Reasons for performing study: Damage to the flexor tendons, particularly the superficial digital flexor tendon (SDFT), is one of the most common musculoskeletal injuries sustained by horses competing in all disciplines. Our previous work has shown that SDFTs from different individuals show a wide variation in mechanical strengths; this is important clinically as it may relate to predisposition to injury. The high mechanical strength of tendon relies on the correct orientation of collagen molecules within fibrils and stabilisation by the formation of chemical cross-links between collagen molecules. It is not known whether the variation in SDFT mechanical properties between individuals relates to differences in collagen cross-link levels.
Hypothesis: Enzyme-derived, intermolecular cross-linking of tendon collagen correlates with mechanical properties of the SDFT.
Methods: SDFTs were collected from 38 horses and mechanically tested to failure. Structural and material properties were calculated from the load/deformation plot and cross-sectional area for each tendon. Following mechanical testing, pyrrolic cross-link levels were measured using a spectrophotometric assay for Ehrlich's reactivity and pyridinoline levels were quantified by HPLC. Cross-link levels were correlated with mechanical properties and statistical significance tested using a Pearson's correlation test.
Results: Pyrrole cross-link levels showed a significant positive correlation with ultimate stress (P = 0.004), yield stress (P = 0.003) and elastic modulus (P = 0.018) of the tendons, despite being a minor cross-link in these tendons. There was no significant correlation of mechanical properties with either hydroxylysyl- or lysyl-pyridinoline levels.
Conclusions: Given the low absolute levels of pyrrole, we suggest that the correlation with high mechanical strength is through an indirect mechanism. Understanding the nature of the relationships between pyrrole cross-links, other matrix characteristics and tendon material properties may allow development of strategies to identify horses at risk from tendon injury and be of value in informing training practices.
Tendon and ligament injuries are among the most common musculoskeletal injuries sustained by competition horses and represent a significant cause of morbidity and mortality. A study of injuries occurring on British racecourses found that 46% of all limb injuries in racehorses were to tendons and ligaments (Williams et al. 2001). The incidence rate is also high for horses in training, at 1.9/100 horse months, recorded in National Hunt training yards in the UK over 2 racing seasons (Ely et al. 2009). A similar picture is seen in other parts of the world; 15% of Thoroughbred flat racehorses in training in Japan in one year suffered from a tendon or ligament injury (Kasashima et al. 2004). On Kentucky racecourses 28.1% of musculoskeletal injuries were to the superficial digital flexor tendon (SDFT) or suspensory ligament (Peloso et al. 1994), while tendon injury was the most common reason for retirement in racing Thoroughbreds in Hong Kong (Lam et al. 2007). Although much less epidemiological information is available for horses competing in other disciplines, tendon injury is generally recognised as a significant problem. A study in event horses found that injuries to tendons and ligaments were the most common reason (43.4%) preventing horses from competing (Singer et al. 2008).
The majority of tendon injuries in racehorses (97–99%) occur to the forelimb tendons (Kasashima et al. 2004; Lam et al. 2007), with the SDFT being injured in 75–93% of cases (Kasashima et al. 2004; Ely et al. 2004, 2009). In event horses the SDFT was also the most commonly injured tendon or ligament (Singer et al. 2008). The high susceptibility of the SDFT to injury is not surprising when considering the function of this tendon. In addition to preventing hyper-extension of the metacarpo-phalangeal joint, the SDFT plays a vital role in energy storage and release thereby increasing the efficiency of locomotion (Alexander 1991). Energy storing tendons are required to stretch and recoil and are therefore subjected to high strains. Maximum in vivo strain for the common digital extensor tendon, a positional tendon, has been estimated at 2.5% (Birch 2007), which is almost 4 times lower than the failure strain of 9.7% recorded in vitro (Batson et al. 2003). In contrast, during gallop, in vivo strains of 16% have been recorded in the SDFT (Stephens et al. 1989), similar to the failure strains of 15–17% recorded in vitro (Dowling et al. 2002; Gerard et al. 2005). Therefore, the SDFT is working with a very low safety margin. However, despite this, some horses race and compete at the very top level without ever suffering from a tendon problem.
