Article Information

• PDF (488 kb)
• Full Text with Large Figures
• Cited by in Scopus (5)

PubMed

• Articles by Markus J. Buehler
• Articles by Sophie Y. Wong

Related Articles

• Contour Length and Refolding Rate of a Small Protein Control...

  Contour Length and Refolding Rate of a Small Protein Controlled by Engineered Disulfide Bonds Biophysical Journal, Volume 92, Issue 1, 1 January 2007, Pages 225-233 Sri Rama Koti Ainavarapu, Jasna Brujić, Hector H. Huang, Arun P. Wiita, Hui Lu, Lewyn Li, Kirstin A. Walther, Mariano Carrion-Vazquez, Hongbin Li and Julio M. Fernandez Abstract The introduction of disulfide bonds into proteins creates additional mechanical barriers and limits the unfolded contour length (i.e., the maximal extension) measured by single-molecule force spectroscopy. Here, we engineer single disulfide bonds into four different locations of the human cardiac titin module (I27) to control the contour length while keeping the distance to the transition state unchanged. This enables the study of several biologically important parameters. First, we are able to precisely determine the end-to-end length of the transition state before unfolding (53Å), which is longer than the end-to-end length of the protein obtained from NMR spectroscopy (43Å). Second, the measured contour length per amino acid from five different methods (4.0±0.2Å) is longer than the end-to-end length obtained from the crystal structure (3.6Å). Our measurement of the contour length takes into account all the internal degrees of freedom of the polypeptide chain, whereas crystallography measures the end-to-end length within the “frozen” protein structure. Furthermore, the control of contour length and therefore the number of amino acids unraveled before reaching the disulfide bond () facilitates the test of the chain length dependence on the folding time (). We find that both a power law scaling ∝ with =4.4, and an exponential scaling with fit the data range, in support of different protein-folding scenarios. Abstract | Full Text | PDF (403 kb)

• Stabilization Provided by Neighboring Strands Is Critical fo...

  Stabilization Provided by Neighboring Strands Is Critical for the Mechanical Stability of Proteins Biophysical Journal, Volume 95, Issue 8, 15 October 2008, Pages 3935-3942 Deepak Sharma, Gang Feng, Dingyue Khor, Georgi Z. Genchev, Hui Lu and Hongbin Li Abstract Single-molecule force spectroscopy studies and steered molecular dynamics simulations have revealed that protein topology and pulling geometry play important roles in determining the mechanical stability of proteins. Most studies have focused on local interactions that are associated with the force-bearing -strands. Interactions mediated by neighboring strands are often overlooked. Here we use Top7 and barstar as model systems to illustrate the critical importance of the stabilization effect provided by neighboring -strands on the mechanical stability. Using single-molecule atomic force microscopy, we showed that Top7 and barstar, which have similar topology in their force-bearing region, exhibit vastly different mechanical-stability characteristics. Top7 is mechanically stable and unfolds at ∼150pN, whereas barstar is mechanically labile and unfolds largely below 50pN. Steered molecular dynamics simulations revealed that stretching force peels one force-bearing strand away from barstar to trigger unfolding, whereas Top7 unfolds via a substructure-sliding mechanism. This previously overlooked stabilization effect from neighboring -strands is likely to be a general mechanism in protein mechanics and can serve as a guideline for the de novo design of proteins with significant mechanical stability and novel protein topology. Abstract | Full Text | PDF (561 kb)

• Inferring the Diameter of a Biopolymer from Its Stretching R...

  Inferring the Diameter of a Biopolymer from Its Stretching Response Biophysical Journal, Volume 89, Issue 1, 1 July 2005, Pages 80-86 Ngo Minh Toan, Davide Marenduzzo and Cristian Micheletti Abstract We investigate the stretching response of a thick polymer model by means of extensive stochastic simulations. The computational results are synthesized in an analytic expression that characterizes how the force versus elongation curve depends on the polymer structural parameters: its thickness and granularity (spacing of the monomers). The expression is used to analyze experimental data for the stretching of various different types of biopolymers: polypeptides, polysaccharides, and nucleic acids. Besides recovering elastic parameters (such as the persistence length) that are consistent with those obtained from standard entropic models, the approach allows us to extract viable estimates for the polymers diameter and granularity. This shows that the basic structural polymer features have such a profound impact on the elastic behavior that they can be recovered with the sole input of stretching measurements. Abstract | Full Text | PDF (154 kb)

• …more

Copyright © 2007 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 93, Issue 1, 37-43, 1 July 2007

doi:10.1529/biophysj.106.102616

Biophysical Theory and Modeling

Entropic Elasticity Controls Nanomechanics of Single Tropocollagen Molecules

Markus J. Buehler^*, ,   and Sophie Y. Wong^†

^* Laboratory for Atomistic and Molecular Mechanics, Department of Civil and Environmental Engineering, Cambridge, Massachusetts
^† Division of Biological Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts

Address reprint requests to M. J. Buehler, Tel.: 626-628-4087.

