Collagen fibrils from both positional and energy-storing tendons exhibit increased amounts of denatured collagen when stretched beyond the yield point

Abstract

Collagen molecules are the base structural unit of tendons which become denatured during mechanical overload. We recently demonstrated that during tendon stretch, collagen denaturation occurs at the yield point of the stress-strain curve in both positional and energy-storing tendons. We were interested in investigating how this load is transferred throughout the collagen hierarchy, and sought to determine the onset of collagen denaturation when collagen fibrils are stretched. Fibrils are one level above the collagen molecule in the collagen hierarchy, allowing more direct probing of the effect of strain on collagen molecules. We isolated collagen fibrils from both positional and energy-storing tendon types and stretched them using a microelectromechanical system device to various levels of strain. We stained the fibrils with fluorescently labeled collagen hybridizing peptides that specifically bind to denatured collagen, and examined whether samples stretched beyond the yield point of the stress-strain curve exhibited increased amounts of denatured collagen. We found that collagen denaturation in collagen fibrils from both tendon types occurs at the yield point. Greater amounts of denatured collagen were found in post-yield positional fibrils than in energy-storing fibrils. This is despite a greater yield strain and yield stress in fibrils from energy-storing tendons compared to positional tendons. Interestingly, the peak modulus of collagen fibrils from both tendon types was the same. These results are likely explained by the greater crosslink density found in energy-storing tendons compared to positional tendons. We believe that insights gained from this study could help management of tendon and other musculoskeletal injuries.

Graphical Abstract

Statement of Significance

When tendons are stretched or torn, this can lead to collagen denaturation (damage). Depending on their biomechanical function, tendons are considered positional or energy-storing with different crosslink profiles. By stretching collagen fibrils instead of fascicles from both tendon types, we can more directly examine the effect of tensile stretch on the collagen molecule in tendons. We found that regardless of tendon type, collagen denaturation in fibrils occurs when they are stretched beyond the yield point of the stress-strain curve. This provides insight into how load affects different tendon sub-structures during tendon injuries and failure which will help clinicians and researchers understand mechanisms of injuries, leading to improved treatment outcomes.

1. Introduction

Collagen is the most abundant protein found in mammals, and the major structural component of musculoskeletal tissues such as tendons, ligaments, and cartilage [1]. Within tendons, fibrous collagen exists in a hierarchical structure, with the collagen molecule as the base unit. The triple helical collagen molecule can become damaged in the form of collagen denaturation in injuries and diseases [2]. One form of this denaturation involves pullout or scission of one or more of the three amino acid collagen strands that makes up the triple helical collagen molecule. A tool that recently became available for easily characterizing such denatured collagen is collagen hybridizing peptide (CHP), which is a short polypeptide chain with detection moieties that bind selectively to denatured, but not intact collagen [3, 4]. Our lab recently used fluorescently labeled CHPs to demonstrate that collagen denaturation occurs in tendons due to mechanical overload during monotonic stretch and creep fatigue [2, 5, 6]. Denatured collagen accumulated in tendons even when stretched to sub-failure levels of strain, indicating that collagen denaturation is an important feature of tendon failure.

We previously established that when tendons are monotonically stretched, the onset of collagen denaturation occurs at the yield point of the stress-strain curve [5]. This was true for both positional and energy-storing tendons, indicating that the fundamental mechanism of failure is the same in both tendon types. This is despite biomechanical and biochemical differences between the two tendon types which reflect their physiological functions [7]. Positional tendons position joints, and are generally stiffer and fail at lower strain levels compared to energy-storing tendons. Energy-storing tendons store and release energy during movement, and can endure greater levels of extension compared to positional tendons [7–9]. Furthermore, positional and energy-storing tendons generally feature different enzymatic crosslinks profiles. Energy-storing tendons feature predominantly trivalent crosslinks and generally have a greater overall crosslink density compared to divalent crosslinks in positional tendons [8–11]. The greater crosslink content and valency found in energy-storing tendons leads to more brittle failure compared to plastic failure in positional tendons, which was demonstrated by the higher amount of denatured collagen measured in positional tendons when stretched to failure compared to energy-storing tendons [5]. This difference in crosslink density influencing failure mode has been observed even in positional common digital extensor and energy-storing superficial digital flexor tendons, which both play weight-bearing roles in mammals [8, 10]. Therefore, crosslink density is a defining feature differentiating positional and energy-storing tendons.

All of our previous studies have investigated the effect of tissue level load on the collagen molecule [2, 5, 6]. However, load transfer throughout the collagen hierarchy in tendons and its effect on the collagen molecule is poorly understood. For example, the micro-scale (fiber and fibril) strain to tissue level strain ratio is less than one for both positional and energy-storing tendons [12]. This is partially due to complex interactions between the tendon sub-structures, such as interfiber and interfibril sliding during tissue elongation [8, 13]. Regardless, collagen molecule denaturation occurring at the yield point of the stress-strain curve during tensile stretch was the common defining feature in both tendon types that we tested [5]. Computational models of collagen microfibrils predicted that collagen crosslinks facilitate collagen molecular stretching and resist intermolecular sliding until the yield point, at which point, the shear-dominant unfolding mechanism occurs, leading to α-chain pull-out creating binding sites for CHPs [2]. Collagen fibrils are one level above the collagen molecule in the collagen hierarchy. Therefore, investigating collagen molecular denaturation during tensile overload of individual collagen fibrils would allow more direct probing of the effect of tension on the collagen molecule and its crosslinks.