Our previous work has shown that equine SDFTs from different individuals show a wide variation in mechanical properties (Birch et al. 2000). The variation in tendon strength and stiffness could not be accounted for by horse size, breed or age. Furthermore the material properties of the tendon tissue (ultimate stress and elastic modulus) also showed considerable differences between horses. It is not known which aspect of tendon structure is responsible for imparting superior material properties, although this may relate to injury risk and therefore be of clinical significance.
The mechanical response of tendon and ligament to load is governed by their complex and highly organised hierarchical structure from nano- to macroscale. At the nanoscale collagen molecules are aligned longitudinally to form fibrils, which are embedded in a proteoglycan rich medium to form microscale fibres. Collagen fibres in turn pack to form macroscopic fascicles and finally fascicles form the whole tendon. This composite structure results in a mechanical response that is nonlinear and time dependant (Puxkandl et al. 2002). The high mechanical strength of tendon relies on the correct orientation of collagen molecules within the fibrils and their stabilisation by the formation of chemical cross-links between collagen molecules. These cross-links are formed at the ends of the collagen molecules in the telopeptide regions following the action of the enzyme lysyl oxidase on specific lysine and hydroxylysine residues. The formation of enzyme-derived cross-links increases the mechanical strength of the collagen fibrils (Avery and Bailey 2005).
The type of crosslink formed depends on the degree of lysine hydroxylation in the helical part of the molecule and nonhelical telopeptide region (for a review of collagen cross-linking see Bailey et al. 1998). Lysyl oxidase converts lysine and hydroxylysine residues into aldehydes; if hydroxylation is extensive in the telopeptide region more hydroxylysine aldehyde will be formed. Lysine aldehydes and hydroxylysine aldehydes undergo a spontaneous reaction with lysine or hydroxylsine residues in the helical region of neighbouring molecules to form the immature bivalent cross-links. Immature cross-links undergo further spontaneous reactions to form mature trivalent pyridinoline and pyrrole cross-links. Extensive hydroxylation of lysine residues throughout the collagen molecule results in the formation of hydroxylysyl-pyridinoline (HL-Pyr). In tissues where less hydroxylation of the helical or telopeptide lysine residues occurs the mature cross-links lysyl-pyridinoline (L-Pyr) and pyrrole form, respectively. A further mature crosslink, histidino-hydroxylysinonorleucine (HHL), found predominantly in skin but also present in some tendons, can form from the interaction between an immature cross-link and histidine residue. We have shown previously that HHL, although present in the equine extensor tendon, is not present in SDFT tissue (Birch et al. 2008a), suggesting a connection between crosslink type and tendon mechanical function.
In this study we tested the hypothesis that lysyl oxidase derived intermolecular cross-linking of tendon collagen is associated with the mechanical properties of the SDFT.
Materials and methods
Tendon collection and storage
The distal part of the left forelimb (n = 37) or right forelimb (n = 1) was collected from horses (n = 38) aged 1–28 years (mean ± s.d. 11.6 ± 7.2 years; median 10 years) destroyed for reasons other than tendon injury at a commercial equine abattoir or hunt kennels. The limbs from 16 horses were collected on the day of euthanasia and information on horse age, size, sex and type recorded. On return to the laboratory, the SDFT was dissected immediately from the fresh limbs after marking the mid-metacarpal level on the tendon. The cross-sectional area (CSA) at the mid-metacarpal level was measured using an alginate paste casting technique that has been shown previously to be accurate to within 0.8% (Goodship and Birch 2005). The fresh tendons were then mechanically tested to failure as described below. The remaining limbs (22 horses) were stored frozen prior to transport to our laboratory, thawed on arrival and the SDFT dissected from the limb after marking the mid-metacarpal level on the tendon. The dissected tendons were wrapped in cling film and stored at −20°C prior to mechanical testing. On the day of testing the tendons were re-thawed at room temperature and the CSA and mechanical properties measured in exactly the same way as the fresh tendons. We have previously shown that freeze thawing does not affect mechanical properties of equine tendons (unpublished data).