Collagen, the most abundant protein on earth, is a fibrous structural protein with superior mechanical properties, and provides an intriguing example of a hierarchical biological nanomaterial. Collagen consists of triple helical tropocollagen (TC) molecules that have highly conserved lengths of L ≈ 300nm, roughly 1.5nm in diameter 1,2,3,4,5,6,7. Staggered arrays of TC molecules form fibrils, which arrange to form collagen fibers 1,2,8.

Collagen plays an important role in many biological tissues, including tendon, bone, teeth, cartilage, and in the cardiovascular system. Collagen is typically subjected to mechanical tensile load under physiological conditions 8,9,10.

Due to lack of understanding of nanomechanical properties of collagen, the source of the elasticity of individual TC molecules, fibrils, and collagen fibers remains controversial.

There are several reports of experimental studies focused on the mechanics of single tropocollagen fibers 7,15,16,17,18,19,20. However, tropocollagen molecules have rarely been studied using molecular dynamics (MD) studies. In a recent investigation, the mechanical properties of collagen fibers, using MD studies, focus on their Young’s modulus 11 for tensile strains up to 4%. Another recent study was carried out using molecular mechanics, considering the stretching mechanics of tropocollagen molecules up to 10% deformation 5. Other studies focused on the stability of collagen molecules 12 or the effect of point mutations on the stability 13,14. Some researchers modeled collagen at the continuum scale, using techniques such as finite element modeling 15.

However, no studies of the large deformation behavior of tropocollagen molecules, including elastic and fracture regimes, have been reported. Further, most existing atomistic studies were limited to short sequences of tropocollagen proteins that did not exceed 10nm, thus significantly below the length of molecules extracted from tissue, which typically features molecular lengths ≈300nm 1,2,3,4,5,6,7. Atomistic investigations of the bending mechanics and the mechanics of long tropocollagen molecules at finite temperature are still missing. Such studies are vital to understand the effect of entropic contributions to elasticity and its competition with energetic terms.

Therefore, limited understanding exists about the transition from entropic to energetic elasticity, under which conditions either one dominates, and how properties change when individual TC molecules assemble into collagen fibrils 8. Understanding the nanomechanical properties of single TC molecules is essential to develop theories that link molecular and tissue properties of collagen.

Early experiments suggested that TC molecules are rigid, rod-shaped structures 16,17. However, based on hydrodynamic methods and transmission electron microscopy (TEM), it was found that TC molecules feature some flexibility, which can be measured by the persistence length. Hydrodynamic methods suggested a persistence length of 130–180nm 18,19,20; TEM-based estimates 21 range between 40nm and 60nm. Recent experiments using optical tweezers suggested a much lower persistence length, with values ranging between 11nm and 15nm 22,23,24. Thus far, no satisfying explanation for these discrepancies has been provided. No molecular model of TC has been developed capable of describing different regimes of deformation.

Here we report hierarchical modeling consisting of sequential atomistic and molecular simulation that confirms optical tweezer results quantitatively, without relying on empirical parameters. We employ a mesoscale molecular model as a strategy to upscale the mechanical properties of single TC molecules to large timescales to enable capturing entropic effects. Our model is capable of carrying our molecular stretching experiments, with high precision, at timescales of tens of microseconds, approaching experimental deformation rates.

We find that when the TC molecule is stretched below its contour length (x<L), entropic elasticity governs elasticity. When the TC molecule is stretched close to and beyond its contour length (x>L), we observe a transition to energetic elasticity, featuring a two-regime hyperelastic or nonlinear elastic behavior due to unfolding and breaking of hydrogen bonds and stretching of the polypeptide backbone, followed by molecular fracture. We find that assembling TC molecules into fibrillar structures leads to rapid increase of the persistence length, rendering entropic elasticity less significant.

Our molecular, mesoscale model is derived from a series of full atomistic simulations, similar to those described in the literature 25,26. We note that the focus of this article is not development of the mesoscale model. Instead, here we focus on applying the mesoscale model to describe deformation of ultra-long TC molecules.

