Collagen Peptide Simulated Bending After Applied Axial Deformation

Abstract

Structural proteins in the extracellular matrix are subjected to a range of mechanical loading conditions, including varied directions of force application. Molecular modeling suggests that these mechanical forces directly affect collagen’s conformation and the subsequent mechanical response at the molecular level is complex. For example, tensile forces in the axial direction result in collagen triple helix elongation and unwinding, while perpendicular forces can cause local triple helix disruption. However, the effects of more complicated mechanical loading, such as the effect of axial pretension on collagen bending and triple helix microunfolding are unknown. In this study we used steered molecular dynamics to first model a collagen peptide under axial tension and then apply a perpendicular bending force. Axial tension causes molecular elongation and increased the subsequent perpendicular bending stiffness, but surprisingly did not increase the predicted collagen triple helix microunfolding threshold. We believe these results elucidate a key potential mechanism by which microscale mechanical loads translate from cellular and micro scales down to the nano and atomistic. Further, these data predict that cryptic force-induced collagen triple helix unwinding is axial-deformation independent, supporting the possibility that cell traction forces could be a key molecular mechanism to alter the cellular matrix microenvironment to facilitate collagen enzymatic degradation and subsequent cellular migration, such as in tumor extravasation.

1. Introduction

Structural proteins are key mechanical components of the extracellular matrix throughout the body. These proteins are subjected to a wide range of mechanical conditions, including varied magnitudes of force, a range of directions of force application, and variable rates of force application. Accumulating data now suggests that collagen’s triple helical conformation is affected by mechanical loading conditions, such that the triple helix can be locally disrupted^1–3. These physiologically relevant protein conformation changes have not been fully revealed by NMR and x-ray crystallographic studies^4,5.

Steered molecular dynamics (SMD) simulations modeling tensile forces in the axial direction (axial deformation) resulted in a variety of conformational changes; most notable was progressive helical unwinding with increasing axial deformation^2,6. These computational loading simulations further indicated that collagen molecules exhibit strain-rate dependent mechanical properties during tensile loading², including “creep” behavior, that is, increasing strain with a constant applied load⁷. These observations during mechanical loading indicated that the single molecule has intrinsic viscoelastic behavior in tension.

Structural collagen in tissues (e.g., tendons, skin) is a fibrous protein that is subjected to axial forces. However at the molecular level, the intermolecular collagen cross-links are involved in force transmission⁸ and contribute to complex non-axial molecular deformation patterns¹. Previous molecular simulations reported in literature indicated that tensile loads transmitted through an enzymatic cross-link results in globular telopeptide disruption before triple helical domain failure⁹. In an earlier publication we modeled bending of the collagen triple-helical molecule from local force transmission across a single simulated spontaneous non-enzymatic cross-link by applying a force through a side chain directed perpendicularly to the long axis of the triple helix. We found that this non-axial bending load was predicted to induce local triple helix microunfolding at a force below any previously described collagen hierarchical structure failure mechanisms¹. Although not able to determine precise collagen molecular conformation, mechanical overload studies have been shown to sensitize collagen to atypical serine protease degradation^10,11. In addition, repeated overload resulted in a decrease in the enthalpy of denaturation indicative of molecular-level damage to the triple helix¹¹. A collagen hybridizing peptide designed to bind and label unfolded collagen was also shown to label mechanically loaded rat-tail tendon¹². These peptide-labeling results provide direct experimental evidence for molecular-level triple helix loss, and which the authors attributed (based on molecular modeling results) to collagen damage caused by shear-dominant loading induced collagen damage¹².