Mechanical testing and data analysis
Tendons were mounted vertically, with the proximal end uppermost, in a servo-hydraulic materials testing machine (Zwick Testing Machines Ltd)1 with a 50 kN load cell, using cryoclamps cooled with liquid CO2 (Riemersma and Schamhardt 1982) with the clamps set at 10 cm apart. The mid-metacarpal region of the tendon was centred between the clamps which gave a homogeneous length of tendon to be tested. The tendons were preloaded with 100 N, which represents a negligible load of approximately 1% of the failure load and allows determination of a resting length. The distance between the 2 freeze lines was measured to give the effective gauge length. Tendons were preconditioned to reach a steady state with 20 cycles of load between 100 N and 4 kN using a sine wave at a frequency of 0.5 Hz and returned to a 100 N pre-load. The tendons were then loaded to failure using a ramp load at a rate of 80%/s. Data were collected at 4 ms intervals and used to determine ultimate strength, ultimate strain, ultimate stress, yield stress, stiffness and elastic modulus.
The resting length of the tendon was taken as the length with the preload of 100 N applied after preconditioning. Displacement data were used to calculate tendon strain, and CSA at zero load was used to calculate tendon stress from the force data. A stress/strain curve was plotted. Ultimate properties were calculated from the highest load recorded during the break test. The point at which the gradient of the stress/strain curve reached a maximum in the linear portion of the curve was identified and 2 data points either side of the maximum were included to determine elastic modulus and tendon stiffness. The yield point was determined by the use of Meredith's construction (Gersak 1998), a method whereby a straight line is constructed from the initiation point of the curve to the failure point. The line is then transposed upwards until it becomes a tangent to the stress strain curve. The point at which the transposed line becomes a tangent to the curve is taken as the yield point.
Tendon tissue preparation for cross-link determination
Immediately following mechanical testing to failure, a sample of tissue (approximately 1 g wet weight) was taken from the mid-metacarpal region of the tendon and trimmed to remove the surrounding epitenon. Tissue was frozen at −70°C, lyophilised using a freeze drier (Modulyo)2 and ground to a powder using a mikro-dismembrator3.
Pyridinoline cross-link assay
Hydroxylysyl-pyridinoline and L-Pyr were quantified using high-performance liquid chromatography (HPLC). Approximately 15 mg powdered tendon was hydrolysed in 3 ml 6 mol/l HCl at 110°C for 24 h, dried and re-dissolved in 500 µl deionised H2O with 1% trifluoroacetic acid (TFA). The samples were filtered (0.2 µm filter) and 50 µl of sample was injected into the HPLC system (Series 200)4. The crosslinks were eluted using a Hypercarb column (150 × 4.6 mm, 7 µm i.d.)5 and a method adapted from Bailey et al. (1995). Mobile phase A was deionised H2O with 0.5% TFA (v/v) and mobile phase B was tetrahydrofuran with 0.5% TFA (v/v). A gradient programme was run as follows: 0–5 min 5% mobile phase B, 5–35 min 5–20% mobile phase B, flow rate 1 ml/min. Fluorescence was measured at λex 295 nm and λem 405 nm to detect HL-Pyr and L-Pyr cross-links. HL-Pyr and L-Pyr concentration was calculated using a commercially available standard (Metra Pyd/Dpd HPLC Calibrator)6. Collagen content was calculated by measuring the concentration of the amino acid hydroxyproline as previously described (Birch et al. 1998), assuming it to be present in collagen at 14%, in a 10 µl aliquot of the hydrolysate used for cross-link analysis. Cross-link concentration was expressed as mole/mole collagen.
Pyrrole cross-link assay
Lyophilised powdered tendon (25–50 mg) was suspended in 1.5 ml of TAPSO buffer (100 mmol/l TAPSO, pH 8.2, 10 mmol/l CaCl). Samples were heated at 90°C for 35 min and allowed to cool to 37°C before adding 0.75 ml of trypsin solution (10,000 u/ml in TAPSO). Samples were digested at 37°C for 16 h with shaking and centrifuged at 12,000 g for 20 min before measuring pyrrole levels. An aliquot of the digest was removed for hydroxyproline assay to determine collagen content as above. Pyrrole concentration was measured by mixing 0.8 ml of the digested tendon with 160 µl of Ehrlich's Reagent (500 mg 4-dimethylaminobenzaldehyde in 4.4 ml 60% perchloric acid made up to 10 ml with deionised H2O). Samples were left at room temperature for 10 min before measuring the absorbance at 570 nm in a UV-160A spectrophotometer7. In addition, each digest was measured in the absence of 4-dimethylaminobenzaldehyde and any background reading subtracted from the reading with Ehrlich's Reagent. Pyrrole concentrations were calculated by comparison with a standard curve prepared by mixing 0.8 ml 1-methyl pyrrole (0–10 µmol/l in TAPSO) with 160 µl of Ehrlich's Reagent. Results are expressed as mole/mole collagen.