Collagen denaturation at sub-failure levels of strain has been previously demonstrated in collagen fibrils from positional tendons by Iqbal et al [14]. This study provided evidence that, like tendon level tests, the onset of collagen denaturation in collagen fibrils occurs before tensile failure. While they did not associate the onset of collagen denaturation with a point on the stress-strain curve, they saw an increase in CHP fluorescence relative to applied strain at approximately 10%, which closely correlates with the yield point on the stress-strain curve when the same fibril type was tested to rupture by Quigley et al [9]. Computational models of collagen fibrils also showed that during the linear region of the stress-strain curve, collagen enzymatic crosslinks resist intermolecular sliding until the yield point [15]. This was found to be true regardless of the different crosslink profiles found in positional and energy-storing tendons. Therefore, similar to tissue level tests, collagen fibrils seemed to exhibit increased amounts of denatured collagen when they are stretched to strain levels beyond the yield point. Since we previously established that the onset of collagen denaturation during tendon level load occurs at the yield point, we hypothesized that the same behavior may occur when stretching individual collagen fibrils.

The goal of this study was to establish the onset of collagen denaturation in collagen fibrils from both positional and energy-storing tendons in comparison to the stress-strain curve. We predicted that, like tissue level tests, we would see increased amounts of denatured collagen when collagen fibrils are stretched beyond the yield point of the stress-strain curve. We also expected the yield point to occur at a greater strain level in energy-storing fibrils compared to positional fibrils, because the greater crosslink density found in energy-storing tendons compared to positional tendons enables them to endure greater levels of molecular stretching and intermolecular sliding up until yield [8]. We also wanted know if the differing crosslink densities would lead to plastic and brittle failure in positional and energy-storing fibrils, respectively. We predicted that post-yield, lower crosslink densities in positional fibrils would favor intermolecular sliding, leading to lower a post-yield modulus and plastic deformation compared to energy-storing fibrils, which would favor molecular stretching, leading to a greater post-yield modulus and brittle deformation [15]. To study this, we tested collagen fibrils from the two tendon types using a microelectromechanical system (MEMS) device that was previously used for testing collagen fibrils [16–18].

2. Materials and Methods

2.1. MEMS fabrication

We used MEMS devices to perform tensile testing on collagen fibrils (Fig. 1). The MEMS devices were fabricated following a previously published method [16–18]. Briefly, silicon-on-insulator (SOI) wafers with buried oxide and top device layers of 2 μm and 4 μm, respectively, were purchased (Ultrasil LLC, Hayward, CA). Photoresist was applied to the wafers and a photomask and photolithography were used to etch the MEMS design onto the wafer. The wafer was plasma etched with SF₆, and the wafer was then diced into squares with each square consisting of 9 MEMS devices in a 3×3 arrangement. The MEMS squares were released by wet etching with 49% hydrofluoric acid for 20 min. to free the moveable pads of the MEMS devices and then thoroughly rinsed with isopropyl alcohol. Polystyrene microbeads (Thermo Scientific Opti-Link, Waltham, MA), 0.9 μm in diameter, were injected underneath the moveable pads to prevent stiction between the moveable pads and the wafer substrate. The MEMS squares were then critical point dried to yield working MEMS devices.

Figure 1:

A schematic of the MEMS device used for collagen fibril tensile testing. The gray and blue pads represent the anchored and moveable pads, respectively. A collagen fibril (blue line) is adhered to the MEMS device using epoxy droplets (yellow dots). The sample spans a test region, which is attached to a force gauge beam that deflects based on the amount of applied load. The sample also spans a control region that does not undergo loading. A tungsten probe in the pinhole holds the moveable pad in place during testing.

2.2. Fibril isolation and mounting

Rat tails and rat hindlimbs from 12–16 week old male Sprague-Dawley rats were purchased (BioIVT, Westbury, NY). Positional rat tail tendon fascicles (RTTF) and energy-storing flexor digitorum longus (FDL) tendons were dissected from the rat tails and hindlimbs, respectively. The enzymatic and non-enzymatic crosslink content in tendons changes depending on age and the type of tendon, which can affect their material properties. Non-enzymatic advanced glycation end-product (AGE) crosslink content was reported to be similar between 16-week old tendons and 4-week old tendons, indicating a lack of age-induced AGE accumulation [11]. Although the FDL tendon has not been shown to store and return energy during motion specifically in rats, it has been demonstrated in other species [19–21]. Additionally, the biochemical and biomechanical properties of rat FDL tendons are representative of energy-storing tendons. Published results indicate that the 12-week old positional RTTFs and 16-week old energy-storing Achilles tendons feature primarily divalent and trivalent enzymatic crosslinks, respectively, as measured by acid solubility following sodium borohydride reduction and HPLC analysis [11, 22]. It is unclear whether the crosslink content measured in rat Achilles tendons can be extended to the rat FDL tendons used in this present study. However, we are confident that the crosslink densities of RTTF and FDL tendons are representative of positional and energy-storing tendons, respectively. This is based on biomechanical differences we observed in our previous study [5] comparing these two tendons, where we saw biomechanical behavior representative of positional and energy-storing tendons (i.e. plastic vs. brittle deformation). Furthermore, other studies comparing bovine positional and energy-storing tendons found that energy-storing flexor tendons had more thermally stable crosslinks and greater overall crosslink density compared to positional extensor tendons, when characterized using hydrothermal isometric tension [10]. Our previous study [5] also saw energy-storing FDL tendons exhibit greater resistance to heat denaturation compared to positional RTTFs, despite being heated for longer amounts of time. While thermally stable crosslinks are a combination of trivalent and divalent (ketoamine) crosslinks, based on our previous results and data published by other groups, we believe that FDL tendons likely contain trivalent crosslinks and at least have a greater overall crosslink density compared to RTTFs.