However, extending earlier work 25, here we pay particular attention to the bending stiffness of TC molecules and its dependence on the loading rate. Therefore, we carry out additional molecular dynamics simulations of bending a short sequence of a TC molecule.

For the full atomistic simulations, we use the classical force field CHARMM 27,28, implemented in the molecular dynamics (MD) program NAMD 29. The CHARMM force field 27,28 is widely used in the protein and biophysics community, and provides a basic description of the atomistic interactions in proteins. It is based on harmonic and anharmonic terms describing covalent interactions, in addition to long-range contributions describing van der Waals interactions and ionic (Coulomb) interactions as well as hydrogen-bonding.

We visualize the molecular structures using the visual molecular dynamics 30 program. We consider the atomic coordinates of a single TC molecule based on the crystal structure PDB ID 1QSU 31 with 1.75Å resolution, a short sequence of a TC molecule with length L ≈ 8.4nm. The 1QSU is a triple-helical type II collagen-like molecule with sequence (Pro-Hyp-Gly)₄-Glu-Lys-Gly-(Pro-Hyp-Gly)₅.

This model serves as a simple representation for the amino-acid sequence that makes up the sequence of longer tropocollagen molecules extracted from live tissue.

We use the psfgen program of NAMD 29 to add hydrogen atoms and to assign charges. The molecular model is set up according to pH 7, resembling conditions in experiment 24. No electrolytes are present in our atomistic setup.

We include crystallographic water in the simulations, and also add a skin of water of a few Ångströms surrounding the tropocollagen molecule. We note that electrostatic interactions can be significantly weakened by screening due to solvent molecules, which can lead to interactions that are weaker than van der Waals interactions and that decay over distances of several Ångströms 32.

Simulations with forces applied are carried out after energy minimization and finite temperature relaxation at T=300K. All simulations reported in this article are carried out at T=300K with temperature control (NVT ensemble), using a time step Δt=1 fs. We apply forces to the molecule using an SMD scheme 33. Forces and fixed constraints are applied at the C_α atom in each chain. Forces are applied on each of the three α-chains equally. In the case of a fixed end, each C_α at the end of the molecule is fixed. No torsional loads are applied since the end of the tropocollagen molecule is free to rotate. This is essential to allow for proper unfolding of the molecular structure.

Forces in the SMD scheme are applied by adding a potential energy term to the Hamiltonian of the system,

(1)

To carry out the tensile experiments, we use a SMD spring constant k_SMD=10kcal/mol/Ų, with Å/step.

Postprocessing of the simulations is done using the CMDF framework 34, a computational Python-based simulation environment capable of seamless integration of various file formats and computational engines, including molecular and crystal building tools. CMDF Python scripts are used to analyze and postprocess the simulation data, as required, for example, to compute statistical averages of force-displacement curves.

Fig. 1 depicts results of stretching calculations of a single TC molecule, based on the CHARMM model and the ReaxFF model (for details and simulation conditions, see 25). Fig. 2 shows snapshots of the atomistic geometry of the tropocollagen molecule as it undergoes tensile deformation.

Display large version of this figure

Figure 1

Force-displacement response of a single TC molecule with L ≈ 8.4nm under tensile loading, showing results obtained using different methods (nonreactive CHARMM force field (A) 27, ReaxFF (B), and the reactive mesoscale model (C) 25). We distinguish four regimes as the molecule undergoes deformation: Regime I is characterized by uncoiling of the triple helical structure, and Regime II with a larger tangent modulus corresponds to stretching of covalent bonds. In Regime III, we observe molecular fracture followed by rapid decay of the force in Regime IV.

Display large version of this figure

Figure 2

Snapshots during deformation of a single tropocollagen molecule, for different levels of tensile strain ranging from 0% to 50%.

The curves reveal the highly nonlinear or hyperelastic stress-strain behavior of a single TC molecule as the tensile strain increases up to 50%. We observe four regimes as the molecule undergoes large tensile deformation: Regime I is characterized by uncoiling of the triple helical structure; Regime II with a larger tangent modulus corresponds to stretching of covalent bonds in the protein backbone. In Regime III, we observe molecular fracture due to breaking of peptide bonds, followed by rapid decay of the force in Regime IV. The stiffness ratio of the second to first regime is ∼8, underlying the significant stiffening effect at tensile deformation beyond ∼30%.