Statistical tests were performed using SPSS (Version 17). All data were found to be normally distributed using a Kolmogorov-Smirnov test. Correlations were performed using a bivariate 2-tailed Pearson's correlation test. The level of significance was P<0.05.
Tendon strength and stiffness showed a wide variation between individual horses (Table 1) as found previously but this did not correlate with horse age, size, sex or type. The weakest tendon failed at a load of 7441 N whereas the strongest tendon was able to withstand a load 2.4 times greater (17,533 N) before gross failure. The material properties of the tendon tissue also showed a wide variation in mechanical properties (Table 1). The ultimate stress values ranged from 83.0–187.3 MPa, equalling the variation in structural properties. The elastic modulus showed a significant (P<0.001) positive correlation with the ultimate stress showing that stiffer tendon tissue is stronger. The yield stress also showed a significant (P<0.001) positive correlation with ultimate stress showing that the load at which the tendon begins to fail is a good indicator of the ultimate strength.
Table 1. Mechanical properties of SDFT (data are from 38 horses)
Mean ± s.d.
Ultimate strength (N)
12,379 ± 2494
Ultimate stress (MPa)
128.1 ± 74.7
Ultimate strain (%)
17.7 ± 3.9
Yield stress (MPa)
107.3 ± 19.1
1299.1 ± 180.0
Elastic modulus (MPa)
1217.0 ± 199.4
Collagen cross-link levels
The HL-Pyr cross-link was the predominant cross-link in all tendon samples (Fig 1). L-Pyr and pyrrole cross-links were detected in all samples but at much lower levels (Fig 1). The total cross-link levels equate to approximately one cross-link per collagen molecule (Fig 1). There were no relationships between the levels of the different cross-links in the tendon tissue.
Relationship between mechanical properties and collagen cross-link levels
The ultimate stress, yield stress and elastic modulus of tendons did not correlate with the HL-Pyr or L-Pyr levels measured in the tendon tissue (Fig 2). However, pyrrolic cross-link levels showed a significant positive correlation with ultimate stress (P = 0.004) (Fig 3), yield stress (P = 0.003) (Fig 4) and elastic modulus (P = 0.018) (Fig 5).
The results of this study support our previous findings demonstrating that equine tendons vary considerably between individuals in their structural and material properties (Birch et al. 2000). The collagen cross-link profile was as expected for tendon, with HL-Pyr being the dominant crosslink and L-Pyr present at much lower levels (Eyre et al. 1984). The collagen cross-link levels measured in this study are similar to those reported previously by our group (Birch et al. 2008a) and others (Lin et al. 2005) for equine SDFT. The formation of mature trivalent cross-links increases the tensile strength of collagen fibres (Avery and Bailey 2005), while inhibition of cross-link formation using inhibitors of lysyl oxidase significantly reduces the stiffness and strength of tendon (Puxkandl et al. 2002). However, our hypothesis that variation in the collagen cross-link level is responsible for variation in tendon material properties is only partially supported as the predominant cross-link, HL-Pyr, does not correlate with mechanical properties whereas the minor pyrrolic cross-link shows a significant positive correlation with tendon tissue ultimate strength and stiffness.
