. Denny, and M.E. Demont, Spider Silk as Rubber. Nature, 1984. 309(5968): p. 551-552. Vollrath, F. and D.P. Knight, Liquid crystalline spinning of spider silk. Nature, 2001. 410(6828): p. 541-548. Du, N., et al., Structural Origin of the Strain-Hardening of Spider Silk. Advanced Functional Materials, 2011. 21(4): p. 772-778. Giesa, T. and M.J. Buehler, Spidermans Geheimnis. Physik in unserer Zeit, 2013. 44(2): p. 72-79. Nova, A., et al., Molecular and Nanostructural Mechanisms of Deformation, Strength and Toughness of Spider Silk Fibrils. Nano Letters, 2010. 10(7): p. 2626-2634. Grubb, D.T. and L.W. Jelinski, Fiber morphology of spider silk: The effects of tensile deformation. Macromolecules, 1997. 30(10): p. 2860-2867. Gosline, J., et al., Elastic proteins: biological roles and mechanical properties. Philosophical Transactions of the Royal Society of London Series BBiological Sciences, 2002. 357(1418): p. 121-132. Giesa, T., et al., Nanoconfinement of spider silk fibrils begets superior strength, extensibility and toughness. Nano Letters, 2011. 11(11): p. 5038-5046. Keten, S. and M.J. Buehler, Nanostructureand molecular mechanics of spider dragline silk protein assemblies. Journal of the Royal Society Interface, 2010. 7(53): p. 1709-1721. Keten, S., et al., Nanoconfinement controls stiffness, strength and mechanical toughness of beta-sheet crystals in silk. Nature Materials, 2010. 9(4): p. 359367. Vollrath, F. and D. Porter, Spider silk as a model biomaterial. Applied Physics A - Materials Science & Processing, 2006. 82(2): p. 205-212. 179 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. Frische, S., A.B. Maunsbach, and F. Vollrath, Elongate cavities and skin-core structure in Nephila spider silk observed by electron microscopy. Journal of Microscopy-Oxford, 1998. 189: p. 64-70. Thiel, B.L. and C. Viney, Spider major ampullate silk (drag line): Smart composite processing based on imperfect crystals. Journal of MicroscopyOxford, 1997. 185: p. 179-187. Poza, P., et al., Fractographicanalysis of silkworm and spider silk. Engineering Fracture Mechanics, 2002. 69(9): p. 1035-1048. Li, S.F.Y., A.J. Mcghie, and S.L. Tang, New Internal Structure of Spider Dragline Silk Revealed by A tomic-ForceMicroscopy. Biophysical journal, 1994. 66(4): p. 1209-1212. Vollrath, F., et al., Structural organization of spider silk. Proceedings of the Royal Society of London Series B-Biological Sciences, 1996. 263(1367): p. 147-151. Du, N., et al., Design of superiorspider silk: From nanostructure to mechanical properties. Biophysical Journal, 2006. 91(12): p. 4528-4535. Perez-Rigueiro, J., et al., Fracture surfaces and tensile properties of UVirradiated spider silk fibers. Journal of Polymer Science Part B-Polymer Physics, 2007. 45(7): p. 786-793. Murphy, R.B., D.M. Philipp, and R.A. Friesner, A mixed quantum mechanics/molecular mechanics (QA4/A4M) method for large-scale modeling of chemistry in protein environments. Journal of Computational Chemistry, 2000. 21(16): p. 1442-1457. Park, S., X. Yang, and J.G. Saven, Advances in computational protein design. Current Opinion in Structural Biology, 2004. 14(4): p. 487-494. Klepeis, J.L., et al., Long-timescale molecular dynamics simulations of protein structure and function. Current Opinion in Structural Biology, 2009. 19(2): p. 120-127. Alder, B.J. and T.E. Wainwright, Studies in Molecular Dynamics. I. General Method. The Journal of Chemical Physics, 1959. 31(2): p. 459-466. Bernal, J.D., The Bakerian Lecture, 1962. The Structure of Liquids. Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences, 1964. 280(1382): p. 299-322. Hao, M.-H. and H.A. Scheraga, Monte Carlo Simulation of a First-Order Transition for Protein Folding. The Journal of Physical Chemistry, 1994. 98(18): p. 4940-4948. Ratje, A.H., et al., Head swivel on the ribosome facilitates translocation by means of intra-subunittRNA hybrid sites. Nature. 468(7324): p. 713-716. Brooks, B.R., et al., CHARMM: The biomolecular simulation program. Journal of Computational Chemistry, 2009. 30(10): p. 1545-1614. Buehler, M.J., Atomistic modeling of materials failure2008, New York: Springer. 180 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. Carrion-Vazquez, M., et al., Mechanical design of proteins studied by singlemolecule force spectroscopy and protein engineering. Progress in Biophysics and Molecular Biology. 74(1-2): p. 63-91. Qin, Z. and M.J. Buehler, Molecular Dynamics Simulation of the alpha -Helix to beta -Sheet Transition in Coiled Protein Filaments: Evidence for a Critical Filament Length Scale. Physical Review Letters. 104(19): p. 198304. Sotomayor, M. and K. Schulten, Single-Molecule Experiments in Vitro and in Silico. Science, 2007. 316(5828): p. 1144-1148. Sugita, Y. and Y. Okamoto, Replica-exchange molecular dynamics method for protein folding. Chemical Physics Letters, 1999. 314(1-2): p. 141-151. Nadler, W. and U.H.E. Hansmann, Optimized explicit-solvent replica exchange molecular dynamics from scratch. Journal of Physical Chemistry B, 2008. 112(34): p. 10386-10387. Feig, M., J. Karanicolas, and C.L. Brooks, MMTSB Tool Set: enhanced sampling and multiscale modeling methods for applicationsin structural biology. Journal of Molecular Graphics & Modelling, 2004. 22(5): p. 377-395. Levy, R.M., et al., Evaluation of the Configurational Entropy for Proteins Application to Molecular-Dynamics Simulations of an Alpha-Helix. Macromolecules, 1984. 17(7): p. 1370-1374. Schlitter, J., Estimation of Absolute and Relative Entropies of Macromolecules Using the Covariance-Matrix.Chemical Physics Letters, 1993. 215(6): p. 617621. Carlsson, J. and J. Aqvist, Absolute and relative entropies from computer simulation with applications to ligand binding. Journal of Physical Chemistry B, 2005. 109(13): p. 6448-6456. Bennett, C.H., Efficient Estimation of Free-Energy Differences from MonteCarlo Data. Journal of Computational Physics, 1976. 22(2): p. 245-268. Eisenhaber, F., Hydrophobic regions on protein surfaces. Derivation of the solvation energy from their area distribution in crystallographic protein structures. Protein Science, 1996. 5(8): p. 1676-1686. Eisenhaber, F., et al., The Double Cubic Lattice Method - Efficient Approaches to Numerical-Integration of Surface-Area and Volume and to Dot Surface Contouring of Molecular Assemblies. Journal of Computational Chemistry, 1995. 16(3): p. 273-284. Available from: Williams, L. Molecular Interactions. 2014; http: / / ww2.chemistry.gatech.edu/ lw26/ structure/ molecular interactio ns/mol int.html. Barrow, G., Physical Chemistry. 5th Edition ed2008: McGraw-Hill Education (India) Pvt Limited. Humphrey, W., A. Dalke, and K. Schulten, VMD: Visual molecular dynamics. Journal of Molecular Graphics & Modelling, 1996. 14(1): p. 33-38. - 67. 181 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. Crupi, V., et al., Vibrational Density of States and Elastic Propertiesof CrossLinked Polymers: Combining Inelastic Light and Neutron Scattering. Journal of Physical Chemistry B, 2014.118(2): p. 624-633. Meyer, R. and D. Comtesse, Vibrational density of states of silicon nanoparticles.Physical Review B, 2011. 83(1). Griffith, A.A., Phenomena of Rupture and Flow in Solids. Philosophical Transactions of the Royal Society of London, Series A, 1921. 221: p. 163198. Tada, H., P.C. Paris, and G.R. Irwin, The stress analysis of cracks handbook. 3rd ed2000, New York: ASME Press. 677. Salib, I.G., et al., Using Mesoscopic Models to Design Strong and Tough Biomimetic Polymer Networks. Langmuir, 2011. 27(22): p. 13796-13805. Mikos, A.G. and N.A. Peppas, Polymer-Chain Entanglements and BrittleFracture.Journal of Chemical Physics, 1988. 88(2): p. 1337-1342. Miehe, C. and L.-M. Schaenzel, Phase field modeling of fracture in rubbery polymers. Part I: finite elasticity coupled with brittle failure. Journal of the Mechanics and Physics of Solids, 2014. 65: p. 93-113. Meyers, M.A. and K.K. Chawla, Mechanical behavior of materials. 2nd ed2009, Cambridge; New York: Cambridge University Press. xxii, 856 p. Williams, J.G., Fracture mechanics of polymers. Ellis Horwood series in engineering science1984, Chichester, New York: E. Horwood; Halsted Press. 302 p. Camanho, P.P., et al., Modeling the inelastic deformation and fracture of polymer composites - Part II: Smeared crack model. Mechanics of Materials, 2013. 59: p. 36-49. Anderson, T.L., Fracture mechanics : fundamentals and applications. 3rd ed2005, Boca Raton, FL: Taylor & Francis. 621 p. Rice, J.R., A Path Independent Integral and Approximate Analysis of Strain Concentration by Notches and Cracks. Journal of Applied Mechanics, 1968. 35(2): p. 379-&. Eshelby, J.D., Elastic Energy-Momentum Tensor. Journal of Elasticity, 1975. 5(3-4): p. 321-335. Maugin, G.A., Eshelby Stress in Elastoplasticity and Ductile Fracture. International Journal of Plasticity, 1994. 10(4): p. 393-408. Brunig, M., Eshelby stress tensor in large strain anisotropic damage mechanics. International Journal of Mechanical Sciences, 2004. 46(12): p. 1763-1782. Eischen, J.W., Fracture of Nonhomogeneous Materials. International Journal of Fracture, 1987. 34(1): p. 3-22. Dugdale, D.S., Yielding of Steel Sheets Containing Slits. Journal of the Mechanics and Physics of Solids, 1960. 8(2): p. 100-104. Williams, J.G. and H. Hadavinia, Analytical solutions for cohesive zone models. Journal of the Mechanics and Physics of Solids, 2002. 50(4): p. 809825. 182 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. Giesa, T., et al., Mhat's Inside the Box? - Length-Scales that Govern Fracture Processesof Polymer Fibers. Advanced Materials, 2014. 26(3): p. 412-417. Knight, D.P., M.M. Knight, and F. Vollrath, Beta transition and stressinduced phase separation in the spinning of spider dragline silk. International Journal of Biological Macromolecules, 2000. 27(3): p. 205-210. Greving, I., et al., Shear-Induced Self-Assembly of Native Silk Proteins into Fibrils Studied by Atomic Force Microscopy. Biomacromolecules, 2012. 