One RTTF and one FDL tendon from separate rats were used to isolate all positional and energy-storing fibrils tested in this study, respectively. A scalpel was used to slice and scrape the surface and bulk of the tendon, disrupting the epitenon and interfascicular matrix, so that the fibrils could be separated from the bulk tissue. The tendon was agitated in phosphate buffer solution (PBS), causing it to swell. The swollen tissue was then transferred to a glass petri dish. PBS (1.5 mL) was placed on the swollen tissue, and the tissue was then combed with tweezers, causing collagen fibrils to be released into the PBS solution [23]. The solution containing collagen fibrils was collected. The collagen fibrils were visualized using darkfield microscopy (BX51WI, Olympus, Waltham, MA), and individual collagen fibrils were pulled from the solution using micromanipulators (MMO-203 and MMN-1, Narishige, Amityville, NY) with attached micropipettes. The micropipettes were custom made from glass capillaries (TW150F-3, World Precision Instruments, Sarasota, FL) using an automatic pipette puller (PC-10, Narishige, Amityville, NY). The collagen fibrils were placed on the MEMS device, and epoxy micro-droplets (50:50 mixture of Devcon 5 Minute and 2 Ton Clear epoxies, ITW Performance Polymers, Danvers, MA) were deposited using micropipettes to adhere the fibril to the MEMS device. The collagen fibrils became dehydrated in air as the epoxy was allowed to cure overnight.

2.3. CHP staining and imaging

The MEMS squares with mounted collagen fibrils were stained with CHP prior to testing to establish the baseline amount of denatured collagen in unstretched fibrils. Our previous studies have all used a fluorescein conjugated CHP (F-CHP) [2, 5, 6]. However, during preliminary studies, we observed significant photo-bleaching of the fluorescein. This prompted us to use a biotinylated CHP in conjunction with a streptavidin conjugated Alexa Fluor, which has greater resistance to photo-bleaching. Endogenous biotin was first blocked using a commercially available kit (E21930, Invitrogen, Waltham, MA) to ensure that any biotin that may be present on the collagen fibrils did not bind to streptavidin Alexa Fluor. The MEMS squares were then stained overnight in a 3 mL solution of 15 μM biotinylated CHP (B-CHP, 3Helix Inc., Salt Lake City, UT) in PBS at 4°C, with gentle agitation. The MEMS squares were rinsed three times, ten min. each time, with PBS at 4°C to remove unbound CHP. They were then stained with 3 mL of a 5 μg/mL solution of streptavidin Alexa Fluor 488 (S11223, Invitrogen, Waltham, MA) in 1% bovine serum albumin (BSA) at 4°C for two hr. The MEMS squares were then rinsed ten times, ten min. each time, with PBS at 4°C to remove excess streptavidin Alexa Fluor 488 that had not bound to biotin. The rinse times and numbers for both B-CHP and streptavidin Alexa Fluor 488 were determined during pilot studies based on when the fluorescence intensity of CHP stained heat denatured collagen fibrils stopped changing. The CHP stained collagen fibrils were imaged using an upright microscope (BX51WI, Olympus, Waltham, MA) with a water immersion 60× objective using a LED fluorescent light source (X-Cite XYLIS, Excelitas Technologies Corp., Waltham, MA) and a FITC filter cube (Fig. 2).

Figure 2:

Representative images of a collagen fibril on the MEMS devices. Left: A brightfield image of a collagen fibril adhered with epoxy droplets. Right: A fluorescent image of the collagen fibril when stretched beyond the yield strain stained with CHP. The outline of the MEMS device (violet) was drawn for clarity. The portion of the sample in the tested region had greater fluorescence intensity than in the control region. Scale bar: 65 μm

2.4. Fibril tensile testing

The MEMS square was adhered using double-sided tape to a piezoelectric stage (Nano-H100, Mad City Labs, Madison, WI) and the collagen fibrils were rehydrated in PBS for at least 30 min. prior to testing occurring in PBS. A tungsten probe was placed in the pinhole of the MEMS device to hold the moveable pad in place. While the pad was held in place, a displacement was applied with the piezoelectric stage, applying a displacement to the sample. The portion of the sample in the test region of the MEMS device was attached to a force gauge beam, which deflected based on the amount of load applied to the sample (Fig. 1). Therefore, the total displacement applied by the piezoelectric stage was the sum of the displacements of the sample and the force gauge beam. As such, there was no way of knowing the amount of strain that was applied until the test was concluded, since the sample and force gauge beam displacements were coupled. Therefore, a wide elongation range was used to create a wide range of applied strains on both sides of the yield point of the stress-strain curve. The displacement levels that were selected ranged from 0 to 12 μm at 0.5 μm increments, for a total of 25 different displacement levels. One of each of these 25 displacement levels was applied to an individual fibril from both tendon types at a rate of 0.05 μm/sec using the piezoelectric stage. Therefore, a total of 50 collagen fibrils (n=25 per tendon type) were tested. After the target displacement was reached, the fibril was unloaded. The CHP staining and imaging steps were repeated as above after mechanical testing to quantify the denatured collagen that arose due to mechanical load.