Interestingly, our results are similar to those in chicken bones where a significant positive correlation was found between pyrrole cross-link levels and 3-point bending strength in the diaphysis but no difference in the HL-Pyr and L-Pyr levels (Knott et al. 1995). However, the collagen in mineralised tissues, such as bone, has a much lower level of lysine hydroxylation than collagen in nonmineralising tissues. This gives rise to higher levels of L-Pyr and pyrrole as these crosslinks are favoured by low hydroxylation (Knott and Bailey 1998). For comparison, the mean level of pyrrole in our tendon samples was approximately 0.05 compared to 0.4 moles/mole collagen for the chicken bone samples (Knott et al. 1995). A study of adult vertebral cancellous bone by the same group (Banse et al. 2002) found that the HL-Pyr/L-Pyr ratio was a significant predictor of strength with a higher ratio resulting in stronger and stiffer bone. However, the vertebral bone samples were tested in compression whereas the chicken bones were tested in 3-point bending, which leads to a tensile mode of failure (Banse et al. 2002). Knott et al. (1995) suggested that the pyrrole cross-link, in contrast to the pyridinolines, could act as an interfibrillar cross-link and therefore have a more dramatic effect on tensile strength. The divalent immature cross-links occur between the telopeptide lysine- or hydroxylysine aldehydes and a lysine or hydroxylysine residue in the helical part of a neighbouring molecule, made possible by the quarter staggered arrangement of collagen molecules. The reaction of another lysine- or hydroxylysine aldehyde can only occur by the additional collagen molecule being in parallel alignment and therefore may occur between micro-fibrils (Bailey et al. 1998).
Given the low levels of pyrrole within tendon tissue and scatter of data around the linear relationship, it seems likely that the pyrrole cross-links influence the mechanical properties by an indirect mechanism. There are several ways in which this may occur. First, over-hydroxylation of telopeptidyl lysine residues has been shown to interfere with collagen fibrillogenesis, resulting in smaller diameter collagen fibrils (Pornprasertsuk et al. 2005). Increased lysine hydroxylation favours the formation of HL-Pyr rather than pyrrole cross-links. Therefore, higher pyrrole cross-link levels may reflect larger collagen fibril diameters, which are known to increase the strength and stiffness of tendon tissue.
Alternatively, the presence of different types of cross-links at the telopeptides may influence the susceptibility of the collagen to degradation. There is evidence to suggest that lysine hydroxylation decreases susceptibility to enzymatic degradation. In the field of fibrosis research, van der Slot et al. (2003) showed that fibroblasts derived from fibrotic skin of systemic sclerosis patients had a marked increase in expression of lysyl hydroxylase 2, the enzyme responsible for lysine hydroxylation in the telopeptide region of collagen. Furthermore, increased pyridinoline cross-link levels were found in the matrix deposited by systemic sclerosis fibroblasts, demonstrating a clear link between mRNA levels of the lysyl hydroxylase gene and the hydroxylation of lysine residues. Brinckmann et al. (1999) showed that a substantial increase in the cross-link HL-Pyr, which is thought to play a major part in the hardening of sclerotic tissue, is observed in fibrotic skin of lipodermatosclerosis patients. It may be that equine tendons with high pyrrole content are more readily degraded by proteolytic enzymes and, although this may at first appear to be a disadvantage, the ability to degrade and remove damaged matrix components is just as important as synthesis of new matrix for maintenance of tendon mechanical competence.
Tendon is a composite material and matrix components other than collagen may contribute to the mechanical properties. The noncollagenous protein content is predominantly proteoglycan (Birch et al. 2008b), which is thought to allow sliding movement between collagen fibrils (Scott 2003). However, partial removal of the proteoglycan component in mouse Achilles tendon and rat tail tendon did not change the elastic modulus or elastic behaviour of the tendon (Fessel and Snedeker 2009; Rigozzi et al. 2009). The collagen undergoes other chemical modifications, in addition to enzyme derived cross-link formation, which influence mechanical behaviour. A second nonenzymatic adventitious mechanism of cross-linking takes place through the reaction of lysine and arginine residues with tissue glucose; a process known as glycation (Avery and Bailey 2005). Glycation becomes increasingly important when collagen turnover is low following maturation of the tissue (Avery and Bailey 2005) and the process is known to increase the stiffness and decrease solubility of tissues (Kohn et al. 1984; Monnier et al. 1984). In addition, there is evidence to suggest that partially degraded collagen accumulates within the equine SDFT in mature horses, which has implications for mechanical behaviour of the tendon (Thorpe et al. 2010). However, the variation between horses may be greater than any age related change as we were unable to find a relationship between tissue stiffness and horse age.
Understanding the nature of the relationships between pyrrole cross-links, other matrix characteristics and tendon material properties may allow individual horses at risk of tendon injury to be identified using appropriate markers and provide the opportunity to monitor training regimes and develop conditioning programmes to enhance tendon properties.
The authors are grateful to The Horse Trust for their generous funding of this work and to Dr Tracey Smith for technical assistance with the mechanical testing of tendons.