In vivo collagen is subjected to a variety of mechanical loading conditions and many of the reported collagen biomechanical studies focused on axial tension. However, single chondrocytes and fibroblasts were shown to bend single collagen fibrils (ranging from below measurable diameter to 500 nm diameter)¹³. Chondrocytes were reported to generate 0.3–0.6 nN of force on a collagen-proteoglycan matrix¹⁴ while fibroblasts generated 1 nN of force¹⁵. This type of collagen bending by cells has implications for the tumor microenvironment¹⁶, where cell traction forces in model tumor spheroids were found to precede cell extravasation from the spheroid¹⁷. Cell traction force is also considered a biomarker of metastatic potential¹⁸. Simple mechanical engineering principles describe how tensile force acting along the long axis of a beam will increase the beam’s resistance to transverse bending from a perpendicular force, however material changes inside the structure (beam) are unaccounted for in structural mechanics based models.

In a previous study¹, single-chain pullout (microunfolding) from a perpendicular force was attributed primarily to hydrogen bond (H-bond) loss¹, however in that study the collagen triple helical coil was still intact and no axial deformation had been applied. The purpose of this study was to quantify the bending mechanics and potential triple helix microunfolding pathways in a molecule loaded perpendicular to its long axis after it had already undergone axial deformation and partial triple helical unwinding. We hypothesized that an axial force causing a pre-axial elongation of the collagen molecule would not increase the microunfolding threshold force for a single α-chain but would increase the structural bending stiffness of the entire collagen peptide. We tested these hypotheses using steered molecular dynamics to apply increasing axial deformations prior to applying a perpendicular (bending) force through the C-β atom of the side chain of Arg44 in the middle of the peptide. The bending force was applied in this manner as a simplified model of force transmission across physiologically relevant covalent non-enzymatic cross-links, as previously introduced by Bourne and Torzilli¹. The results from this study were then compared with our previous microunfolding results¹, including an analysis using a structural mechanics beam bending model¹⁹.

2. Methods

2.1. Equilibration and SMD Procedures

The simulation system in explicit water was built with Visual Molecular Dynamics (VMD) 1.8.6²⁰. In brief, the crystal structure of a collagen type III peptide (PDB: 1BKV)⁴ was used in this study, and consists of a homotrimer containing a 12 residue flexible region of collagen type III that is flanked on both ends by a (Pro-Hydroxyproline (Hyp)-Gly)₃ segment and totaled 30 residues per α-chain⁴. The collagen peptide was protonated for a physiologic pH and immersed in a water box, with a 1.5 nm thick layer of water on the lateral sides and a 3 nm thick layer of water surrounding the triple helix along the axial direction (N and C termini). Na⁺ and Cl⁻ ions were added to bring the solution to 0.15 M and bring the system to neutral charge. The system contains ~18,000 atoms and measured 3.6 × 14.6 × 3.7 nm¹.

The molecular dynamics (MD) simulations were carried out with NAMD 2.6 molecular dynamics software package²¹ using CHARMM 27 force field²² and the TIP3P water model. The force field parameters for hydroxyproline are based on those described by In ‘t Veld and Stevens²³. The simulation system was first minimized using the conjugate gradient minimization algorithm for 5,000 steps²¹, and the electrostatic cutoff was 12 angstroms.

The system was then equilibrated at 310 Kelvin in three stages under adiabatic (NVE) conditions using periodic boundary conditions. Rigid bonds were used to constrain the hydrogen atoms so that a time step of 2 femtoseconds could be employed to reduce computational costs². The Cα atoms of the first residue of chain A and last residue of chain C were fixed for 250 picoseconds (ps) of first stage, then released for the next 500 ps in the second stage for the peptide to further equilibrate with the water phase. In the third stage, five separate 250 ps simulations were restarted from the end point of the second stage with no atoms restrained. The most linearly extended conformation of these five 250 ps simulations from the third stage was then used as the relaxed starting structure of the unloaded collagen peptide. Together, these three stages represented a total duration of 1,000 ps of equilibration simulation. This 3-stage equilibration strategy was chosen to avoid the use of an initial external steering force to straighten the molecule. This resulting 1,000 ps conformation was then used as the initial relaxed model for the subsequent tensile SMD simulations. Additional details about the 3-stage equilibration and the RMSD measurements were previously described¹.