13(3): p. 676-682. Xia, X.X., et al., Native-sized recombinant spider silk protein produced in metabolically engineered Escherichia coli results in a strong fiber. Proceedings of the National Academy of Sciences of the United States of America, 2010. 107(32): p. 14059-14063. Heidebrecht, A., et al., Bionimetic Fibers Made of Recombinant Spidroins with the Same Toughness as Natural Spider Silk. Advanced Functional Materials, 2015. Elices, M., et al., Bioinspired Fibers Follow the Track of Natural Spider Silk. Macromolecules, 2011. 44(5): p. 1166-1176. Liu, X. and K.-Q. Zhang, eds. Silk Fiber - Molecular Formation Mechanism, Structure- Property Relationship and Advanced Applications. Oligomerization of Chemical and Biological Compounds2014. Sponner, A., et al., Characterization of the protein components of Nephila clavipes dragline silk. Biochemistry, 2005. 44(12): p. 4727-4736. Lefevre, T., et al., Diversity of Molecular Transformations Involved in the Formationof Spider Silks. Journal of Molecular Biology, 2011. 405(1): p. 238253. Lefevre, T., et al., Conformation of spider silk proteins in situ in the intact major ampullate gland and in solution. Biomacromolecules, 2007. 8(8): p. 2342-2344. Heim, M., D. Keerl, and T. Scheibel, Spider Silk: From Soluble Protein to Extraordinary Fiber. Angewandte Chemie-International Edition, 2009. 48(20): p. 3584-3596. Romer, L. and T. Scheibel, The elaborate structureof spider silk Structure and function of a naturalhigh performancefiber. Prion, 2008. 2(4): p. 154-161. Askarieh, G., et al., Self-assembly of spider silk proteins is controlled by a pHsensitive relay. Nature, 2010.465(7295): p. 236-U125. Jin, H.J. and D.L. Kaplan, Mechanism of silk processing in insects and spiders. Nature, 2003. 424(6952): p. 1057-1061. Gronau, G., Z. Qin, and M.J. Buehler, Effect of sodium chloride on the structure and stability of spider silk's N-terminal protein domain. Biomaterials Science, 2013.1(3): p. 276-284. Rammensee, S., et al., Assembly mechanism of recombinant spider silk proteins. Proceedings of the National Academy of Sciences of the United States of America, 2008.105(18): p. 6590-6595. 183 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. Lefevre, T., et al., Conformational and orientational transformation of silk proteins in the major ampullate gland of Nephila clavipes spiders. Biomacromolecules, 2008. 9(9): p. 2399-2407. Lefevre, T., M.E. Rousseau, and M. Pezolet, Protein secondary structure and orientation in silk as revealed by Raman spectromicroscopy. Biophysical journal, 2007. 92(8): p. 2885-2895. Jenkins, J.E., et al., Solid-state NMR evidence for elastin-like beta-turn structure in spider dragline silk. Chemical Communications, 2010. 46(36): p. 67146716. Sampath, S., et al., X-ray diffraction study of nanocrystallineand amorphous structure within major and minor ampullate dragline spider silks. Soft Matter, 2012. 8(25): p. 6713-6722. Ma, B.Y. and R. Nussinov, Molecular dynamics simulations of alanine rich beta-sheet oligomers: Insight into amyloid formation. Protein Science, 2002. 11(10): p. 2335-2350. Giesa, T. and M.J. Buehler, Nanoconfinement and the Strength of Biopolymers. Annual Review of Biophysics, 2013. 42: p. 651-673. Michal, C.A. and L.W. Jelinski, Rotational-echo double-resonance in complex biopolymers: a study of Nephila clavipes dragline silk. Journal of Biomolecular Nmr, 1998.12(2): p. 231-241. van Beek, J.D., et al., The molecular structure of spider dragline silk: Folding and orientation of the protein backbone. Proceedings of the National Academy of Sciences of the United States of America, 2002. 99(16): p. 10266-10271. Kummerlen, J., et al., Local structure in spider dragline silk investigated by two-dimensional spin-diffusion nuclear magnetic resonance. Macromolecules, 1996. 29(8): p. 2920-2928. Plaza, G.R., et al., Relationship betw'een microstructure and mechanical properties in spider silk fibers: identificationof two regimes in the microstructural changes. Soft Matter, 2012. 8(22): p. 6015-6026. Paquet-Mercier, F., et al., Evidence by infrared spectroscopy of the presence of two types of beta-sheets in major ampullate spider silk and silkworm silk. Soft Matter, 2013. 9(1): p. 208-215. Rousseau, M.E., et al., Nephila clavipes spider dragline silk microstructure studied by scanning transmission X-ray microscopy. Journal of the American Chemical Society, 2007.129(13): p. 3897-3905. Fossey, S.A. and S. Tripathy, Atomistic modeling of interphases in spider silk fibers. International Journal of Biological Macromolecules, 1999. 24(2-3): p. 119-125. Hayashi, C.Y., N.H. Shipley, and R.V. Lewis, Hypotheses that correlate the sequence, structure, and mechanical properties of spider silk proteins. International Journal of Biological Macromolecules, 1999. 24(2-3): p. 271275. 184 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. Bratzel, G. and M.J. Buehler, Sequence-structurecorrelationsand size effects in silk nanostructure:Poly-Ala repeat of N. clavipes MaSpi is naturally optimized. Journal of the Mechanical Behavior of Biomedical Materials, 2012. 7: p. 3040. Phillips, J.C., et al., Scalable molecular dynamics with NAMD. Journal of Computational Chemistry, 2005. 26(16): p. 1781-1802. Essmann, U., et al., A Smooth Particle Mesh Ewald Method. Journal of Chemical Physics, 1995.103(19): p. 8577-8593. Kabsch, W. and C. Sander, Dictionary of Protein Secondary Structure Pattern-Recognition of Hydrogen-Bonded and Geometrical Features. Biopolymers, 1983. 22(12): p. 2577-2637. Cloutier, I., et al., Solid-state nuclear magnetic resonance (NMR) spectroscopy reveals distinctive protein dynamics in closely related spider silks. Canadian Journal of Chemistry-Revue Canadienne De Chimie, 2011. 89(9): p. 10471054. Lefevre, T., et al., Structure of silk by raman spectromicroscopy: From the spinning glands to the fibers. Biopolymers, 2012. 97(6): p. 322-336. van Beek, J.D., et al., Solid-state NMR determination of the secondary structure of Samia cynthia ricini silk. Nature, 2000. 405(6790): p. 1077-1079. Shao, Z.Z. and F. Vollrath, Materials: Surprising strength of silkworm silk. Nature, 2002. 418(6899): p. 741-741. Ortlepp, C.S. and J.M. Gosline, Consequences of forced silking. Biomacromolecules, 2004. 5(3): p. 727-731. Perez-Rigueiro, J., et al., The effect of spinningforces on spider silk properties. Journal of Experimental Biology, 2005. 208(14): p. 2633-2639. Breslauer, D.N., L.P. Lee, and S.J. Muller, Simulation of Flow in the Silk Gland. Biomacromolecules, 2009. 10(1): p. 49-57. Kojic, N., et al., Ex vivo rheology of spider silk. Journal of Experimental Biology, 2006. 209(21): p. 4355-4362. Humenik, M., M. Magdeburg, and T. Scheibel, Influence of repeat numbers on self-assembly rates of repetitive recombinant spider silk proteins. Journal of Structural Biology, 2014. 186(3): p. 431-437. Granick, S., et al., Macromolecules at surfaces: Research challenges and opportunitiesfrom tribology to biology. Journal of Polymer Science Part BPolymer Physics, 2003. 41(22): p. 2755-2793. Stuart, M.A.C., et al., Emerging applications of stimuli-responsive polymer materials. Nature Materials, 2010. 9(2): p. 101-113. Burman, M., A. Arinstein, and E. Zussman, Freeflight of an oscillated string pendulum as a tool for the mechanical characterizationof an individual polymer nanofiber. Applied Physics Letters, 2008. 93(19): p. 193118. Arinstein, A. and E. Zussman, Electrospun Polymer Nanofibers: Mechanical and Thermodynamic Perspectives. Journal of Polymer Science Part B-Polymer Physics, 2011.49(10): p. 691-707. - 130. 185 147. 148. 149. 150. 151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. Cuenot, S., S. Demoustier-Champagne, and B. Nysten, Elastic modulus of polypyrrole nanotubes. Physical Review Letters, 2000. 85(8): p. 1690-1693. Shin, M.K., et al., Size-dependent elastic modulus of single electroactive polymer nanofibers. Applied Physics Letters, 2006. 89(23). Qiu, W.G., et al., Wet-Spinning of Recombinant Silk-Elastin-Like Protein Polymer Fibers with High Tensile Strength and High Deformability. Biomacromolecules, 2009. 10(3): p. 602-608. De Rosa, I.M., et al., Morphological, thermal and mechanical characterizationof okra (Abelmoschus esculentus) fibres as potential reinforcement in polymer composites. Composites Science and Technology, 2010. 70(1): p. 116-122. Buell, S., K.J. Van Vliet, and G.C. Rutledge, Mechanical Propertiesof Glassy Polyethylene Nanofibers via Molecular Dynamics Simulations. Macromolecules, 2009. 42(13): p. 4887-4895. Liivak, 0., et al., A microfabricatedwet-spinning apparatusto spin fibers of silk proteins. Structure-property correlations. Macromolecules, 1998. 31(9): p. 2947-2951. Giesa, T., N.M. Pugno, and M.J. Buehler, Natural stiffening increases flaw tolerance of biologicalfibers. Physical Review E, 2012. 86(4): p. 041902. Wong, S.C., A. Baji, and S.W. Leng, Effect of fiber diameter on tensile properties of electrospun poly(epsilon-caprolactone). Polymer, 2008. 49(21): p. 4713-4722. Torres, J.M., C.M. Stafford, and B.D. Vogt, Elastic Modulus of Amorphous Polymer Thin Films: Relationship to the Glass Transition Temperature. Acs Nano, 2009. 3(9): p. 2677-2685. Lee, J.H., J.Y. Chung, and C.M. Stafford, Effect of Confinement on Stiffness and Fractureof Thin Amorphous Polymer Films. Acs Macro Letters, 2012. 1(1): p. 122-126. Keten, S. and M.J. Buehler, Geometric confinement governs the rupture strength of H-bond assemblies at a critical length scale. Nano Letters, 2008. 8(2): p. 743-748. Brown, C.P., et al., Rough Fibrils Provide a Toughening Mechanism in Biological Fibers. Acs Nano, 2012. 6(3): p. 1961-1969. Peterlik, H., et al., From brittle to ductile fracture of bone. Nature Materials, 2006. 5(1): p. 52-55. Bansal, A., et al., Quantitative equivalence between polymer nanocomposites and thin polymer films. Nature Materials, 2005. 4(9): p. 693-698. Wang, Y.M., Confinement free energy and chain conformations of homopolymers confined between two repulsive walls. Journal of Chemical Physics, 2004. 