The MEMS devices were sputter-coated with 10 nm of Au/Pd and imaged under vacuum using scanning electron microscopy (Quanta 600F, FEI Company, Hillsboro, OR). The diameter of the collagen fibril in the control region (where the fibril did not undergo testing) was measured at three separate points and averaged to calculate the mean diameter of each tested collagen fibril. The mean diameter of the collagen fibril under vacuum was multiplied by 2.2 to calculate the hydrated collagen fibril diameter, based on published results [18].

2.5. Data analysis

CHP fluorescence intensity was measured to quantify the amount of denatured collagen in the collagen fibrils. It is reasonable to assume that CHP fluorescence is linearly related to the amount of denatured collagen molecules. This is based on results in our previous studies finding a linear relationship between CHP fluorescence and percent of collagen denatured using the trypsin-hydroxyproline assay. The trypsin-hydroxyproline assay quantifies the amount of hydroxyproline in denatured and intact collagen fractions, and hence directly measures denatured collagen molecules [5, 24, 25].

During preliminary studies, we found that there was inhomogeneous CHP fluorescence along the length of the collagen fibrils (i.e. the test region vs. the control region) before mechanical testing. This is likely due to differences in the amount of denatured collagen along the fibril. We tested several different methods of quantifying increased amounts of denatured collagen after mechanical load, one of which accounted for the fluorescence intensity in both the control and test region both before and after testing. Ultimately, we found no difference in trend (increased collagen denaturation vs. strain) between this method and the method we describe below; we used the latter method since it was more intuitive and introduced less potential error by analyzing only the test region.

The test region was defined as the region of interest for each sample, and a threshold value for this region was calculated using Otsu’s method [26]. The mean intensity of pixels above the threshold value was calculated for the pre-test and post-test images for each sample. During preliminary testing, we observed photo-bleaching despite our best efforts and using Alexa Fluor 488 instead of fluorescein. Therefore, we carefully characterized the level of photo-bleaching relative to the number of imaging events, and found that each imaging event reduced the fluorescence intensity by 10%. The post-test and pre-test intensities were adjusted accordingly to account for the drop in fluorescence during imaging (Supplementary Information). The post-test intensity was divided by the pre-test intensity to calculate the Fluorescence Ratio. This value represents the relative fluorescence increase from additional CHP binding, indicating the amount of denatured collagen purely caused by mechanical load.

The amount of force applied to the collagen fibril was calculated based on the displacement of the force gauge beam, which is a non-linear relationship. Therefore, a model of the MEMS device was created and analyzed using FEBio [27]. The model used TET10 elements [28] and the material properties of polysilicon (Young’s modulus of 160 GPa; Poisson’s ratio of 0.22) [17]. The SOI wafers used to construct the MEMS devices were intended for semiconductor applications, and are of high purity. Therefore, the polysilicon was modeled as a linear isotropic material and the material properties used in the model are well established. The hydrated collagen fibril diameter and the predicted applied force from the finite element analysis were used to calculate the stress applied to each sample. The stress and elongation were used to construct a stress-strain curve for each sample. Each individual stress-strain curve was analyzed to calculate the peak modulus, yield point, and post-yield modulus. The peak modulus was determined by the maximum slope of a moving linear fit over a 2% strain window [11]. We defined the yield point to be when the peak modulus had reduced to 50% of its maximum value [5, 29]. We calculated the post-yield modulus of fibrils that reached a yield point, by determining the minimum slope of the moving linear fit over a 2% strain window beyond the yield point. The percentage of the post-yield modulus out of each fibril’s peak modulus was also calculated. Data analysis was performed using MATLAB (R2019a, MathWorks, Natick, MA).

2.6. Acid solubility

The acid solubility protocol was adapted from Svensson et al. [11] and performed on both RTTF and FDL tendons to characterize the trivalent crosslink density of the two tendon types. Five RTTF were dissected from a single rat tail, and five FDL tendons were dissected from three separate rats. The tendons were cut such that the wet weight was approximately 1.0 g. The tendons were placed in vials and treated in 1.0 mL of 50 mM acetic acid, under gentle agitation, at 4°C overnight. The samples were then centrifuged; 0.5 mL of the supernatant was transferred to new vials, leaving 0.5 mL of supernatant and the remaining tissue. The amount of hydroxyproline in the two vials was quantified using a previously described protocol [25, 30]. In keeping with the protocol used by Svensson et al., the acid solubility was calculated as the ratio of the hydroxyproline in the top 0.5 ml of supernatant divided by the hydroxyproline in the bottom 0.5 ml of supernatant plus the pellet. We express this ratio as a percentage.

2.7. Statistical analysis

Power analysis was performed using G*Power 3.1 [31] to determine the number of samples needed for CHP quantification in the groups that did and did not reach yield. A sample size of 10 was calculated from preliminary studies using an effect size of 1.40 and two-tailed α of 0.05 to provide a power of 0.80. All values presented represent Mean±SD. Datasets were tested for normality using the Shapiro-Wilk test. Statistical significance (α = 0.05) was tested for using a two-tailed t-test between the mean Fluorescence Ratio values for samples that did and did not yield, in addition to the yield strain, yield stress, post-yield modulus, post-yield modulus percentage, and the fibril diameters between collagen fibrils from the two tendon types. The difference in peak modulus between collagen fibrils from different tendon types was tested using the Mann-Whitney U Test, with α = 0.05, because the two datasets failed normality. These statistical comparisons were performed using SigmaPlot 13.0 (Systat Software, Inc., San Jose, CA).