The SMD simulations (in silico mechanical loading simulations) were performed at constant velocity with periodic boundary conditions and conducted in an isovolumetric and isothermal (NVT) ensemble as previously described⁶. Tensile deformations were conducted as described by Buehler and Wong²⁴. In brief, the Cα atoms of all three chains were fixed in space at the N-terminal end while the pseudo atom at the center of mass of the Cα atoms of all three chains at the C-terminal end was pulled in line with the molecular axis² to elongate the triple helix at an elongation rate (velocity) of 1 nm/ns (strain rate of 1.25 × 10⁸% s⁻¹). Such a tensile deformation configuration (boundary conditions) results in the pulled end being able to freely rotate, consistent with published data^2,23. In our simulations, the applied tensile strains (elongation/L_o) were <15%. These tensile deformation models were then used as the starting configuration (elongated length L) for subsequent bending simulations.

For the bending simulations, once the triple helix was elongated the Cα atom of each N and C termini residues of all three chains were fixed in space to limit overall displacement of the entire structure, and to serve as a simplified model of the long-range interactions of the peptide. These terminal residues would normally be connected to the adjacent residues within the triple helix in the overall linear structure with other structural cross-links and external forces that resist translation of the overall collagen molecule. The bending force, F, was applied through the side chain of Arg44 near the middle of the peptide through the Cβ atom. The arginine side chain was selected due to its ability to form covalent AGE cross-linking in vivo¹. The force applied to the Cβ atom was oriented perpendicular to and directed away from the long axis of the molecule through a theoretical spring (spring constant k = 10 kcal/mol/Ų)¹, the force moving at a displacement rate (velocity) of 1 nm/ns.

3. Results

3.1. Apparent Young’s Modulus From Tension in the Axial Direction

Separate tensile SMD simulations were performed whereby five different axial displacements (0.25, 0.50, 0.75, 1.00 and 1.50 nm) were applied to the ends of the model collagen type III peptide^1,4. Each tensile force-displacement curve for the five axial displacements was separately fit with a linear line, and the slopes used to calculate a representative average force and displacement curve, Figure 1A. The peptide’s tensile stiffness was estimated from the linear slope to be 1,605 ± 119 pN/nm (slope ± standard error, Figure 1A). Representative stress and strain values were calculated from the force-displacement curve, where the peptide geometry was approximated as a cylinder with radius of 0.95 nm based on peptide measurements calculated using VMD²⁰. The apparent Young’s modulus, E, calculated from the slope of linear line of the stress-strain curve, was 4.56 ± 0.34 GPa (Pa=N/M²). The measured apparent Young’s modulus was in reasonable agreement with the apparent average Young’s modulus of 5.4 GPa previously calculated from a combination of experimental and computational studies reported by Gautieri et al⁷. The actual resulting tensile elongations of the peptide were measured using VMD²⁰ to be 0.163, 0.360, 0.482, 0.765, and 1.181 nm, respectively. The peptide’s initial and final lengths were used to calculate the tensile strains of 2.0, 4.5, 6.0, 9.5 and 14.7%, respectively. The axially elongated peptide models and % strain values were then used for the subsequent SMD bending simulations.

Figure 1. Axial Tension Simulation Results.

(A) Collagen peptide tensile force and tensile elongation are plotted for five separate tensile deformation simulations. The tensile stiffness was determined from the slope of the linear fit to the data (solid line ± 95% confidence internal dotted line).

(B) Peptide helical unwinding is illustrated by representative collagen conformations’ changes in the amount of helical twist at approximately 0, 5, 10, and 15% axial strain.

Our results of the tensile simulations on this collagen type III model peptide showed unwinding with increasing tensile deformation (Figure 1B), which is consistent with previous tensile simulations conducted on this peptide²³. Published work by Gautieri et al.² found similar collagen unwinding during tension loading using a different model collagen peptide. Comparing our simulation result to that of Gautieri et al.² at a equivalent strain and deformation rate of ~10% and 1 nm/ns, respectively, found similar helical unwinding (rotation) and a comparable force-deformation response^2,23. In addition to the unwinding during peptide elongation, we found that the tensile deformation resulted in a 4% mean decrease in the spacing between the peptide backbones of the three collagen α-chains (decreased from 1.53 nm to 1.47 nm, n=5), suggesting that as the triple helical conformation rearranges due to axial tension the molecule’s apparent diameter decreases slightly (“necks down”) due to the tension (apparent Poisson’s ratio ~ 0.30).