121(8): p. 3898-3904. Muller, M., Chain conformations and correlations in thin polymer films: A Monte Carlo study. Journal of Chemical Physics, 2002.116(22): p. 9930-9938. 186 163. 164. 165. 166. 167. 168. 169. 170. 171. 172. 173. 174. 175. 176. 177. 178. Cifra, P. and I. Teraoka, Confined polymer chains in a Theta solvent: A model with polymer-solvent interactions. Macromolecules, 2003. 36(25): p. 96389646. Ok, S., et al., Confinement Effects on Chain Dynamics and Local Chain Order in Entangled Polymer Melts. Macromolecules, 2010. 43(10): p. 4429-4434. Wu, Y.Y., et al., Composite mesostructures by nano-confinement. Nature Materials, 2004. 3(11): p. 816-822. Zhang, S.Q. and M.S. Cheung, Manipulating biopolymer dynamics by anisotropicnanoconfinement. Nano Letters, 2007. 7(11): p. 3438-3442. Shin, K., et al., Enhanced mobility of confined polymers. Nature Materials, 2007. 6(12): p. 961-965. Ostermeir, K., K. Alim, and E. Frey, Confinement induces conformational transition of semiflexible polymer rings to figure eight form. Soft Matter, 2010. 6(15): p. 3467-3471. Zhou, H.X., G.N. Rivas, and A.P. Minton, Macromolecular crowIding and confinement: Biochemical, biophysical, and potential physiological consequences. Annual Review of Biophysics, 2008. 37: p. 375-397. Zhou, H.X. and K.A. Dill, Stabilization of proteins in confined spaces. Biochemistry, 2001. 40(38): p. 11289-11293. Zhou, H.X., Loops, linkages, rings, catenanes, cages, and crowoders: Entropybased strategiesfor stabilizingproteins. Accounts of Chemical Research, 2004. 37(2): p. 123-130. Cheung, M.S., D. Klimov, and D. Thirumalai, Molecular crowvding enhances native state stability and refolding rates of globularproteins. Proceedings of the National Academy of Sciences of the United States of America, 2005. 102(13): p. 4753-4758. Samiotakis, A., P. Wittung-Stafshede, and M.S. Cheung, Folding, stability and shape of proteins in crowded environments: experimental and computational approaches. International Journal of Molecular Sciences, 2009. 10(2): p. 572588. Langhe, D.S., A. Hiltner, and E. Baer, Melt crystallization of syndiotactic polypropylene in nanolayer confinement impacting structure. Polymer, 2011. 52(25): p. 5879-5889. Kline, R.J., M.D. Mcgehee, and M.F. Toney, Highly oriented crystals at the buried interface in polythiophene thin-film transistors. Nature Materials, 2006. 5(3): p. 222-228. Micheletti, D., et al., Effect of nanoconfinement on liquid-crystal polymer chains. Journal of Chemical Physics, 2005.123(22). Ellis, R.J. and A.P. Minton, Cell biology - Join the crozd. Nature, 2003. 425(6953): p. 27-28. Alcoutlabi, M. and G.B. McKenna, Effects of confinement on material behaviour at the nanometre size scale. Journal of Physics-Condensed Matter, 2005. 17(15): p. R461-R524. 187 179. 180. 181. 182. 183. 184. 185. 186. 187. 188. 189. 190. 191. 192. 193. Barrat, J.L., J. Baschnagel, and A. Lyulin, Molecular dynamics simulations of glassy polymers. Soft Matter, 2010. 6(15): p. 3430-3446. Baeumchen, 0., et al., Reduced Glass Transition Temperatures in Thin Polymer Films: Surface Effect or Artifact? Physical Review Letters, 2012. 109(5): p. 055701-1 - 055701-5. Zhang, C., Y.L. Guo, and R.D. Priestley, Glass Transition Temperature of Polymer Nanoparticles under Soft and Hard Confinement. Macromolecules, 2011. 44(10): p. 4001-4006. Rittigstein, P., et al., Model polymer nanocomposites provide an understanding of confinement effects in real nanocomposites. Nature Materials, 2007. 6(4): p. 278-282. Pye, J.E. and C.B. Roth, Two Simultaneous Mechanisms Causing Glass Transition Temperature Reductions in High Molecular Weight Freestanding Polymer Films as Measured by Transmission Ellipsometry. Physical Review Letters, 2011. 107(23). Hudzinskyy, D., et al., Effects of Strong Confinement on the Glass-Transition Temperature in Simulated Atactic Polystyrene Films. Macromolecules, 2011. 44(7): p. 2299-2310. Oh, H. and P.F. Green, Polymer chain dynamics and glass transition in athermal polymer/nanoparticlemixtures. Nature Materials, 2009. 8(2): p. 139143. Pryamitsyn, V. and V. Ganesan, A Comparison of the Dynamical Relaxations in a Model for Glass Transition in Polymer Nanocomposites and Polymer Thin Films. Macromolecules, 2010. 43(13): p. 5851-5862. Yousfi, M., et al., A Novel Method for Studying the Dynamics of Polymers Confined in Spherical Nanoparticles in Nanoblends. Macromolecules, 2009. 42(6): p. 2190-2197. Arinstein, A., et al., Effect of supramolecular structure on polymer nanofibre elasticity. Nature nanotechnology, 2007. 2(1): p. 59-62. Aryal, M., K. Trivedi, and W.C. Hu, Nano-Confinement Induced Chain Alignment in Ordered P3HT Nanostructures Defined by Nanoimprint Lithography. Acs Nano, 2009. 3(10): p. 3085-3090. Ao, Z.M. and S. Li, Temperature- and thickness-dependent elastic moduli of polymer thin films. Nanoscale Research Letters, 2011. 6: p. 243. Evans, C.M., et al., Modulus, Confinement, and Temperature Effects on Surface Capillary Wave Dynamics in Bilayer Polymer Films Near the Glass Transition. Physical Review Letters, 2012. 109(3): p. 038302-1 - 038302-5. Torres, J.M., C.M. Stafford, and B.D. Vogt, Impact of molecular mass on the elastic modulus of thin polystyrene films. Polymer, 2010. 51(18): p. 4211-4217. Evans, E., Probing the relation between force - Lifetime - and chemistry in single molecular bonds. Annual Review of Biophysics and Biomolecular Structure, 2001. 30: p. 105-128. 188 194. 195. 196. 197. 198. 199. 200. 201. 202. 203. 204. 205. 206. 207. 208. 209. Konda, S.S.M., et al., Chemical reactions modulated by mechanical stress: Extended Bell theory. Journal of Chemical Physics, 2011. 135(16). Ackbarow, T., et al., Hierarchies, multiple energy barriers, and robustness govern the fracture mechanics of alpha-helical and beta-sheet protein domains. Proceedings of the National Academy of Sciences, 2007. 104(42): p. 1641016415. Spivak, D.I., et al., Category Theoretic Analysis of Hierarchical Protein Materials and Social Networks. PLoS ONE, 2011. 6(9). Keten, S. and M.J. Buehler, Asymptotic strength limit of hydrogen-bond assemblies in proteins at vanishing pulling rates. Physical Review Letters, 2008. 100(19): p. 198301-1 - 198301-4. Adamcik, J. and R. Mezzenga, Proteins Fibrils from a Polymer Physics Perspective. Macromolecules, 2012. 45(3): p. 1137-1150. Marszalek, P.E., et al., Mechanical unfolding intermediates in titin modules. Nature, 1999. 402(6757): p. 100-103. Ritort, F., Single-molecule experiments in biological physics: methods and applications. Journal of Physics-Condensed Matter, 2006. 18(32): p. R531R583. Binnig, G., C.F. Quate, and C. Gerber, Atomic Force Microscope. Physical Review Letters, 1986. 56(9): p. 930-933. Domke, J. and M. Radmacher, Measuring the elastic propertiesof thin polymer films with the atomicforce microscope. Langmuir, 1998. 14(12): p. 3320-3325. Du, B.Y., et al., Study of elastic modulus and yield strength of polymer thin films using atomicforce microscopy. Langmuir, 2001. 17(11): p. 3286-3291. Neuman, K.C. and A. Nagy, Single-molecule force spectroscopy: optical tweezers, magnetic tweezers and atomic force microscopy. Nature Methods, 2008. 5(6): p. 491-505. Isralewitz, B., M. Gao, and K. Schulten, Steered molecular dynamics and mechanical functions of proteins. Current Opinion in Structural Biology, 2001. 11(2): p. 224-230. Bustamante, C., Z. Bryant, and S.B. Smith, Ten years of tension: singlemolecule DNA mechanics. Nature, 2003. 421(6921): p. 423-427. Sharma, D., et al., Single-molecule force spectroscopy reveals a mechanically stable protein fold and the rational tuning of its mechanical stability. Proceedings of the National Academy of Sciences of the United States of America, 2007.104(22): p. 9278-9283. Shi, Q., et al., Mechanism of Adhesion between Polymer Fibers at Nanoscale Contacts. Langmuir, 2012. 28(10): p. 4663-4671. Persson, B.N.J. and S. Gorb, The effect of surface roughness on the adhesion of elastic plates with application to biological systems. Journal of Chemical Physics, 2003.119(21): p. 11437-11444. 189 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 222. 223. 224. 225. 226. Persson, B.N.J., et al., On the nature of surface roughness wvith application to contact mechanics, sealing, rubber friction and adhesion. Journal of PhysicsCondensed Matter, 2005. 17(1): p. R1-R62. Pugno, N.M., Spiderman gloves. Nano Today, 2008. 3(5-6): p. 35-41. Rowland, H.D., et al., Molecular Confinement Accelerates Deformation of Entangled Polymers During Squeeze Flow. Science, 2008. 322(5902): p. 720724. Si, L., et al., Chain entanglement in thin freestanding polymer films. Physical Review Letters, 2005. 94(12). Fischer, H., Polymer nanocomposites: from fundamental research to specific applications. Materials Science & Engineering C-Biomimetic and Supramolecular Systems, 2003. 23(6-8): p. 763-772. Ji, B.H. and H.J. Gao, Mechanical properties of nanostructure of biological materials. Journal of the Mechanics and Physics of Solids, 2004. 52(9): p. 1963-1990. Tjong, S.C., Structural and mechanical properties of polymer nanocomposites. Materials Science & Engineering R-Reports, 2006. 53(3-4): p. 73-197. Paul, D.R. and L.M. Robeson, Polymer nanotechnology: Nanocomposites. Polymer, 2008. 49(15): p. 3187-3204. Giannelis, E.P., Polymer-layered silicate nanocomposites: Synthesis, properties and applications. Applied Organometallic Chemistry, 1998. 12(10-11): p. 675-680. Gao, H.J., et al., Materials become insensitive to flaws at nanoscale: Lessonsfrom nature. Proceedings of the National Academy of Sciences of the United States of America, 2003. 100(10): p. 5597-5600. He, M.R., et al., Quantifying the defect-dominated size effect offracture strain in single crystalline ZnO nanowires. Journal of Applied Physics, 2011. 109(12). Mishra, V., G.H. Fredrickson, and E.J. Kramer, Effect of Film Thickness and Domain Spacing on Defect Densities in Directed Self-Assembly of Cylindrical Morphology Block Copolymers. Acs Nano, 2012. 