Probing multi-scale mechanical damage in connective tissues using X-ray diffraction

The accumulation of microstructural collagen damage following repetitive loading is linked to painful and debilitating tendon injuries. As a hierarchical, semi-crystalline material, collagen mechanics can be studied using X-ray diffraction. The aim of the study was to describe multi-structural changes in tendon collagen following controlled plastic damage (5% permanent strain). We used small angle X-ray scattering (SAXS) to interrogate the spacing of collagen molecules within a ﬁbril, and wide angle X-ray scatter- ing (WAXS) to measure molecular strains under macroscopic loading. Simultaneous recordings of SAXS and WAXS patterns, together with whole-tissue strain in physiolog- ically hydrated rat-tail tendons were made during increments of in situ tensile loading. Results showed that while tissue level modulus was unchanged, ﬁbril modulus decreased signiﬁcantly, and molecular modulus signiﬁcantly increased. Further, analysis of higher order SAXS peaks suggested structural changes in the gap and overlap regions, possibly localising the damage to molecular cross-links. Our results provide new insight into the fundamental damage processes at work in collagenous tissues and point to new directions for their mitigation and repair. X-ray scattering techniques. Detailed analysis of higher order diffraction peaks suggested damage is localised to molecular cross-links. The results are consistent with previous X-ray scattering studies of tendons and also with recent thermal stability studies on damaged material. Detailed understanding of damage mechanisms is essential in the development of new therapies promoting tissue repair.


Introduction
The mechanical properties of tendons are determined by collagen structure, and microdamage accumulation in the collagen fibrils has been associated with tendon injury [1]. Tendon injury and pathology are painful and disabling, and understanding the mechanisms underlying collagen damage can lead to novel strategies to diagnose and prevent injury, and facilitate regeneration. This requires comprehensive knowledge of collagen mechanics, cellmatrix interactions, and of the structural and biochemical response to loading and damage accumulation. [2] Tendons are composite materials, where collagen accounts for 86% of dry mass [3]. Three alpha chains, consisting of 1056, 0.29-nm spaced amino-acid residues arranged in a Gly-X-Y repeat pattern, combine to form 300 nm long triple-helical collagen molecules. These assemble into five-staggered microfibrils with a characteristic 67 nm D-period, forming the base crystallographic unit of collagen; as described by the consensus axial packing model, formulated by Hodge and Petruska in 1964 [4,5]. The fibrillar D-period is made up of a gap region of lower electron density, and an overlap region, where laterally adjoining, non-helical molecule terminals are cross-linked, stabilising the structure [6,7]. Interlaced microfibrils form fibrils, which combine with a proteoglycan-rich matrix, that maintains physiological hydration, to form fibres [8]. Tenocytes, specialised fibroblastic cells of elongated cytoplasm morphology, sparsely populate the tendon, and maintain homeostasis of the tissue. [2] Due to this complex hierarchical organisation, strain partitioning between structural levels and damage mechanisms in tendons are not well understood. Whole tendon deformation results from fibrillar straining and intra-fibrillar matrix deformation. Fibrillar straining, in turn, is a compound effect of molecular deformation and molecular rearrangement [8,9]. Micro-damage accumulation at small length scales may lead to pain and loss of functionality and ultimately macroscopic failure [10].
Deformation mechanisms and mechanical properties at the nanoscale can be probed using in situ loading experiments with X-ray diffraction, due to the quasi-crystalline organisation of collagen at fibril and molecular level. Small angle diffraction (SAXS) has been used to measure the 67 nm D-period between repeat units in fibrils [6,11], while wide-angle diffraction (WAXS) can measure the 0.29 nm intra-amino acid spacing along the molecular axis [12]. Fibrillar and molecular strain can be estimated from the shifting of diffraction peaks during in situ loading [9]. Further analysis of the diffraction patterns can provide structural information: The ratio of integrated intensities of the second and third order SAXS peaks is directly correlated to the ratio of the overlap region length to that of the whole D-period [9,13], capturing structural changes due to load. Furthermore, lateral packing and fibril distribution affect damage mechanisms, and have been probed in recent years using X-ray ptychography [14]. Previous X-ray diffraction (XRD) investigations of non-physiological damage to fibrillar collagen through dehydration and heating have shown changes to lateral packing tightening and fibrillar D-period shrinking to 64 nm, resulting in a loss of gap-overlap periodicity and a shearing of molecular packing [6,15]. However physiological mechanical damage mechanisms have not been investigated using these techniques. The aim of this investigation is to describe collagen material property changes at three length scales following mechanical damage applied under phyisiological conditions, and to suggest potential underlying deformation and damage mechanisms. This was achieved using in situ mechanical loading and X-ray scattering to compute the multi-scale mechanical properties at each hierarchical level, to investigate the relationship between these properties and pre-applied mechanical damage.

Tendon harvesting
Tendons were harvested from 300 to 350 g (10-12 week old), male Sprague-Dawley rats, sacrificed by cervical dislocation for an unrelated study, by adapting the method described in [16]. Briefly, individual fascicles were separated, and selected based on diameter (0.22 mm ±0.05 mm, measured using an optical micrometer (Keyence, Milton Keynes, UK; precision 1 Â 10 À5 mm)). A tendon sub-population was pre-damaged, and all samples were fixed to neodymium magnets (8 mm diameter by 10 mm) using epoxy putty, and frozen at À20 in Phosphate Buffered Saline (PBS, Gibco, UK) until use, to not alter the mechanical properties of the tendons [13]. Black ink marks (see arrows in Fig. 1(d)) were made on all samples, for optical measurement of whole tissue strain.