When measuring acid solubility, all tail tendons came from one rat (Rat 1), while the five FDL tendons came from three separate rats (two from Rat 2, two from Rat 3, and one from Rat 4). Therefore, the acid solubility for the two tendon types was compared using a nested t-test (mixed effects linear regression) using GraphPad Prism (9.4.1, GraphPad Software, San Diego, CA), with significance set at α = 0.05.

3. Results

Collagen fibrils from both positional and energy-storing tendons had visibly greater fluorescence intensity in the test region compared to the untested control region when they were stretched beyond the yield point and stained with CHP (Fig. 2, right).

Fibrils from both positional RTTF and energy-storing FDL tendons were stretched to varying levels of strain. Four of the 25 total fibrils within each tissue type ruptured during testing. These samples were discarded due to unreliable fluorescence quantification from the recoiled fibril ends. The final strain levels for the 21 RTTF collagen fibrils that did not rupture ranged from 0 to 120% (Fig. 3). Two of the positional fibrils that did not reach yield began to display strain hardening during stretching, while the rest exhibited linear behavior. One of these fibrils was considered an outlier based on its peak stress value, and discarded. The final strain levels for the 21 FDL tendon collagen fibrils that did not rupture ranged from 0 to 150% (Fig. 4). Two of the fibrils that were stretched beyond yield exhibited both strain hardening and softening, while the remaining fibrils exhibited strain softening after reaching yield.

Figure 3:

Stress-strain curves for collagen fibrils from positional RTTF. Left: Stress-strain curves for samples that did not reach the yield strain (n=9). Most samples feature the typical linear behavior with no softening, while some exhibit strain hardening. Right: Stress-strain curves that reached a yield strain (n=11). All fibrils exhibited softening behavior as the modulus decreases and the sample undergoes plastic deformation.

Figure 4:

Stress-strain curves for collagen fibrils from energy-storing FDL tendons. Left: Stress-strain curves for samples that did not reach the yield strain (n=11). Most samples feature the typical linear behavior with no softening, with one exhibiting strain hardening. Right: Stress-strain curves that reached a yield strain (n=10). All fibrils exhibited softening behavior as the modulus decreases and the sample undergoes plastic deformation.

The Fluorescence Ratio increased with additional applied strain beyond the yield point for collagen fibrils from both tendon types (Fig. 5). The Fluorescence Ratio values for RTTF collagen fibrils that did and did not reach yield were 20±20 (n=11) and 3±1 (n=9), respectively (Fig. 5). The difference between these two groups was statistically significant (p = 0.0034). The Fluorescence Ratio values for FDL collagen fibrils that did and did not reach yield were 6±4 (n=10) and 2±1 (n=11), respectively (Fig. 6). The difference between these two groups was also statistically significant (p = 0.0018).

Figure 5:

Fluorescence Ratio vs. Strain. Left: Data for RTTF fibrils. Right: Data for FDL fibrils. Generally, fluorescence increases relative to applied strain when stretched past the yield point for both tendon types.

Figure 6:

The mean Fluorescence Ratio for stretched collagen fibrils from both positional RTTF and energy-storing FDL tendons that did and did not reach yield. In fibrils from both tendon types, the difference in Fluorescence Ratio between samples that did and did not reach yield was statistically significant (α < 0.05). There was less relative fluorescence increase in FDL fibrils compared to RTTF fibrils. Mean±SD.

The peak modulus, yield strain, yield stress, post-yield modulus, post-yield modulus percentage, and collagen fibril wet diameter were compared between collagen fibrils from the two tendon types (Table 1). Collagen fibrils from energy-storing FDL tendons had a greater yield strain and yield stress compared to collagen fibrils from positional RTTF tendons. FDL collagen fibrils also had a smaller wet diameter compared to RTTF collagen fibrils. These differences were all statistically significant. However, there was no significant difference in peak modulus between collagen fibrils from the two tendon types. The post-yield modulus was greater for energy-storing compared to positional fibrils, but this difference was not statistically significant (p = 0.155). When the post-yield modulus was calculated in terms of percentage of the fibril’s peak modulus, it was significantly greater in energy-storing fibrils compared to positional fibrils (p = 0.0237).

Table 1:

Material properties of the collagen fibrils from positional RTTF and energy-storing FDL tendons. The peak modulus for both fibril types was the same. Collagen fibrils from energy-storing tendons exhibited a greater yield strain, yield stress, and post-yield modulus in terms of percentage of the fibril’s peak modulus. The post-yield modulus of energy-storing fibrils was also greater than positional fibrils, but the difference was not statistically significant. Collagen fibrils from positional tendons had a greater wet fibril diameter. Mean ±SD.