6(4): p. 3641-3641. Buehler, M.J., et al., Cracking and adhesion at small scales: atomistic and continuum studies offlaw tolerant nanostructures. Modelling and Simulation in Materials Science and Engineering, 2006.14(5): p. 799-816. Kumar, S., M.A. Haque, and H. Gao, Notch insensitive fracture in nanoscale thin films. Applied Physics Letters, 2009. 94(25): p. 253104. Zhang, T., et al., Flaw Insensitive Fracture in Nanocrystalline Graphene. Nano Letters, 2012. 10.1021/n1301908b. Libonati, F., et al., Fracture mechanics of hydroxyapatite single crystals under geometric confinement. Journal of the Mechanical Behavior of Biomedical Materials, 2012. In submission. Pugno, N.M., New quantized failure criteria: application to nanotubes and nanowires. International Journal of Fracture, 2006. 141(1-2): p. 313-323. 190 227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. Gao, H.J. and S.H. Chen, Flaw) tolerance in a thin strip under tension. Journal of Applied Mechanics-Transactions of the ASME, 2005. 72(5): p. 732-737. Wei, X.D., M. Naraghi, and H.D. Espinosa, Optimal Length Scales Emerging from Shear Load Transfer in Natural Materials: Application to Carbon-Based Nanocomposite Design. Acs Nano, 2012. 6(3): p. 2333-2344. Okumura, K. and P.G. de Gennes, Why is nacre strong? Elastic theory and fracture mechanics for biocomposites tith stratified structures. European Physical Journal E, 2001. 4(1): p. 121-127. Dimas, L.S., T. Giesa, and M.J. Buehler, Coupled continuum and discrete analysis of random heterogeneous materials: elasticity and fracture. Journal of the Mechanics and Physics of Solids, 2013. 63: p. 481-491. Dimas, L.S., et al., Random Bulk Properties of Heterogeneous Rectangular Blocks With Lognormal Young's Modulus: Effective Moduli. Journal of Applied Mechanics-Transactions of the Asme, 2015. 82(1). Knowles, T.P., et al., Role of intermolecular forces in defining material propertiesof protein nanofibrils. Science, 2007. 318(5858): p. 1900-1903. Ackbarow, T., et al., Alpha-Helical Protein Netzorks Are Self-Protective and Flawt-Tolerant. PLoS ONE, 2009. 4(6): p. e6015. Porter, D. and F. Vollrath, Nanoscale toughness of spider silk. Nano Today, 2007. 2(3): p. 6. Gatesy, J., et al., Extreme Diversity, Conservation, and Convergence of Spider Silk FibroinSequences. Science, 2001. 291(5513): p. 2603-2605. De Rosa, I.M., et al., Tensile behavior of New? Zealand flax (Phormium tenax) fibers. Journal of Reinforced Plastics and Composites, 2010. 29(23): p. 34503454. Tan, E.P.S., S.Y. Ng, and C.T. Lim, Tensile testing of a single ultrafine polymericfiber. Biomaterials, 2005. 26(13): p. 1453-1456. Gao, H.J. and H.M. Yao, Shape insensitive optimal adhesion of nanoscale fibrillar structures. Proceedings of the National Academy of Sciences of the United States of America, 2004. 101(21): p. 7851-7856. Guerette, P.A., et al., Nanoconfined beta-Sheets Mechanically Reinforce the Supra-Biomolecular Netwoork of Robust Squid Sucker Ring Teeth. Acs Nano, 2014. 8(7): p. 7170-7179. Porter, D., J. Guan, and F. Vollrath, Spider Silk: Super Material or Thin Fibre? Advanced Materials, 2012. 25(9): p. 1275-1279. Gao, H.J., Application of fracture mechanics concepts to hierarchical biomechanics of bone and bone-like materials. International Journal of Fracture, 2006. 138(1-4): p. 101-137. Bazant, Z.P., Scaling theory for quasibrittle structuralfailure. Proceedings of the National Academy of Sciences of the United States of America, 2004. 101(37): p. 13400-13407. 191 243. 244. 245. 246. 247. 248. 249. 250. 251. 252. 253. 254. 255. 256. 257. 258. Chen, H.B. and J.S. Wu, Understandingthe underlying physics of the essential work offracture on the molecular level. Macromolecules, 2007. 40(12): p. 43224326. Ashby, M.F. and A.L. Greer, Metallic glasses as structural materials. Scripta Materialia, 2006. 54(3): p. 321-326. Pugno, N.M. and R.S. Ruoff, Quantized fracture mechanics. Philosophical Magazine, 2004. 84(27): p. 2829-2845. Kruzic, J.J., et al., Indentation techniques for evaluating the fracture toughness of biomaterials and hard tissues. Journal of the Mechanical Behavior of Biomedical Materials, 2009. 2(4): p. 384-395. Litteken, C.S. and R.H. Dauskardt, Adhesion of polymer thin-films and patterned lines. International Journal of Fracture, 2003. 119(4-2): p. 475-485. Bougaut, 0. and D. Rittel, On crack-tip cooling during dynamic crack initiation. International Journal of Solids and Structures, 2001. 38(15): p. 2517-2532. Wu, J.S. and Y.W. Mai, The essential fracture work concept for toughness measurement of ductile polymers. Polymer Engineering and Science, 1996. 36(18): p. 2275-2288. Evans, A.G. and K.T. Faber, Crack-Growth Resistance of Microcracking Brittle Materials. Journal of the American Ceramic Society, 1984. 67(4): p. 255-260. Bazant, Z.P. and G. Pijaudiercabot, Measurement of CharacteristicLength of Nonlocal Continuum. Journal of Engineering Mechanics-Asce, 1989. 115(4): p. 755-767. Cedolin, L., S.D. Poli, and I. Iori, Experimental-Determinationof the Fracture Process Zone in Concrete. Cement and Concrete Research, 1983. 13(4): p. 557-567. Shah, S.P. and S. Choi, Nondestructive techniques for studying fracture processes in concrete. International Journal of Fracture, 1999. 98(3-4): p. 351359. Nakhodchi, S., et al., Fracture Process Zones in Polygranular Graphite in Bending. Advances in Fracture and Damage Mechanics Ix, 2011. 452-453: p. 93-96. Smith, I. and S. Vasic, Fracture behaviour of softwood. Mechanics of Materials, 2003. 35(8): p. 803-815. Seitz, J.T., The Estimation of Mechanical-Propertiesof Polymersfrom MolecularStructure. Journal of Applied Polymer Science, 1993. 49(8): p. 1331-1351. Grzybowski, A., K. Koperwas, and M. Paluch, Scaling of volumetric data in model systems based on the Lennard-Jones potential. Physical Review E, 2012. 86(3). Ballarini, R., et al., Biological structures mitigate catastrophicfracture through various strategies. International Journal of Fracture, 2005. 135(1-4): p. 187197. 192 259. 260. 261. 262. 263. 264. 265. 266. 267. 268. 269. 270. 271. 272. 273. 274. Kang, Y.K., Which functional form is appropriatefor hydrogen bond of amides? Journal of Physical Chemistry B, 2000. 104(34): p. 8321-8326. Meier, C. and M.E. Welland, Wet-Spinning of Amyloid Protein Nanofibers into Multifunctional High-Performance Biofibers. Biomacromolecules, 2011. 12(10): p. 3453-3459. Kushner, A.M., et al., A Biomimetic Modular Polymer with Tough and Adaptive Properties. Journal of the American Chemical Society, 2009. 131(25): p. 8766-+. Hui, C.Y., et al., Crack blunting and the strength of soft elastic solids. Proceedings of the Royal Society of London Series a-Mathematical Physical and Engineering Sciences, 2003.459(2034): p. 1489-1516. Porter, D. and F. Vollrath, Silk as a Bionimetic Ideal for Structural Polymers. Advanced Materials, 2009. 21(4): p. 487-492. Varnik, F., L. Bocquet, and J.L. Barrat, A study of the static yield stress in a binary Lennard-Jones glass. Journal of Chemical Physics, 2004. 120(6): p. 2788-2801. Cornetti, P., et al., Finitefracture mechanics: A coupled stress and energy failure criterion. Engineering Fracture Mechanics, 2006. 73(14): p. 2021-2033. Lillie, M.A. and J.M. Gosline, The viscoelastic basis for the tensile strength of elastin. International Journal of Biological Macromolecules, 2002. 30(2): p. 119-127. Riekel, C. and F. Vollrath, Spider silk fibre extrusion: combined wide- and small-angle X-ray microdiffraction experiments. International Journal of Biological Macromolecules, 2001. 29(3): p. 203-210. Vollrath, F. and D. Porter, Spider silk as archetypal protein elastomer. Soft Matter, 2006. 2(5): p. 377-385. Sapede, D., et al., Nanofibrillar structure and molecular mobility in spider dragline silk. Macromolecules, 2005. 38(20): p. 8447-8453. Fernandes, A.N., et al., Nanostructureof cellulose microfibrils in spruce wood. Proceedings of the National Academy of Sciences of the United States of America, 2011.108(47): p. E1195-E1203. Meyers, M.A., et al., Biological materials: Structure and mechanical properties. Progress in Materials Science, 2008. 53(1): p. 1-206. Jin, J., et al., A biomimetic composite from solution self-assembly of chitin nanofibers in a silk fibroin matrix Advanced Materials, 2013. 25(32): p. 44824487. Hagenau, A., et al., StructuralAnalysis of ProteinaceousComponents in Byssal Threads of the Mussel Mytilus galloprovincialis. Macromolecular Bioscience, 2009. 9(2): p. 162-168. Leung, K.C.F., et al., Supramolecular self-assembly of dendronized polymers: Reversible control of the polymer architectures through acid-base reactions. Journal of the American Chemical Society, 2006. 128(33): p. 10707-10715. 193 275. 276. 277. 278. 279. 280. 281. 282. 283. 284. 285. 286. 287. 288. 289. 290. 291. Erk, K.A., K.J. Henderson, and K.R. Shull, Strain Stiffening in Synthetic and Biopolymer Networks. Biomacromolecules, 2010. 11(5): p. 1358-1363. Broedersz, C.P. and F.C. MacKintosh, Molecular motors stiffen non-affine semiflexible polymer networks. Soft Matter, 2011. 7(7): p. 3186-3191. Capadona, J.R., et al., Stimuli-responsive polymer nanocomposites inspired by the sea cucumber dermis. Science, 2008. 319(5868): p. 1370-1374. Ko, F.K., EngineeringPropertiesof Spider Silk Fibers, in Natural Fibers, Plastics and Composites, F.T. Wallenberger and N.E. Weston, Editors. 2004, Kluwer Academic Publishers. Keten, S. and M.J. Buehler, Nanostructure and molecular mechanics of spider dragline silk protein assemblies. Journal of the Royal Society Interface, 2010c. 7(53): p. 1709-1721. Agnarsson, I., et al., Spider silk as a novel high performance biomimetic muscle driven by humidity. Journal of Experimental Biology, 2009. 212(13): p. 19891993. Blackledge, T.A., et al., Howv super is supercontraction?Persistentversus cyclic responses to humidity in spider draglinesilk. Journal of Experimental Biology, 2009. 212(13): p. 1980-1988. Guan, J., F. Vollrath, and D. Porter, Two mechanisms for supercontractionin Nephila spider dragline silk. Biomacromolecules, 2011. 