To determine the bending stiffness, we fix both ends of the TC molecules. This is achieved by fixing the terminal C_α atom in each of the three polypeptides. We apply a force in the center of mass of the TC molecule using steered molecular dynamics (using k_SMD=0.01kcal/mol/Ų. To study the variation of bending stiffness with deformation rate, we simulate this three-point bending experiment at varying displacement rates, where =0.01 to 0.1m/s. The applied force is orthogonal to the axial direction of the TC molecule, leading to bending of the molecule upon increasing force. From the force-displacement data obtained by atomistic modeling, the bending stiffness EI is given by

(2)

Fig. 3 depicts the bending stiffness as a function of displacement rate. The results suggest a linear relationship between the bending stiffness and the deformation rate. Extrapolation to zero deformation rate yields an estimate for the quasi-static bending stiffness, EI ≈ 1.247×10⁻²⁹ Nm².

Display large version of this figure

Figure 3

Dependence of bending stiffness EI of a single TC molecule on the deformation rate, L ≈ 8.4nm, at T=300K. The results indicate that EI decreases linearly with decreasing loading rate. A linear fit to the data enables us to extrapolate to The resulting bending stiffness for vanishing deformation rate is used as input parameter for the mesoscale molecular model.

We note that the TC molecule is several thousand times stiffer than a carbon nanotube of comparable diameter (e.g., a 5,5 carbon nanotube).

The full atomistic calculation results in feed parameters for the molecular model, representing TC molecules at a mesoscale, as shown schematically in Fig. 4. Our molecular model is implemented in LAMMPS 35, and parameters are identical to the model described in the literature 25,26. Here we chose a smaller bead-to-bead distance of 7Å and adapt the bending stiffness to the extrapolated value for zero deformation rate (see Fig. 3). The effects of the solvent are treated implicitly in the model.

Display large version of this figure

Figure 4

Schematic showing development of the coarse-grained molecular model from a full atomistic description. The full atomistic representation of the triple-helical TC structure is replaced by a collection of beads, each of which represents ≈10 protein atoms plus water molecules 25, with r₀=7Å. The bottom subplot depicts how the load is applied to the long tropocollagen molecule. The outermost particle of the molecule is kept fixed during the simulation, and a slowly increasing force is applied on the right end.

Since the parameters in our molecular model are fitted directly to match the nanomechanical properties of the single TC molecule (PDB ID 1QSU), our model intrinsically assumes that the sequence along the length of the tropocollagen molecule does not vary. This is an approximation since collagen extracted from live tissue displays sequence variations along the axial directions. These sequence variations may induce modifications of the mechanical properties, including changes to the bending stiffness. These effects are neglected in the present study and left to future work.

The bottom subplot in Fig. 4 depicts how the load is applied to the long tropocollagen molecule using the SMD scheme 33. The outermost particle of the molecule is kept fixed during the simulation, and a slowly increasing force is applied on the right end. The SMD spring constant is chosen to be k_SMD=1kcal/mol, and we simulate deformation at a pulling velocity of 0.025m/s (for the SMD formulation, see Eq. (1); the SMD model in the atomistic and molecular model have the same formulation). The continuously increasing force slowly stretched the molecule from a convoluted initial shape to a straight configuration, followed by further extension and fracture. During the simulation, we record snapshots of magnitude of force applied and corresponding end-to-end distance of the molecule. This force-extension data is averaged into bins after the simulation is over to obtain a smooth curve. We find that further reduction of loading rate shows negligible changes in the force-extension curve.

We consider a stretching experiment of a TC molecule with L=301.7nm. The time step used in molecular modeling is Δt=20×10⁻¹⁵s, and we reach a total simulation time of t_tot ≈ 20×10⁻⁶s. Each mesoscale particle has a mass of m=1358.7amu, accounting for the mass of the TC segment it represents plus a skin of water molecules in a radius of 3Å.

All calculations are carried out in three dimensions.

Figure 5a shows the entire force-displacement curve of stretching a single TC molecule (x denotes the end-to-end distance of the TC molecule).

Display large version of this figure

Figure 5

Force-displacement (F(x)) curves of stretching a single TC molecule, L=301.7nm, at 300K. Subplot (a): Force-displacement curve over the entire deformation range, covering four stages: (I) uncoiling of the entangled configuration, (II) uncurling of the triple helix, (III) stretching of covalent bonds in the individual polypeptides, and (IV) rupture of the TC molecule. The dashed line of indicates the contour length of the molecule. The inlay depicts dɛ/dx(x), showing that entropic contributions dominate for x<280nm. Subplot b depicts a subset of the results depicted in the previous figure, focusing on the small force, entropic response (F<14 pN). This plot also depicts experimental results obtained for TC molecules with similar contour lengths 23,24, as well as the prediction of the WLC model with persistence length ξ_p ≈ 16nm.