3.2. Bending Stiffness Perpendicular to the Long Axis

For each applied tensile elongation, the perpendicular or transverse displacement of a single α-chain, in response to the perpendicular force applied through the C-β atom of Arg44 (Figure 2, inset), was measured from the position of the C-α atom of the pulled residue at the approximate mid-point of the collagen peptide. A representative bending force–perpendicular displacement response is shown in Figure 2. Examination of the force–displacement responses suggested two different regions, an initial low stiffness (slope) region followed by an increasing stiffness (higher slope) region, with the transition temporally correlated with the conformational transition (microunfolding) as the perpendicular force increased (Figure 2).

Figure 2. Force-Displacement Response of a Collagen Peptide During Bending.

A collagen peptide was axially extended 5% (tension), and a perpendicular force was then applied to a single α-chain via Arg44 (inset, top left). The horizontal dotted line indicates the average force threshold of triple helical microunfolding, and the black solid lines indicate linear fits of the pre (^▲) and post (^●) microunfolded regimes. Overlaid on the graph are the linear fits of the pre-microunfolded regime of the separate 0% axial extension-bending simulation (short-dashed blue line) and the also separate and independent simulation of the 14.7% axial extension-bending case (long-dashed red line). As indicated by the respective slopes, the 0% axial extension-bending case (short-dashed blue line) had the lowest stiffness, while the 14.7% axial extension-bending case (long-dashed red line) had a higher stiffness than the 5% axial-extension bending case (solid black line).

To compare pre- and post- microunfolding regions of the curve the force-displacement response was divided into an initial bending (pre-microunfolding) region, which represented at least 500 ps of SMD simulation, and a second post-microunfolding bending region at >500 ps. The bending stiffness for each of the elongated (and partially unwound) collagen peptides were then determined from the slope of each of the bending force–perpendicular displacement regions by fitting a linear line to the respective data. For the pre-microunfolding region, increasing the initial axial tensile strain or force resulted in an increase in the resistance to bending (Figure 3). The bending stiffness increased from approximately 350 pN/nm with no axial strain to over 600 pN/nm at 14.7% strain, at a rate of 1,664 ± 321 pN/nm (bending stiffness/tensile strain; rate of change in bending stiffness per tensile force was 0.115 ± 0.025 nm⁻¹). However, in the post-microunfolding region the bending stiffness did not change with increasing initial axial tensile strain, with the mean bending stiffness = 1,116 ± 16.5 pN/nm.

Figure 3. Perpendicular Bending Stiffness vs. Applied Axial Strain.

Collagen peptides underwent an axial tensile strain and then a perpendicular force was applied through the Arg44 side chain to simulate force transmission through a covalent cross-link. SMD simulations found that increasing tensile strains resulted in higher measured perpendicular (bending) stiffness.

3.3. Microunfolding Results

Past molecular modeling data suggested that α-chain loops can be pulled out from the collagen triple helix due to non-tensile force, and that local triple helical microunfolding could be observed at ~900 pN¹. To assess whether tensile pre-deformation affected microunfolding thresholds we used a previously described interchain spacing (ICS) analysis to determine force thresholds for triple helix loop pullout¹. This ICS approach takes a hypothetical slice across the peptide, normal with the long axis, and measures how far apart the three collagen α-chains are in space by following a representative C-α position from each α-chain as a function of time or applied force. Based on the ICS data the microunfolding threshold was 795 ± 63 pN (mean ± standard deviation; range 725 to 860 pN), and was not statistically different (P = 0.1268, t = 1.6823, df = 9, unpaired two-tailed t-test, QuickCalcs, GraphPad Software Inc., San Diego, CA) from the previous unstrained ICS results¹.