12(11): p. 4030-4035. Holland, G.P., R.V. Lewis, and J.L. Yarger, WISE NMR characterizationof nanoscale heterogeneity and mobility in supercontractedNephila clavipes spider dragline silk. Journal of the American Chemical Society, 2004. 126(18): p. 5867-5872. Yang, Z.T., et al., Supercontractionand backbone dynamics in spider silk: C-13 and H-2 NMR studies. Journal of the American Chemical Society, 2000. 122(37): p. 9019-9025. Shao, Z.Z. and F. Vollrath, The effect of solvents on the contraction and mechanical propertiesof spider silk. Polymer, 1999. 40(7): p. 1799-1806. Vollrath, F. and D. Porter, Spider silk as a model biomaterial. Applied Physics A, 2006. 82(2): p. 205-212. Bratzel, G. and M.J. Buehler, Molecularmechanics of silk nanostructuresunder varied mechanical loading. Biopolymers, 2012a. 97(6): p. 408-417. Brown, C.P., et al., The critical role of water in spider silk and its consequence for protein mechanics. Nanoscale, 2011. 3(9): p. 3805-3811. Liu, Y., Z. Shao, and F. Vollrath, Extended wet-spinning can modify spider silk properties.Chemical Communications, 2005b(19): p. 2489-2491. Work, R.W., A Comparative-Study of the Super-Contraction of Major Ampullate Silk Fibers of Orb-Web-Building Spiders (Araneae). Journal of Arachnology, 1981. 9(3): p. 299-308. Vollrath, F., Spider Webs and Silks. Scientific American, 1992. 266(3): p. 7076. 194 292. 293. 294. 295. 296. 297. 298. 299. 300. 301. 302. 303. 304. 305. 306. 307. 308. Guinea, G.V., et al., Self-tightening of spider silk fibers induced by moisture. Polymer, 2003. 44(19): p. 5785-5788. Agnarsson, I., et al., Spider silk as a novel high performance biomimetic muscle driven by humidity. Journal of Experimental Biology, 2009b. 212(13): p. 1990-1994. Blackledge, T.A., et al., How super is supercontraction?Persistent versus cyclic responses to humidity in spider dragline silk. Journal of Experimental Biology, 2009. 212: p. 1981-1989. Vollrath, F., D. Porter, and C. Holland, The Science of Silks. MRS Bulletin, 2013. 38(1): p. 73-80. Agnarsson, I., et al., Supercontractionforces in spider dragline silk depend on hydration rate. Zoology, 2009a. 112(5): p. 325-331. Qin, Z. and M.J. Buehler, Spider silk: webs measure up. Nature Materials, 2013. 12(3): p. 185-187. Liu, Y., Z. Shao, and F. Vollrath, Relationships between supercontractionand mechanical propertiesof spider silk. Nature Materials, 2005a. 4(12): p. 901-905. Elices, M., et al., The hidden link between supercontraction and mechanical behavior of spider silks. Journal of the Mechanical Behavior of Biomedical Materials, 2011. 4(5): p. 658-669. Ene, R., P. Papadopoulos, and F. Kremer, Supercontraction in Nephila spider dragline silk - Relaxation into equilibrium state. Polymer, 2011. 52(26): p. 6056-6060. Gosline, J.M., M.W. Denny, and M.E. DeMont, Spider silk as rubber. Nature, 1984. 309: p. 551-552. Elices, M., et al., The hidden link between supercontraction and mechanical behavior of spider silks. Journal of the Mechanical Behavior of Biomedical Materials, 2011. 4(5): p. 658-669. Grubb, D.T. and L.W. Jelinski, Fiber Morphology of Spider Silk: The Effects of Tensile Deformation. Macromolecules, 1997. 30(96): p. 2860-2867. Bratzel, G. and M.J. Buehler, Sequence-structure correlationsin silk: Poly-Ala repeat of N. clavipes MaSpi is naturally optimized at a critical length scale. Journal of the Mechanical Behavior of Biomedical Materials, 2012b. 7: p. 30-40. Keten, S. and M.J. Buehler, Atomistic model of the spider silk nanostructure. Applied Physics Letters, 2010a. 96(15). Buehler, M.J., S. Keten, and T. Ackbarow, Theoretical and computational hierarchical nanomechanics of protein materials: Deformation and fracture. Progress in Materials Science, 2008. 53(8): p. 1101-1241. Humphrey, W., A. Dalke, and K. Schulten, VMD: Visual Molecular Dynamics. Journal of Molecular Graphics, 1996.14: p. 33-38. Shao, Z.Z., R.J. Young, and F. Vollrath, The effect of solvents on spider silk studied by mechanical testing and single-fibre Raman spectroscopy. 195 309. 310. 311. 312. 313. 314. 315. 316. 317. International Journal of Biological Macromolecules, 1999. 24(2-3): p. 295300. Siamwiza, M.N., et al., Interpretation of Doublet at 850 and 830 Cm-1 in Raman-Spectra of Tyrosyl Residues in Proteins and Certain Model Compounds. Biochemistry, 1975. 14(22): p. 4870-4876. Shao, Z., et al., Analysis of spider silk in native and supercontractedstates using Raman spectroscopy. Polymer, 1999. 40(10): p. 2493-2500. Bini, E., D.P. Knight, and D.L. Kaplan, Mapping domain structures in silks from insects and spiders related to protein assembly. Journal of Molecular Biology, 2004. 335(1): p. 27-40. Savage, K.N., P.A. Guerette, and J.M. Gosline, Supercontraction stress in spider webs. Biomacromolecules, 2004. 5(3): p. 675-679. Boutry, C. and T.A. Blackledge, Evolution of supercontraction in spider silk: structure-function relationship from tarantulas to orb-weavers. Journal of Experimental Biology, 2010. 213(20): p. 3505-3514. Ramakris.C and Ramachan.Gn, Stereochemical Criteria for Polypeptide and Protein Chain Conformations .2. Allowed Conformations for a Pair of Peptide Units. Biophysical journal, 1965. 5(6): p. 909-&. Schrauber, H., F. Eisenhaber, and P. Argos, Rotamers - to Be or Not to Be - an Analysis of Amino-Acid Side-Chain Conformations in Globular-Proteins. Journal of Molecular Biology, 1993. 230(2): p. 592-612. Scouras, A.D. and V. Daggett, The dynameomics rotamer library: Amino acid side chain conformations and dynamics from comprehensive molecular dynamics simulations in water. Protein Science, 2011. 20(2): p. 341-352. Buehler, M.J., Computational and Theoretical Materiomics: Properties of Biological and de novo BioinspiredMaterials. Journal of Computational and Theoretical Nanoscience, 2010. 7(7): p. 1203-1209. 196 7.5 List of Figures Figure 1 1 Complex hierarchical structures found in natural materials. (a) Scanning electron microscopy (SEM) pictures displaying the intricate hierarchical porous silica wall structure of diatoms. Figure adapted from [12], with permission from Elsevier. (b) SEM pictures of the mineralized skeletal system of Eucleptella. The caged structure consists of struts (bundled spicules) which themselves are a ceramic composite with laminated silica layers and organic interlayers. Figure adapted from [8], copyright @ 2005, with permission from the American Association for the Advancement of Science. .................................... 15 Figure 2 1 Hierarchical structures of biological materials such as spider silk and bone. Biological materials are designed bottom-up to overcome fundamental strength limits at the nanoscale. Spider dragline silk, specifically the protein MaSpi, consists of P-sheet nanocrystals embedded in an amorphous phase. These are aligned along the fiber axis to form fibrils of size 20-150 nm. Hundreds of fibrils are spun together in a fiber that eventually forms the frame of an orb web. Compact bone is composed of osteons that surround and protect blood vessels. Osteons have a lamellar structure. Each individual lamella is composed of fibers arranged in geometrical patterns. These fibers are the result of several collagen fibrils, each linked by an organic phase to form fibril arrays. Each array makes up a single collagen fiber. The mineralized collagen fibrils are the basic building blocks of bone. Bone figure adapted from Reference [20], copyright D 2010, Annual Reviews. Web image courtesy of Charles J. Sharp. Silk figure composition courtesy of Dr. James Weaver, Harvard University. ..................... 18 Figure 3 1 Density vs. failure strength of synthetic and natural materials as Ashby plot. Adapted from Reference [34]............................................................... 21 Figure 4 | Silk features a hierarchical structure, where P-sheet crystals play a key role in defining the mechanical properties by providing stiff and orderly crosslinking domains embedded in a semi-amorphous 197 matrix that consists predominantly of less ordered structures. These (-sheet nanocrystals, bonded by means of assemblies of H-bonds, have dimensions of a few nanometers and constitute roughly 15-50% of the silk volume. Adapted from Reference [43]...... 22 Figure 5 1 Stress-strain behavior for a defect-free silk fiber, noting the key transition points between the four regimes marked by molecular events at the molecular scale. The transition from Regime I to Regime II marks the onset of unfolding of the semi-amorphous phase of silk; the transition from Regime II to Regime III marks the onset of stretching of the P-sheet nanocrystal phase. In Regime IV P-sheet nanocrystals fail via a stick-slip mechanism, eventually leading to failure. Figure adapted from Reference [43]........................................ 25 Figure 6 1 Microscopic images of Nephila madagascariensis dragline silk fibers showing the skin-core structure as well as flaws and cavities in the material. The white arrows point in the axial fiber direction, and the red ellipses highlight some of the defects found in the structure. Pictures reprinted from [51], 9 copyright 1998, with permission from John Wiley & Sons, Inc. ......................... 26 Figure 7 1 Approximate length and time scale regimes of the tools for multiscale engineering. Computational tools predict and explain phenomena that are observed experimentally, but are limited to certain regimes due to constraints on computational performance. While mesoscale and continuum modeling (subpanel a) cannot capture atomistic details, they are trained by atomistic results from Density Functional Theory (DFT) and Molecular Dynamics (MD) simulations. They cover the same length scale range as experimental tools (e.g. atomic force microscopy (AFM, subpanel b), optical/ magnetical tweezers, microelectromechanical systems (MEMS, subpanel c) and nano-indentation. The lower part indicates classes or scales of protein materials that can be studied with the respective techniques. Figure reprinted from [18], copyright @ 2009, with permission from the Nature Publishing Group. .................................................. 