	Positional	Energy-storing	p - value
Peak modulus (MPa; Mean±SD)	200±100	200±100	0.313
Yield strain (%; Mean±SD)	24±7	50±20	0.00228
Yield stress (MPa; Mean±SD)	20±10	60±20	<0.001
Fibril wet diameter (nm; Mean±SD)	500±100	350±90	<0.001
Post-yield modulus (MPa; Mean±SD)	50±30	70±50	0.155
Post-yield modulus, percentage of fibril’s peak modulus (%; Mean±SD)	33±6	40±9	0.0237

The mean acid solubility of RTTF (Rat 1) was 92±8%; the mean acid solubility of FDL tendons from Rats 2, 3, and 4 were 13±2%, 23±1%, and 4%, respectively (Fig. 7). The difference in acid solubility between the two tendon types was statistically significant (p = 0.016).

Figure 7:

Acid solubility of positional RTTF and energy-storing FDL tendons. RTTF tendons were significantly more susceptible to acid solubility compared to FDL tendons (p = 0.016).

4. Discussion

We found that collagen fibrils from both positional and energy-storing tendons exhibit increased amounts of denatured collagen indicated by increased CHP fluorescence when they are stretched beyond the yield point of the stress-strain curve, compared to collagen fibrils that were not stretched beyond yield. This is despite the different biochemical enzymatic crosslink content reported by others in the two tendon types, and which we confirmed in this study using acid solubility [11]. RTTF tendons were significantly more susceptible to acid solubility compared to FDL tendons, indicating that the latter contains a much greater trivalent crosslink density compared to the former. The acid solubility values for both tendon types were very similar to those of age-matched positional and energy-storing tendons (RTTF and Achilles, respectively) reported by Svensson et al. [11]. They also quantified the amount of trivalent crosslinks using HPLC in these tissues, and found that the Achilles tendons had significantly more trivalent crosslinks compared to RTTF tendons. Therefore, it is reasonable to assume that the RTTF and FDL tendons used in this study also feature significantly different trivalent crosslink densities. Computational results of 20 nm diameter fibrils have indicated that the crosslinks, regardless of type, resist intermolecular sliding of collagen molecules during tension until the yield point [15]. Molecular dynamics simulations previously published by our group predict that at yield point, the shear-dominant unfolding mechanism occurs, leading to α-chains in the collagen molecule to slide out, exposing binding sites for CHP [2]. Assuming no stacking faults or other significant inhomogeneities exist in the much larger real-world fibrils we tested, the results of this experiment suggest that this is the failure mechanism in collagen fibrils regardless of tendon type, which is consistent with our previous study using tendon fascicles [5].

Interestingly, despite the differing crosslink densities between the two tendon types, there was no difference in peak modulus between collagen fibrils isolated from positional and energy-storing tendons. There was a considerable range in peak modulus, although this is often encountered when testing collagen fibrils [16, 17]. Both computational and experimental results have also previously shown no difference in peak modulus between collagen fibrils with different enzymatic crosslink profiles [11, 15]; however, the yield strain and yield stress for energy-storing collagen fibrils were nearly double those of positional collagen fibrils. Depalle et al. [15] suggested that the greater crosslink densities such as those found in energy-storing tendons compared to positional tendons are able to endure greater levels of molecular stretching and stress before intermolecular sliding and the shear-dominant unfolding of the collagen triple helix. Consequently, these differing crosslink densities played an important role in the post-yield behavior of the two fibril types. While pre-yield deformation is dictated by molecular stretching, post-yield behavior is a combination of both molecular stretching and intermolecular sliding [15]. Positional fibrils exhibited plastic deformation, indicated by the lower post-yield modulus compared to energy-storing fibrils, which exhibited brittle deformation. The lower crosslink densities in positional fibrils favored intermolecular sliding, leading to greater amounts of denatured collagen compared to energy-storing fibrils, which favored molecular stretching due to greater crosslink densities enhancing intermolecular connectivity.

There was a greater amount of denatured collagen post-yield as indicated by greater Fluorescence Ratio values in collagen fibrils from positional tendons compared to energy-storing tendons. This is consistent with tissue level results [5], and is likely due to the greater crosslink density in energy-storing tendons compared to positional tendons [8]. Miles et al. [32] demonstrated that increasing the crosslink density in positional rat tail tendons increased the collagen denaturation temperature by drawing collagen molecules closer together, which led to dehydration via the polymer-in-a-box mechanism [33]. While that work did not find this crosslink-mediated dehydration to alter the enthalpy of denaturation, hydration is known to play an important role as a plasticizer in collagen constructs [34]; it has been suggested that this could mechanistically be due to water enabling intermolecular collagen sliding [17]. A recent study demonstrated a lack of CHP binding in dehydrated fractured bone samples compared to hydrated bone samples, supporting the notion that hydration is necessary for intermolecular sliding which precedes the shear-dominant unfolding mechanism and collagen denaturation [35]. Energy-storing fibrils have a greater crosslink density, and therefore relatively less hydration and intermolecular sliding compared to positional fibrils. The polymer-in-a-box mechanism could therefore explain both the lower amounts of collagen denaturation that arose during tension and the smaller fibril diameters observed in energy-storing fibrils compared to positional fibrils. While it is possible that dehydrating the collagen fibrils during sample preparation may have affected the intermolecular interactions and were not fully recovered during rehydration, a previous study [36] showed that collagen fibril diameter is recovered following dehydration and rehydration, suggesting that its effect was minimal.