Loading of the TC molecule starts from a coiled entangled configuration of the molecule with end-to-end distance of ≈100nm. During the initial regime (Regime I), the molecule loses its entangled structure, while the applied forces remain relatively low. Once the contour length is reached (x=L), (Regime II) uncurling of the triple helix, (Regime III) stretching of covalent bonds in the individual polypeptides, and (Regime IV) rupture of the TC molecule occurs, followed by a sharp drop of the forces to zero.

The qualitative behavior of TC molecules under stretch is similar to recent experimental studies carried out on collagen fibrils that reach forces on the order of several μN 36, also showing a significant hyperelastic stiffening effect.

Figure 5b depicts a detailed view into Regime I, including a quantitative comparison with experiment 23,24. The plot reveals very good agreement, even though the deformation rate in MD is still much higher.

The elastic properties of materials can be expressed as partial derivatives of the free energy density with respect to strain ɛ. Based on the free energy density A_V=(U−TS)/V (where V stands for the system volume), the scalar stress σ=∂A_V/∂ɛ. At constant temperature, the entropic regime is characterized by ∂U/∂ɛ ≈ 0. In contrast, during the energetic regime, ∂S/∂ɛ ≈ 0.

Changes in the internal energy U can be correlated with changes in the average internal tensile strain in the TC molecule (dɛ(x)/dx), where

(3)

(4)

For x≥280nm, ɛ increases slowly, leading to a change of internal energy that contributes to the elastic response. We find that ɛ ≈ 1.2% when x → L. These changes indicate the onset of a smooth transition from entropic to energetic elasticity at x=L.

Energetic elasticity dominates when x>L and leads to a significantly stiffer response. Entropic contributions to the elastic response decay quickly and approach zero at x ≈ 317nm, suggesting that energetic terms dominate for x>317nm. This is confirmed by considering that dɛ/dx approaches a constant value (see inset of Figure 5a).

By fitting the Marko-Siggia estimate of the worm-like chain (WLC) model 37,38 to the MD results (for F<11 pN), we determine an effective persistence length ξ_p ≈ 16nm, with L=301.7nm.

Even though the WLC model is a reasonable approximation for small forces, it ceases to hold when significant deformation occurs and the molecule is stretched toward its contour length (see Figure 7a, curve f_WLC−f₁).

The significance of entropic effects for a given end-to-end distance x is characterized by the ratio α=L/ξ_p, noting that ξ_p ∼ EI.

Assembly of TC molecules into fibrils leads to strongly increased bending stiffness, thus reducing the significance of entropic elasticity. Fig. 6 depicts the bending stiffness as a function of the number of molecules in a fibril, suggesting a power-law scaling EI ∼ n^p, where p ≈ 2.5. Elastic beam theory predicts p=3, confirming the results of our numerical model.

Display large version of this figure

Figure 6

Increase in bending stiffness of a collagen fibril, as a function of number of TC molecules in the radial direction. The plot shows the bending stiffness normalized by the value for a single TC molecule. The results suggest a strong increase of the bending stiffness with number of molecules. A power-law fit suggests a scaling of EI ∼ n^2.5.

We note that, for simplicity, in this study we only consider a simple stacking without axial displacement, as shown in the inlay of Fig. 6. We note that collagen fibrils show a staggered arrangement of molecules 8. This effect is not accounted for in the present article. To implement the three-point bending test, we fix two outermost left and right particles in the lowest molecule. We apply a slowly increasing force in the center particle of the top molecule using the SMD scheme.

Collagen fibrils have diameters of several tens of nanometers and consist of hundreds of aligned TC molecules. Since α ∼ EI^−p, α<1 for n>10—suggesting that the elasticity in collagen fibrils includes significant energetic contributions, particularly in the regions of dense molecular packing. We have confirmed the effect of increasing EI by direct MD calculations that clearly show the reduction in entropic elasticity, leading to much lower forces in the stretching regime below the contour length (x<L). We note that while the bending stiffness increases significantly after assembly, our model predicts that the tensile stiffness or Young’s modulus decreases continuously with increasing number of molecules.