30 Figure 8 1 Stress variation in the cohesive zone according to the DugdaleBarenblatt model. Adapted from Reference [100]................................................. 198 44 Figure 9 1 Molecular model of the silk fiber assembly process. Silk is processed from spidroin to a solid fiber in the spinning duct under ambient conditions. The dragline silk spidroin consists mainly of MaSpi (studied here) and MaSp2. The sequence is highly repetitive, a ~ 0(100). Shear stresses at the wall together with the removal of water from the protein lead to the formation of a nano-composite having an aligned, fl-sheet rich crystalline phase. A change of ionic conditions during the spinning process is believed to lead to a conformational change in the terminal regions of the silk protein. ........................................................................ 48 Figure 10 1 (a) Model of the silk spidroin in equilibrium. Using Replica Exchange Molecular Dynamics for a total of 1 microsecond the equilibrium structure of the protein is determined for different chain lengths (a = 1,3,5), here denoted as poly-Alanine regions (blue) and remainder of the intervening sequence (red). (b) Model of the silk assembly process. Using Steered Molecular Dynamics at natural pulling speeds a shear flow is modeled and the secondary structure transitions as well as shear stresses in the silk protein chains are determined. Intra- as well as inter-chain interactions are investigated. (c) Model of the final fiber structure. After shearing the assembled structure is simulated in solvent and vacuum to test the stability of the structure as a single-chain or as layered structure, i.e. multiple stacked sheets after shear................................................. 53 Figure 11 1 Two different boundary conditions (i) and (ii), both shear, are tested to investigate the trajectory of secondary structure and shear stress. The part of the sequence colored in blue is the Alanine rich region and the part of the sequence colored in red is the Glycine-rich region. The pulling force is determined with the steered molecular dynamics (SMD) algorithm................. 55 Figure 12 1 Secondary structure transition and shear stress during the assembly of silk. (a) Secondary structure trajectory during the pulling simulation in explicit solvent. The graph shown is one out of four tests for a = 3 (with boundary condition (i)). Starting from the spidroin, a high a-helical and coil content and no P-sheets can be found. During the shear induced assembly all helices and other 199 structures transition into P-sheets or are destroyed in agreement with experimental observations of the processes in the spinning duct. The P-sheet content after pulling for each structure is determined by averaging the P-sheet content for 1 ns around the maximum content as indicated in the figure. (b) Shear stress associated with the secondary structure transition. The transition shear stress is averaged in the same region as that taken to assess the P-sheet content. In all cases, the structure transition happens prior to reaching the strength limit of the material................................................................................... 56 Figure 13 1 P-sheet content versus pulling speed for a set of spring stiffnesses. The observed P-sheet content after shear is insensitive to the simulation p aram eters....................................................................................................................... 60 Figure 14 1 (a) The P-sheet content attained after the shear flow experiment is independent of the chain length and well within the range of experimental observation. (b) The transition shear stress for the silk assembly is calculated to be between 300 and 700 MPa, whereas experimental observations put it between 20-60% of the breaking stress (300-850 MPa). This agrees with the observation that the reorganization of silk requires significant shear stresses. ..................... 62 Figure 15 1 Stability of the silk chains (P-sheet content) after shearing and a further equilibration in explicit solvent. The shorter chains (a = 1, a = 3) cannot retain the P-sheet crystal and the structure returns in its spidroin state; the larger structure a = 5 rem ains relatively stable............................................................... 63 Figure 16 1 Stability in water. Summary of the secondary structure content of the spidroin and the simulated structures after equilibration in explicit solvent. While the shorter chains start to form 310-helices and retain only a low percentage of P-sheets, the larger structure is stable and shows the secondary structure composition of final assembled dragline silk fibers............................. 64 Figure 17 1 Stability in vacuum. Summary of the secondary structure content of the spidroin and the simulated structures after equilibration in vacuum. All chains remain stable independent of the chain length........................................ 200 64 Figure 18 1 Transition probabilities for a = 5 from a-helix to P-sheet. The graph shows the probability for each of the 206 residues of the structure with a = 5 to be in a-helical state after REMD, in P-sheet state after SMD and in in P-sheet state after Equilibration. The dark blue line is the joint probability defined as the product of the three probabilities and indicates the residues that transition from helical to sheet structure with high probability...................................................... 66 Figure 19 1 Transition probabilities for a = 5 from turn/310-helix to P-sheet. Probability for each of the 206 residues of the structure with a = 5 to be in 310helical/turn state (after REMD) and in s-sheet state (after SMD/after E q uilib ration )..................................................................................................................67 Figure 20 1 Structure snapshots (a = 5) with highlighted residues identified from Table 2 of the four transition stages (i - iv), in agreement with .......................... 68 Figure 21 1 P-sheet content of sandwich structures after equilibration in vacuum and solvent. Layered structures are formed from chains with a = 3. Independent of the amount of layers or the orientation 10-15% of the structure stabilize as Psheets in solvent, and 20-30% in vacuum .............................................................. 71 Figure 22 1 Relative elastic modulus and strength of polymer materials as a function of size. (a) In most nanofibers, the relative elastic modulus, as compared with the bulk modulus Eb, increases dramatically at a critical diameter D *. This behavior is explained by the role of spatial confinement on entropy and the dominance of intermolecular interactions in thin nanofibers. Thin polymer films with free surfaces tend to display a decrease in the relative modulus due to the formation of highly mobile surface layers. (b) The alignment of crystallites and the degree of crystallinity in the fiber also improve with smaller diameter, leading to greater strength and toughness, sometimes even exceeding the bulk properties. The data for the fracture strength of thin films show a decrease of the material strength. (The references, critical length scales, and bulk values are summarized in Table 3.) Figure reprinted from Reference [121]........................ 201 77 Figure 23 1 Rupture strength of H-bond clusters and critical sizes of some protein secondary structures. Top: The intrinsic strength limit of H-bonds can be overcome by clusters of three to four H-bonds that then interact synergistically to resist deformation and failure. The H-bond assemblies are loaded in parallel in ahelices, p-sheets, and p-helices. Note that the shear strength curve is derived for a single P-sheet in a pull-out test. The natural load condition for an a-helix is tension, for which an unzipping effect is easily achieved. Bottom: This result explains the cluster size found in natural protein secondary structures (a-helix: N = 3.5; P-helix: N = 5; and P-sheet: N = 2.5 - 8). Figure adapted with permission from [157] Copyright @ 2008 American Chemical Society..............85 Figure 24 1 Intrinsic (plasticity) versus extrinsic (shielding) toughening mechanisms associated with crack extension and R-curve. (a) The illustration shows mutual competition between intrinsic damage mechanisms, which act ahead of the crack tip to promote crack advance and extrinsic crack-tip-shielding mechanisms, which act primarily behind the tip to impede crack advance [24]. Intrinsic toughening results essentially from plasticity and enhances a material's inherent damage resistance; as such it increases both the crack-initiation and crack-growth toughnesses. (b) Toughness behavior of various materials. In many natural materials, it is an order of magnitude tougher than its constituent phases. Figure adapted from [24], copyright 2011, with permission from the Nature Publishing G roup....................................................................................................... 89 Figure 25 1 Confinement and flaw tolerance. The graph shows the concept of flaw tolerance. According to the classical prediction, the strength of a material scales with 1/h or, 1/D respectively. In the case where the strength of the perfect crystal is reached at h = h * (D = D *), the flawed system exhibits no loss of strength. This concept has been experimentally and computationally verified, e.g., for (a) spider dragline silk [47], (b) hydroxyapatite nanocrystals [225], (c) thin metal strips [223], and (d) nanocrystalline graphene [224]. Images in panel (a) reprinted from Reference [47]; images in panel (b) from Reference [225]. 