This is the first time that denatured collagen has been detected in collagen fibrils from energy-storing tendons as a result of mechanical load. This is in contrast to a previous study where Quigley et al. [9] demonstrated a lack of CHP binding in superficial digital flexor (SDF) energy-storing fibrils that were stretched to failure. This is possibly due to the higher strain rate that was used in their study. Whether or not collagen denaturation occurs in tendons during mechanical load is correlated with the transition from localized point failure to plastic failure. This transition has been shown to be a function of both crosslink density and strain rate; tendons with lower crosslink densities stretched lower strain rates are more likely to exhibit collagen denaturation [8]. Quigley et al. tested energy-storing collagen fibrils at a strain rate of about 2%/sec, whereas this study used a strain rate of about 1%/sec. While this difference appears small, Chambers et al. [8] predicted that in energy-storing tendons, this transition occurs between fibril-level strain rates of 0.4–2%/s. This is based on their tissue-level tests of SDF tendons, which found signs of collagen molecular damage using differential scanning calorimetry (DSC) and SEM when the tendons were stretched at 1%/s, which equates to a fibril level strain of 0.4%/s. Therefore, it’s likely that if SDF collagen fibrils were stretched at a low enough strain rate, collagen denaturation would eventually be observed. Similarly, it is possible that stretching energy-storing fibrils at even lower strain rates than those used in this study would yield increasing amounts of denatured collagen, although whether the amount would plateau and at what strain rate is unclear. In vivo strain rates for Achilles tendons during walking is between 1.7 and 10%/s [37]. A fibril strain rate of 1%/s would correspond with a tendon strain rate of 2.5%/s, so the strain rates used in this study corresponds with those seen in energy-storing tendons during daily use. As for collagen fibrils from positional tendons, CHP binding due to sub-failure tensile loads had been previously demonstrated by Iqbal et al [14]. They demonstrated that the amount of collagen denaturation depends on applied strain, which is consistent with our study. However, this is the first study to correlate the onset of mechanical collagen denaturation with the yield point of the stress-strain curve.

Interestingly, very few collagen fibrils tested in this study exhibited a strain-induced strengthening behavior after the yield point, although this second peak modulus was predicted by a computational study in collagen fibrils with sufficiently high crosslink densities [15] and experimentally observed in collagen fibrils with induced advanced glycation end product crosslinks, and fibrils from energy-storing flexor tendons of 24–36 month old steers and human patellar tendon [9, 11, 22]. However, Svensson and coworkers reported that only 1 out of 11 collagen fibrils from 16-week old rat energy-storing Achilles tendons exhibited a second peak modulus, while none of 7 collagen fibrils from 16-week old rat tail tendon fascicles exhibited a second peak modulus [11]. It is unclear how the crosslink density in tendons from 16 week old rats compares to 24–36 month old steers and human patellar tendon, but based on computational results by Depalle et al. [15], it is reasonable to assume that the latter has greater crosslink densities compared to the former.

While the strain levels observed in this study are consistent with strain levels seen in previous applications of the same MEMS device [16, 17], the strain levels are generally greater than those seen using other MEMS designs or different collagen fibril tensile testing methods [9, 11, 14, 38, 39]. For example, of the collagen fibrils that were stretched beyond the yield strain in this current study, most of them had a final strain level greater than 50%. On the other hand, a different MEMS device [38, 39] produced final strain levels of 30–50%; the bowstring method [9] led to fibrils fail at ~30% strain; and a custom tensile test rig by Svensson et al. [11] led to fibrils failing at 8.9% strain. These tests had initial fibril lengths of 30, 50, and 270 μm, respectively [9, 11, 38, 39]. Therefore, one possibility for the difference in final strain levels across these experiments is the different initial gauge lengths used; this current study had initial gauge lengths of 5–10 μm, which is comparatively very short compared to the other fibril tests. The MEMS device was designed with shorter gauge lengths in mind to avoid point defects which affect material properties, while ensuring the sample was long enough to behave as a representative volume element. Molecular dynamics simulations predict that collagen fibrils containing more than 10 collagen molecules in length behave as representative volume elements, which corresponds with approximately 3 μm in length [40]. Therefore, our sample length of 5–10 μm behaves as a representative collagen fibril, while avoiding point defects in the sample. The lack of point defects in the short gauge lengths decrease the chance the fibril failed prematurely, resulting in greater strain levels [11, 22]. Other groups have seen similar behavior where the failure strain decreased when fibrils with longer initial gauge lengths were used [11, 22]. By using shorter gauge lengths and hence reducing the chance that point defects exist in this test region, we were able to more closely examine the effect of crosslink density on collagen fibril mechanics by preventing early failure. Furthermore, different collagen fibril preparation methods different groups used may have also led to different strain levels. Some experiments used reconstituted collagen fibrils, which are unlikely to have enzymatic crosslinks, leading to lower failure strains [39]. Others aggressively dried their fibril samples with nitrogen, which may lead to embrittlement and early failure [11, 14].