However, a possible source of entropic elasticity in collagen fibrils are hydroxyproline-deficient sequences (labile domains) that represent more floppy parts of a collagen fibril with reduced molecular density 6,7,8,39, as they occur, for instance, in the gap regions. The mechanism studied in this article may thus relate to the transition from less ordered to more ordered molecule states, as discussed in the literature 6,7, where it was suggested that disappearance of “thermally activated kinks in the collagen molecules” may account for the entropic elasticity in fibrils. Thus, the entropic regime found in this article may account for the initial regime of stress-strain behavior for fibril strains up to 8%, shown in Fig. 6 in Misof et al. 6.

In summary, the toe region may relate to Regime I (entropic elasticity), and the heel region may relate to Regime II (energetic elasticity) 6,7. The qualitative shape of the force-extension curves showing the strong progressive increase of the tangent slope with strain is found in both our simulations and experiments.

However, since the length of labile domains in collagen fibrils is typically limited to 60 residues (or ∼20nm), quantifying their relative contribution to elasticity of a collagen fibril remains unclear. We leave further investigations of this topic to future research.

Further, variations of the molecule properties along the molecule axis may change the stretching mechanics of single tropocollagen molecules. This is because the variation of sequence along the axial direction leads to locally varying bending stiffnesses and therefore modifications in the local persistence length. Such effects could be analyzed by carrying out studies similar to the one presented here, by considering different amino-acid sequences and then mapping them onto a axial variation of nanomechanical properties. We emphasize that our molecular modeling approach can, in principle, easily be adapted to account for such effects.

We use the MD results to develop a continuum model for the elasticity of TC molecules. The force as a function of distance (F(x)) is given by

(5)

(6)

(7)

The functions f₁ and f₂ are correction functions that describe the transition from entropic to energetic elasticity. We note that f₁(x)=0 for x→0, f₁(x)→∞for x→L (due to the divergence of the WLC model), and f₂(x)=0 for x>>L, but f₂ reaches a finite value for x→L. Note that

Both functions are determined numerically by MD (see Fig. 7). For L=301.7nm, x_break=452.6nm, x₁=39.0nm, k⁽⁰⁾=51.7 pN/nm, k⁽¹⁾=294.8 pN/nm, and ξ_p=16nm. The predictions of our model agree with the numerical results shown in Fig. 5.

Display large version of this figure

Figure 7

Numerical estimates of the functions f_WLC—f₁(x) and f₂(x). These functions describe the smooth transition from entropic elasticity to energetic elasticity. The results show that entropic forces reach a maximum when the molecule is stretched to its contour length, approaching 140 pN.

Our model includes contributions from entropic elasticity, energetic elasticity, and molecular fracture, within a single formulation, and does not contain empirical fitting parameters. All physical properties of TC molecules are determined from full atomistic calculations. Based on our force-stretching data, we find a persistence length of ξ_p ≈ 16nm. Our force-stretching curves are in quantitative agreement with results obtained by optical tweezer experiments (see Figure 5b) 23,24, which suggested persistence lengths between 11 and 15nm.

We note that our estimate of ξ_p from the force-stretching curve is larger than the value obtained by considering

(8)

This suggests that ξ_p determined from force-extension curves in both molecular dynamics and experiment 23,24 represents an effective persistence length. Calculation of the persistence length from the tangent correlation function confirms that suggesting that ξ_p ≈ 30…45nm. Interestingly, this is close to experimental TEM results obtained by optical methods 21 that are based on a similar geometrical analysis scheme. These observations may help to explain the discrepancies noted in earlier studies 23,24.

An alternative reason for the large variations in persistence lengths in experiment 18,19,20 is based on the idea that not a single molecule, but an assembly of several molecules has been stretched, effectively increasing EI that leads to an increased persistence length 24. This possibility is not inconsistent with our model (see Fig. 6).

The observed transition from an entropic to an energetic elastic regime found here may correspond to experimental observations of progressive increase in the tangent slope of the stress-strain curve during stretching of collagen fibrils 6,7. Our results confirm that the transition from less-ordered to more-ordered molecule states as discussed in the literature 6,7 can lead to the characteristic mechanical behavior.

Our results illustrate a possible strategy to develop molecular or mesoscale models of complex biological systems, considering various mechanisms of deformation, ranging from entropic to energetic elasticity, considering fracture mechanics of individual molecules. Similar methods could be used for other biopolymers with similar geometry. Future studies of mechanics of individual and ensembles of TC molecules could help to shed light on assembly processes or help to clarify the mechanisms during cellular production of collagen molecules. Further, the effect of sequence changes along the molecular axis is an important subject of study.