202 Images in panel (c) adapted with permission from Reference [223]. Copyright @ 2009, American Institute of Physics. Image in panel (d) adapted with permission from Reference [224]. Copyright @ 2012 American Chemical Society. Figure adapted with permission from Reference [121]..................................................... 91 Figure 26 1 Critical size of a composite and an adhesion system. Bone-like materials typically consist of fragile, brittle mineral platelets (hydroxyapatite) embedded in protein matrix materials (collagen). (a) The mineral platelets carry a tensile load and the protein transfers carry loads between the platelets via shear. (b) These platelets are confined and optimally arranged to maximize the strength and toughness of the material. (c) Similarly, the adhesion of a spatula on a rigid surface is optimized for a critical diameter D *, e.g., for gecko adhesion [211]. The critical sizes h * and D * determine the point at which the system becomes flawtolerant. The solid line on the left corresponds to the classical fracture mechanics prediction, which breaks down at the length scales at approximately the critical size (on the order of a few nanometers). Figure adapted with permission from R eferen ce [121]................................................................................................................ 92 Figure 27 1 Optimized length scales in a mineral-polymer composite (e.g., nacre). (a) Hierarchical structure of nacre and a schematic of a 2-dimensional continuum model for the composite architecture to predict the critical sizes that maximize the strength of the whole material. (b) Elastic and fracture toughness varying with overlap length normalized by L *. Total elastic strain energy density (squares) maximizes at L = L *, and fracture toughness (circles) exhibits a sudden drop when L > L **. (c) Comparison of overlap lengths for basic building blocks of three natural materials (nacre, tendon, and spider silk) from experimental observation (circles) and model prediction (squares). Figure adapted with permission from Reference [228]. Copyright @ 2012 American Chemical Society. ........................................................................................................................................... 94 Figure 28 1 Robustness of P-sheet nanocrystals as a function of their height. PSheet nanocrystals are especially strong and robust if their height is confined to 203 2-4 nm. This critical dimension is in agreement with experimental results. Figure adapted from Reference [49]. ................................................................................... 96 Figure 29 1 Critical size of spider dragline silk major ampullate spidroin 1 (MaSpi). The graph shows the dependence of the failure strain and failure stress on the fibril size D under various loading conditions (1-4) as well as a direct comparison with experimental results (under the tensile loading condition 1 and the mechanical behavior of a defect-free silk fiber. For decreasing fibril sizes, the perfect material behavior (i.e., ~ 1,400 MPa failure stress and 68% failure strain) is approached and reached at D = D * = 50 + 30 nm. D* is denoted the critical flaw-tolerant size of the fiber. The results show that the high strength and extensibility observed in experimental studies can only be reached by nanoconfinement of fibrils close to D *. Figure adapted from Reference [47] with p erm ission ....................................................................................................................... 98 Figure 30 1 (a) A fiber of diameter D that has no intrinsic flaws. Under tension aO, such a fiber's failure strength af would reach the theoretical strength of the interatomic bonds it consists of, ath. (b) A fiber made of the same bonds without internal structure but containing a flaw of length a would decrease its strength according to Griffith's size scaling as the ratio 10/D and the strength of the fiber become smaller. (c) A possible strategy to maintain the strength of the fiber at the macroscale is to increase the size of the process zone, such that 10 = D. Then, the strength of the fiber will approach the theoretical strength of the internal structure, af = th. Figure adapted from Reference [100].................................... 100 Figure 31 1 Length scales and toughening mechanisms in spider dragline silk. At the lowest hierarchical level (the scale of the atomic bonds) the maximum stress is the theoretical bond stress oth, and rO as well as 10 are small (in the order of nanometers). The nanocrystal is extremely robust because it is geometrically confined to the size of the plastic zone. At the next hierarchical level, the betasheet crystals form a structure that can be understood as a lattice with spacing of rO 10 nm (the distance between the crystals). The intrinsic strength of the 204 lower scale feature - here the crystal and amorphous phase - is scaled up to the next scale. Paired with the unfolding of the semi-amorphous protein domains, the process zone size is then on the order of 20 -150 nm, the size of the fibrils. Through hierarchical assembly, i.e., the weak binding of many layers of flawtolerant fibrils to fibers, the material induces further toughening mechanisms (fibril sliding and delocalization, inducing a process zone of 1 pm) and maintains its toughness at micrometer dimensions. Figure adapted from Reference [100]. ......................................................................................................................................... 1 08 Figure 32 1 Three typical boundary conditions for a fiber under tension. (i) Cylinder with circumferential crack. (ii) Cylinder with inclusion. (iii) 2D tensile specimen with surface crack. Figure adapted from Reference [100]....................109 Figure 33 1 Schematic picture of the hierarchical build-up of materials, where at each level the building blocks are repeated n times such that the total length is confined to r *. At each hierarchical scale, the stress concentrations become delocalized. This is achieved through the confinement of each microstructure to the length scale of the process zone size 10/rO = [0.5,16], where rO is the characteristic size of the building block. The fiber becomes robust to flaws at all these length scales and does not fail in a brittle manner, when part of a largerscale hierarchical structure. The resilience of materials is greatly enhanced through hierarchical structuring from the nanoscale upwards, as deformation and damage processes are translated to larger scales. Figure adapted from R eferen ce [100].............................................................................................................. 111 Figure 34 1 Nephila Clavipes dragline silk nanostructure and supercontraction mechanism- Bridging from experiments to modeling. Supercontraction is the shrinking of silk in water in comparison to its dry state (by up to 50% depending on the silk, around 15% in Nephila Clavipes silk). (a) In the full-atomistic molecular dynamics simulation silk is represented by a unit of silk, with a stable P-sheet crystal and two amorphous phases. The amorphous phase is believed to be responsible for the supercontraction mechanism. (b)-(d) Silk assembles in 205 nanofibrils of size 20-150 nanometers. Hundreds of fibrils form dragline fibers of micrometer size. The spider spins the strong dragline silk as structural support for its webs and as lifeline for escape. (e) Measurement of the supercontraction process on dragline silk fibers in a humidity chamber using tensile testing and in situ Raman spectroscopy. (f) In this multiscale approach the macroscale supercontraction effect is linked to nanoscale changes in the structure. Mutations to the core sequence can be proposed to suppress the supercontraction effect. (d) courtesy of Charles J. Sharp. Figure composition, courtesy of Dr. James Weaver, H arvard University......................................................................................................119 Figure 35 1 In vitro supercontraction process and experimental setup. After increasing the humidity (blue line), the strain in the fiber decreases under isostatic conditions (red line). Image credit: A. Masic. Data collected by R. Schuetz and A. Masic. Figure courtesy A. Masic. ................................................... 121 Figure 36 1 Tensile experiment sequence of a single fiber in the elastic region at three different humidity conditions: at 25%RH (black line), 50%RH (blue line) and after the supercontraction at 90%RH (red line). Data collected by R. Schuetz and A. M asic. Figure courtesy A. M asic...................................................................122 Figure 37 1 Supercontraction measured from Simulation and Experiment. The strand silk model is equilibrated in water and vacuum and supercontraction is measured by average end-to-end length as well as radius of gyration of vacuum versus hydrated model. Results of the simulation in three axis directions are compared with experimental results and good agreement is found for the contraction in the axial direction. .............................................................................. 123 Figure 38 1 Secondary Structure of the MaSp1 dragline silk wildtype determined from molecular dynamics simulation using the STRIDE algorithm.........124 Figure 39 1 Polarized Raman scattering of N. Clavipes dragline silk in wet (85% - RH) and dry (15% RH) conditions. Significant changes are observed in the 830 860 cm - 1 region which can be associated with vibrations in the Tyrosine OHgroup. The relative intensity ratio of two peaks 1860/1833 suggests a role of 206 Tyrosine and specifically of the OH-group in protein folding and the supercontraction of silk. Experimental data collected by R. Schuetz and A. Masic. Figure (left) courtesy A . M asic...................................................................................125 Figure 40 1 The acceptor/Donor (A/D) ratio determined from the hydrogen bonding analysis of the molecular dynamics trajectory is in agreement with the Raman experimental results. Tyrosine tends to be a donor in dry conditions and a donor/acceptor in wet conditions. In the simulation, other polar and/or charged residues in the amorphous part of silk such as Arginine and Serine 126 display a sim ilar behavior. ......................................................................................... Figure 41 | Simulated infrared (IR) spectrum for silk in vacuum (no filter)......128 I Simulated infrared (IR) spectrum for silk in water (no filter). ......... 