3.4. Collagen Peptide Flexural Rigidity Under Combined Axial and Transverse Forces

We previously calculated the apparent flexural rigidity EI (E = elastic modulus, I = area moment of inertia) of the collagen peptide using an elastic three-point beam bending model with fixed ends subjected to a single transverse force¹, an approach initially described by Buehler for the same model but with pinned ends⁶. Using this same approach, EI was calculated using a beam-bending model of the collagen molecule when subjected to a simultaneous axial tensile force (P) and transverse bending force (F). However in this model the principle of superposition of forces (P, F) does not apply, such that the equations for the transverse bending deflection (D) of a beam in three-point bending under combined axial tensile and transverse bending forces involves hyperbolic functions and powers of P/EI (see Young and Budynas, 2002, Table 8.9). Since it was not possible to directly solve for EI, it was necessary to fit the beam equation to the SMD data for the applied axial tensile force P (or elongation), transverse bending force F, and transverse deflection D. However the combined force model could not be used when P=0, so for this case EI was calculated from the simple beam model as previously described¹ (see Young and Budynas, 2002, Table 8.1). Finally, as described above, the ends of the three chains were each fixed in space after the collagen molecule was elongated, prior to application of the transverse force F. Thus the boundary conditions at the ends of the beam can be modeled as being pinned (free to rotate) or fixed (no rotation) (see Figure 4). These end conditions will affect the beam deflection D and apparent flexural rigidity EI. For simple transverse beam bending (P=0), the ratio of pinned-to-fixed end deflections is constant (as is EI), D_pinned/D_fixed = EI_pinned/EI_fixed = 4. However for combined axial tensile and transverse forces these relationships are not constant and vary as a function of the axial tensile force P. Thus EI was calculated using both pinned and fixed ended boundary conditions.

Figure 4. Apparent Bending Flexural Rigidity.

The apparent bending flexural rigidity (EI, mean ± SEM) was calculated for each axial tensile force (P) using the pinned (black bars) and fixed (gray bars) ended models. The EI_fit for all P>0 (n=4) and the EI_mean of the individual EI (n=5) are shown for both models; these were not statistically different within each model. In two cases in the pinned ends model, P=1,416 pN and P=2,124 pN, the EI curve fitting results were not significant (NS).

Shown in Figure 4 is the apparent flexural rigidity EI (mean ± SEM) for the pinned and fixed end conditions calculated from the curve fit for each tensile force P, the mean EI (EI_mean, n=5), and the EI_fit (n=4) calculated by simultaneously fitting all the data for P>0, F and D. For pinned ends, EI_pinned ranged from 6.19×10³ to <<1 pN•nm² for P = 0 to 2,124 pN, respectively, however the fits for P=1,416 pN and 2,124 pN were not significant (NS, p>0.8). The pinned EI_mean (2.56×10³ ± 1.38 pN•nm²) and EI_fit (0.96×10² ± 0.05 pN•nm²) were not statistically different. For the fixed end constraints EI_fixed ranged between 1.41×10³ to 1.43×10⁴ pN•nm², respectively; EI_mean (10.60×10³ ± 0.06 pN•nm²) was also not statistically different from EI_fit (9.63×10³ ± 2.25 pN•nm²). However for P>0 the EIs for the fixed end condition were almost an order of magnitude higher than the pinned end conditions. Using the respective EI_fit values for the pinned and fixed ends, we compared the predicted deflections D_pin and D_fix at the mean microunfolding force (755 pN) for axial tensile forces from 0 to 2,124 pN, and compared these deflections to the SMD model deflections at the corresponding mean microunfolding force (Figure 5). For P>0, D_pin/D_fix decreased from 8.2 to 4.1 with increasing P, the later close to the P=0 model of 4. In general, the fixed end model was a better fit over the full range of axial tensile forces, while the pinned end model appeared more accurate at higher P values. The fixed end model also had smaller coefficients of variance for the mean EI_fix (52%) and EI_fit (0.6%) compared to those for the pinned end values (121% and 5.2%, respectively).