Unintuitively, the fibril strain levels in this study are also much greater than tendon level strains, although this is consistent with previous collagen fibril tensile tests using this MEMS device [16, 17]. It is unlikely that collagen fibrils experience the high levels of strain seen in this study in vivo, and hence indicates that other potential mechanisms behind the etiology of tendon injuries exist. For example, point weaknesses existing along the fibril length are a possible explanation for the strain discrepancies; during tissue level tests, the entire fibril is loaded, engaging all point weaknesses, which would lead to lower failure strain levels than those seen in this experiment. Multiphoton images of CHP strain RTTFs revealed that collagen denaturation during tension is heterogeneous; therefore, tendon failure may occur at stress concentrations at point weaknesses in tendons where collagen denaturation is more susceptible. Additionally, when a tendon is loaded, non-collagenous constituents (e.g. proteoglycans, interfascicular matrix) may play a role in resisting tension; if these constituents suddenly failed, this would cause the collagen fibrils to suddenly experience increased load, providing another possible failure mechanism at the point weaknesses. The strain discrepancy between the fibril and tissue levels may also be due to the lack of shear stress introduced by the MEMS device during fibril level tests. The bowstring method introduces combined shear and tensile loads at the loading point, causing early failure [9]. Interfibrillar shear stress has been shown to be the loading mechanism of collagen fibrils in tendon, possibly mediated by small diameter collagen fibrils [41, 42]. The MEMS device applies load purely in tension, so the effects of shear force on collagen fibrils which occur in vivo and when using the bowstring method were not seen. Regardless, the overall shape of the stress-strain curves seen in this study are consistent with previous work when accounting for the sample source and testing method [11, 16].

There are differences in fibril mechanics between this study and previous studies, depending on the sample preparation. The yield stresses calculated in this study were 16–80 MPa, which are 16–40% of the range of values observed by other groups of (100–200 MPa) [9, 11, 22]. This is likely because the stress values in the previous studies were calculated using the dry fibril diameter, whereas we used a wet fibril diameter which we defined as 2.2 times the dry fibril diameter [18]. This results in a cross-sectional area that is 4.84 times greater for wet fibrils compared to dry fibrils; if we were to also use the dry fibril diameter, the yield stress values would be similar to previous results. The linear region moduli measured in this study are also consistent when compared to previous studies using the MEMS devices [16, 17]. The values measured using this method are generally lower than those measured by other methods, even when accounting for dry vs. wet fibril diameter [9, 11]. One potential explanation is the loading mechanism; Quigley et al. [9] loaded their fibrils using the bowstring method, introducing shear, which may have caused different modulus values. Another possible explanation is sample preparation; Svensson et al. [11] dried their fibrils using nitrogen, which likely caused embrittlement of the samples, leading to higher moduli.

A potential limitation of this study is the use of PBS. Although PBS is extensively used during the mechanical testing of biological tissues, it can cause the tissues to swell, potentially affecting the intermolecular water bridges. Furthermore, the samples were handled extensively throughout the experiment, with the multiple staining and rinsing steps. This may have inadvertently caused mechanical damage, although all samples were handled exactly the same way. Finally, due to the MEMS design, the exactly strain rate could not be controlled, and the final strain level could not be prescribed. We are confident that the strain rate was slow enough to represent quasi-static loading, especially since we saw collagen denaturation in the energy-storing fibrils. However, more careful examination of the effect of strain rate on collagen molecular damage and more precise correlation between the final strain level and CHP fluorescence will require design modifications in the future.

Establishing the onset of collagen denaturation in collagen fibrils relative to the stress-strain curve opens the opportunity for other collagen fibril tensile experiments using CHP and the MEMS device. The collagen fibril testing protocol could be modified to introduce shear forces during the fibril test, allowing investigation of shear force’s effect on the yield strain and collagen denaturation. Our lab has previously demonstrated that, during creep fatigue, the onset of collagen denaturation in tendons is closely correlated to the yield strain [43]. An approach similar to the one employed in this study could be used to test whether this behavior is preserved at the fibril level. While strain-level cyclic tests on collagen fibrils have previously been performed [38, 44], creep fatigue tests have not, and collagen denaturation has not been directly quantified during cyclic tests. With the advent of CHPs in recent years allowing direct probing of molecular level damage in collagen, combined with various techniques available for testing collagen fibrils, the effect of tension on the collagen molecule with varying crosslink densities and loading conditions can be closely examined.

5. Conclusion

We demonstrated that collagen fibrils from both positional and energy-storing tendons exhibit increased amounts of denatured collagen when they are stretched beyond the yield point of the stress-strain curve. Collagen fibrils from energy-storing tendons had a greater yield strain and yield stress compared to fibrils from positional tendons. Despite this, energy-storing fibrils exhibited less collagen denaturation when stretched beyond yield compared to positional fibrils, which is representative of the brittle vs. plastic failure seen between the two tendon types. These differences indicate that collagen fibrils from the two different tendons experience different degrees of molecular stretching before the shear-dominant unfolding mechanism occurs, which creates binding sites for CHP. Collagen denaturation in collagen fibrils due to mechanical load as indicated by CHP binding is also a strain mediated process, similar to tissue level tests. This indicates that the onset of collagen denaturation in synchronized at the yield point at both the fibril and tendon level during tissue level load. The difference in strain levels at the tissue and fibril levels during the onset of collagen denaturation indicates that heterogeneous loading mechanisms may affect fibril-level load when they are embedded in the collagen hierarchy in tendons. Furthermore, our previous study [2] has indicated that the onset of collagen denaturation is heterogeneous within tendons. This, combined with studies that show that tendons can be loaded heterogeneously in vivo [45], indicates that tendon loading may be more complicated than pure tension. Therefore, this study provides insight into how load is transferred throughout the collagen hierarchy in tendons during mechanical load, and how crosslink density affects collagen denaturation during tension, leading to improved understanding of the etiology behind injuries of different tendon types.