128 43 I Simulated VDOS spectrum for silk in water and vacuum versus Figure 42 Figure 129 R am an spectru m . ......................................................................................................... Figure 44 1 Supercontraction strain versus stress for dragline silk fibers determined from experiment and simulation. The simulation data points are found by determining the entropy of spider silk protein in hydrated and dry conditions, and deriving the stress needed to reverse supercontraction. The similarity in shape of the experimental and simulation - despite the orders of magnitude in size difference - provides a strong indication that the supercontraction process is mainly driven by entropic effects in the amorphous structure as well as changes in H-bonding. Experimental data collected by R. Schuetz and A . M asic. ................................................................................................. 134 Figure 45 1 The ratio between supercontraction strain and supercontraction stress is similar among species. This suggests that the free energy scales directly with the supercontracted length. Therefore, the mechanism behind supercontraction is similar for different types of dragline silk. Experimental data collected by R. Schuetz and A . M asic........................................................................ 135 Figure 46 1 Ramachandran plot of the entire silk molecule (excluding Glycine) in dry (left) and wet (middle) state as well as the change in dihedral angle 207 distribution from dry to wet state (right). The distributions are normalized as probabilities. The P-sheets present in the dry state turn into mostly random and helical structures in the w et state............................................................................... Figure 47 Figure 48 Figure 49 Figure 50 Figure 51 Figure 52 Figure 53 137 I Changes in backbone dihedrals of Alanine......................................... 138 I Changes in backbone dihedrals of Glycine......................................... 139 I Changes in backbone and side-chain dihedrals of Serine.................140 I Changes in backbone and side-chain dihedrals of Tyrosine............142 I Changes in backbone and side-chain dihedrals of Leucine.............. 143 I Changes in backbone and side-chain dihedrals of Glutamine.........145 I Changes in backbone and side-chain dihedrals of Arginine............147 Figure 54 1 Dihedral side chain angle distribution of Tyr determined from simulation. The X2-angle describes the torsion angle of the aromatic ring that is coplanar with the hydroxyl group of Tyr. From dry to wet conditions, a peak shift from -90* to -10* as well as a shift from 90* to 170* is observed...................149 Figure 55 1 Two detailed snapshots from the molecular dynamics simulation showing the Tyr local environment. The same residues at a similar simulation time are shown in dry and wet conditions. Three (out of 15) polypeptide chains are visible (blue, red and green), while Tyr and its hydrogen bond partner are highlighted. The peak at -10* and 1700 (1) in the dry state is associated with the intermolecular hydrogen bonding of Tyrosine's OH group with the Gly of the adjacent chain. The hydrogen bond is formed through hydrogen of the Tyr and oxygen of the Gly involved also in forming protein backbone. In this case, Tyr acts as a hydrogen bonds donor. In the wet state, Tyr both accepts hydrogen bond from mobile water within the structure as well as donates H-bonds through intramolecular interactions with the hydrophobic Glycine (2).............................150 Figure 56 1 Dihedral side chain angle distribution of Arg determined from simulation. The X4-angle is the last torsion angle in the long side chain of Arginine where three possible H-bonding sites are present. From dry to wet state, a peak shift from -130* to 1800 is observed. ................................................ 208 151 Figure 57 1 Two detailed snapshots from the molecular dynamics simulation showing the Arg local environment. The same residues at a similar simulation time are shown in dry and wet state. Two (out of 15) polypeptide chains are visible (blue and red), while Arg and its hydrogen bond partner are highlighted. The peak at -130* (1) in the dry state is associated with the intermolecular hydrogen bonding of Arginine's NH2-group with Glycine of an adjacent chain. In the wet state, Arg donates hydrogen bonds to mobile water within the structure (2 )....................................................................................................................................1 51 Figure 58 1 Molecular dynamics simulations of point-mutations of the spider dragline sequence and its effect on supercontraction. Three additional sequences are equilibrated, where polar and/or charged amino acids are substituted by their apolar/ uncharged counterparts. The contraction/expansion of the structure is measured by radius of gyration and average end-to-end length. In the first mutation experiment, Tyrosine is replaced by PhenylAlanine (similar molecular structure, but no hydroxyl group on the aromatic ring). In the second mutation experiment, Arginine (positively charged long side chain) is replaced by Leucine. In a third mutation experiment Tyrosine and Arginine are substituted with additional replacement of Serine with Alanine. While the wildtype shows significant contraction (-15%) in the wet state in comparison to the dry state, a replacement of Tyr leads to suppression of the contraction. Mutation with Arginine even leads to an expansion in the wet state. The effect of the Serine replacement is quite small, which can be explained with its comparatively small side-chain. From these results it becomes clear that Tyrosine and Arginine play a crucial role in the supercontraction mechanism through polar and charged sidechain group that can be related to mostly entropic effects....................................155 Figure 59 | In all of the mutated structures the P-sheet crystal remains intact (and the P-sheet content approximately constant). This suggests that the mechanical properties of the entire structure is unaffected by the point mutations..............156 209 Figure 60 1 Stress-strain curve of wildtype and mutated silk determined with S MD ................................................................................................................................ 210 1 56 7.6 List of Tables Table 1 1 Model structures of the stages during the assembly process. The spider silk dope is stored in the abdomen of the spider in globular form. In the simulation, the general form of the simulated peptide/ protein structure is largely independent of the size of the molecule (left hand column). It remains unassembled and predominantly contains helices and turns. During the shearing process (second column) a large part of the structure transitions into P-sheet structures. Subsequently, equilibration of the structure in vacuum or explicit water in the presence of ions, the silk relaxes again (third and fourth column respectively). In vacuum, irrespective of simulation size, all structures are stable and retain their ft-sheets. In water, the smaller silk structures (two and four polyAlanine stretches; a = 1, a = 3) return to their original largely disordered 'spidroin-like' state and only the larger structure (six poly-Alanine stretches; a = 5) remains stable. The natural condition can be assumed to be an intermediate between fully hydrated and vacuum due to the removal of water and ions from the silk as it is spun into air. ........................................................... 59 Table 2 1 Residues involved in the structural transition of silk chains with length a = 5 (206 residues). From the transition probabilities (a-helices - bold, 310helices/turns - italicized) the residue numbers are identified whose joint probability to transition is higher than 10%. These residue groups are the potential key players in the silk assembly mechanism of the core structure........ 67 Table 3 | Bulk modulus, bulk strength, and critical length scales for polymer m aterials. Taken from Reference [121]................................................................... Table 4 Table 5 78 I Process zone (cohesive zone) for the generalized J potential. ........... 105 I Summary of key structures and associated mechanisms of upscaling from the atomistic to larger scales. Table adapted from Reference [47].......113 Table 6 1 Entropy in J/molK for vacuum and solvated structure split in three independent parts S1, S2, S3. The change in entropy is calculated for the 211 amorphous phase only and the value in brackets gives the change for the entire stru cture ......................................................................................................................... 13 1 I Nomenclature for Secondary Structure Assignments........................... 138 Table 8 I Rotamers and associated secondary structures of Serine.....................141 Table 9 I Rotamers and associated secondary structures of Tyrosine.................142 Table 10 I Rotamers and associated secondary structures of Leucine................144 Table 7 Table 11 I Rotamers and associated secondary structures of Glutamine. .......... 146 Table 12 I Rotamers and associated secondary structures of Arginine. ............. 148 212