Figure 5. Comparative Bending Deflection Between Fixed and Pinned Models.

Transverse bending deflection (D) at the microunfolding force (F = 755 pN) shown as a function of increasing tensile axial force (P>0) for the pinned (black line) and fixed (red line) ended beam models. For comparison, the SMD deflections (blue circles) are also shown, as are the simple beam deflections (P=0) for both end constraints (diamonds).

From the apparent flexural rigidity it is then possible to calculate the persistence length (ξ) for the peptide using ξ = EI/k_BT, where k_B is the Boltzmann constant and T is the temperature. This approach was previously reported to show reasonable agreement between the calculated persistence length from other collagen bending models and experimentally determined values^1,6. Based on the EI_mean for the pinned and fixed end models subjected to axial tensile and transverse bending forces, the persistence length at T = 310 K (37°C) was calculated to be 418 ± 347 nm and 2,730 ± 280 nm, respectively, far greater than previously determined computational and experimental values of approximately ~15–25 nm^1,6,25. However, the EI calculated in this present work simulates collagen as a structural element (a beam) already elongated by a tensile force instead of in a relaxed state. The divergence of the measured persistence length in this study from previous studies is therefore likely due to the new mechanical environment (collagen peptide in pre-tension) being modeled, where the bending deflections and flexural rigidities were an order of magnitude greater. Of interest, the persistence lengths were >> than the molecule length, which supports modeling the collagen axially deformed peptide as an elastic structural member versus a flexible polymer-like element.

Separating EI into the individual E and I components is complicated by the time-dependent changes in peptide conformation. However, if I is instead approximated as a fixed structure, for example a thin rod with a radius of r, I can be estimated from I = 0.25(πr⁴). Assuming a radius of 0.95 nm, the apparent Young’s modulus is estimated to be 4.0 GPa and 16.6 GPa for the pinned and fixed end models with P>0, respectively; with the pinned Young’s modulus being in reasonable agreement with the average Young’s Modulus of 5.4 GPa reported by Gautieri et al.⁷.

4. Discussion

Fibrillar collagens exhibit complex hierarchical structures in vivo, and enzymatic collagen cross-links formed during development play a critical role in both structural organization and mechanical function²⁶. Due in part to the strength of the collagen triple helix, mechanical failure of collagenous structures has been attributed to the breaking of cross-links and the subsequent slippage of collagen molecules^6,8,9,27. To better understand the nanomechanical properties of these cross-links recent molecular modeling has focused on describing the effect of mechanical loading and force transmission, and found that mechanical loads can induce dramatic changes in local conformation^1,9.

In this present work we extend an alternative cross-link loading model previously introduced by Bourne and Torzilli¹. In this revised model we first applied a tensile force to elongate the collagen molecule, and then pulled a side chain perpendicular to the long axis, the later perpendicular force chosen to model force transmission in a cross-linked structure. Due to the complex structure of collagenous tissues, this new type of combination loading pattern serves as an approximation of force transmission in cross-linked load bearing tissue. At a cellular level, this loading pattern can serve as a simplified approximation of the effects of cell traction forces bending collagen during migration through an already axial-tension deformed collagen matrix (e.g., fibroblast migration during wound healing²⁸ and cancer cell migration in the tumor microenvironment^16,29

Based on our SMD tensile-bending experiments we found an increase in the apparent bending stiffness with increasing axial pretension. This new bending stiffness data provides important mechanical property parameters for multi-scale modeling, such as by Sopakayang et al.³⁰, and for model systems that include cross-links (either explicitly or implicitly) in the collagen structure. In addition, the increase in apparent bending stiffness predicted by this study suggests a possible mechanism by which mechanical forces deforming the collagen matrix will change the perceived mechanical microenvironment for cells. For example, this work indicates that a tensile mechanical loading on the collagen matrix that causes axial deformation would result in a perceived increase in bending stiffness to a cell that is exerting a traction force on the collagen, or cells that are otherwise sampling matrix stiffness. These findings represent a theoretical model that potentially connects extracellular matrix mechanical loading with cellular responses to matrix stiffness, as illustrated by the experimental observations reported by Mason et al. who found that AGE cross-linking of collagen gels can alter endothelial cell behavior independent of receptor for AGE (RAGE) activity³¹.

Collagen has been shown to spontaneously denature at body temperature³², and small tensile forces have been predicted to stabilize the collagen triple helix against these thermal fluctuations and subsequent denaturation³³. Although tension may inhibit thermal fluctuations in the triple helix, small tensile deformation does not appear to increase the perpendicular force magnitudes needed to mechanically induce triple helical microunfolding. In our simulations, tensile deformation at the ranges studied did not result in a measurable decrease in the number of predicted hydrogen bonds within the peptide. These H-bond results were in contrast with Gautieri and colleagues, who reported a decrease in H-bond numbers after axial strain exceeded approximately 10%². This difference may be due to the collagen peptide sequences used in the respective studies, and the maximum axial strain applied in this study, which was <15%. Although beyond the scope of the present study, pilot data indicated that slightly higher axial strains in the 1BKV collagen peptide did decrease H-bond numbers, similar to Gautieri’s findings with the [(Gly-Pro-Hyp)₁₀]₃ collagen model peptide². Of relevance to interpreting these data, Ghanaeian and Soheilifard recently reported a sequence effect on H-bond number in response to axial stretching when comparing between different collagen-like model triple helical peptides³⁴.

As triple helix microunfolding is most likely due to force breaking the hydrogen bonds, the unchanged microunfolding threshold results are therefore attributed largely to the H-bonding being unaffected by tension in this study when compared to the earlier perpendicular bending findings with the same 1BKV peptide in the absence of axial tension¹. These data suggest that collagen sequence-specific H-bond characterization may therefore be warranted, and highlights the importance of multi-scale mechanics in order to connect how physiological tissue-level strains deform collagen at its nanoscale molecular level. In light of these computational data, further experimental characterization on non-axial force transmission would be helpful in advancing multi-scale models.

The tensile deformation results found in this study were consistent with previous studies showing helical unwinding and modulus values that agreed with those from both the same peptide²³ and another related collagen model². Although other tensile mechanics simulations have indicated that collagen’s conformation changes are due to tensile loads and deformations, to our knowledge this is the first study that has addressed how “mechano-conformational” changes induced by tensile forces affect subsequent bending loads, deflections, and collagen conformation. These data also supports a potential molecular-level collagen conformational change as a mechanism for previously reported atypical serine protease degradation of mechanically loaded collagen^10,11, and continues to suggest a putative mechanism for why low tensile strain of glycation cross-linked rat-tail tendon sensitized this material to bacterial collagenase degradation when compared to axially deformed native tail tendon³⁵. In addition to shear damage in the collagen molecule¹², these modeling data continues to support the plausibility of a damage-independent mechanically induced collagen conformational change occurring at sub-nanonewton force thresholds. With new technologies now available, such as the collagen hybridizing probe reported by Zitnay et al.¹², these modeling data suggests that interrogating the tumor microenvironment for local helix disruption may be feasible and warranted. These computational predictions also suggest that biaxial stretch of highly uniaxial-aligned collagen tissues or collagen constructs might serve as valuable model systems for future cell mechanobiology and structural studies.

In conclusion, these results provide new insights into the interplay of nanoscale conformational changes and mechanical properties of molecular collagen. In addition, changes in the collagen peptide’s bending stiffness caused by increasing axial deformations suggests an additional potential cellular mechanobiological mechanism that connects cytoskeletal force generation and cellular migration with cell — matrix interactions and collagen conformational changes. These results highlight the complexity of the collagen nanomechanical response and potential molecular damage mechanisms, emphasizing the importance of connecting atomistic mechanical simulations with higher order multiscale modeling and experimental studies to characterize both the conformational response and mechanical behaviors of collagenous tissues.