Pre-damage application
Tendons (n = 6) were pre-damaged using a bench top uniaxial strain machine (Bose Electroforce, Eden Prairie, MN), by applying a continuous 10 MPa load for 500 s. Samples were then placed in PBS for 2 min, and damage quantified by manual measurement of permanent change in length between two ink marks. A length change greater than 5% was selected as the damage threshold. Multiphoton microscopy of other samples damaged to this level confirmed microstructural changes as described previously [17]. This level of damage is physiologically relevant for induction of tendinopathy in animal models of disease [17].

In situ loading
Tendons were loaded in situ using an electromagnetic loading setup, consisting of a remotely operated, current-controlled solenoid, applying forces up to 0.5 N. Samples were placed in PBSfilled Kapton tubing, sealed using Polydimethylsiloxane (PDMS, Corning, UK) to maintain hydration, and held by a custom-built experimental rig, designed to fit within the sample space of the I22 beamline at the Diamond Light Source, UK, and mounted on a linear translation table for positioning ( Fig. 1(a)). Each magnet, weighing 6.1 g, produced a 0.06 N tare load when the solenoid was inactive.

X-ray diffractometry
Diffraction experiments were carried out at beamline I22, Diamond Light Source, UK. The X-ray beam was coarsely focused to 90 Â 110 lm 2 (full width at half maximum), with the largest dimension parallel to the sample axis (z-direction in Fig. 1(a) The optical design of I22 typically generates beamsizes at the sample of 300 Â 100 lm 2 , horizontal Â vertical (Full width at half maximum) by directly reimaging the X-ray source in a 1:1 focussing arrangement. For these experiments the beamsize was reduced to 110 Â 90 lm 2 , horizontal x vertical, using collimating slits with an attendant loss of flux. This represents a compromise between minimising beamsize and maximising flux whilst maintaining the naturally low divergence of I22 (80 Â 50 l rad, horizontal x vertical) essential for high quality SAXS data collection. Smaller beamsizes without loss of flux could be achieved through the use of additional focussing optics, however the increased demagnification here would be offset by increased beam divergence that would negatively impact data quality. A photon energy of 12.4012 ±0.0014 keV was selected DE E ¼ 1:12 Â 10 À4 , corresponding to an X-ray wavelength of 0:0997 AE 1:066 Â 10 À5 nm. SAXS data was collected over a q range of 0:0025 À 0:15A À1 , and over an azimuthal range of 360°using a 2D Pilatus P3-2 M detector (DECTRIS, Switzerland) placed 9 m from the sample and with an exposure time of 5 s. WAXS data was collected for a 2h range of 16-50°on a 1D Pilatus P300k-W detector (DECTRIS, Switzerland), for 5 s at every loading point ( Fig. 1(b) and (c)). A high definition camera placed above the beamline exit, triggered by the beamline software, and focused on the sample was used to record sample images for macro strain computation by image correlation (Fig. 1  (e)).

X-ray diffraction pattern analysis
SAXS diffraction patterns, showing distinct meridional peaks up to the 12th order were converted from polar to Cartesian coordinates, using dried chicken collagen as a calibration reference ( Fig. 1(b)). Hence, a radial line profile for the 12 peaks was extracted, using both upper and lower reflections. Gaussian curves fitted to the peaks were used to compute peak position, full width at half maximum (FWHM) and integrated intensity. WAXS diffraction patterns were collected in 1D due to detector geometry. Calibration of the detector geometry was performed using a CeO 2 standard. A single peak was extracted by fitting a Gaussian curve to the line profile obtained by averaging the detector output in the y direction. Peak position, FWHM and integrated intensities were calculated.

Sample camera tissue strain
High definition, 1292 Â 962 pixel images of samples taken at each load point were used to compute tissue strain with a custom-written Matlab digital image correlation routine.

Statistical analysis
Statistical analysis (paired, two-tailed t-tests) was carried out using Graphpad PRISM. p < 0.05 indicated statistical significance of results (95% confidence interval).

Ethics approval
All tissues were harvested from rats sacrificed for unrelated experiments, and thus no ethical approval was required.

D-period variation with damage
Comparison between damaged (N = 6) and undamaged (N = 6) populations tare load D-period (i.e. before solenoid activation) indicates a significant decrease in both fibrillar D-period and intra-amino-acid spacing with damage application (Fig. 2).
Significant shortening of intra amino-acid spacing and shortening of fibrillar D-period ( Fig. 2a and b) have been shown to be related to X-ray radiation damage [13] and dehydration [18]. Due to minimal exposure to X-rays, and continuous sample hydration by immersion in PBS, neither mechanism could explain the change in D-periods observed here. This was verified by comparison of measured D-period before and after a full scan, showing no significant change, on average, in D-period (data not shown).

Strain partitioning and stiffness changes with damage
During fascicle elongation, the strain induced by applied stress is distributed between the collagen fibrils and the interfibrillar ground substance [8]. The fibril strain is in itself a compound effect of an affine elongation of the collagen molecules, and a rearrangement of the molecular structure within the five-staggered unit [9]. Here, we can evaluate how this partitioning of strain is affected by deformation damage, by simultaneous collection of WAXS, SAXS and Macroscopic high definition sample camera data.
From peak positions of SAXS and WAXS patterns and from macro images, straining of the characteristic measured length was calculated as: where X 0 denotes the average characteristic length (D-period for SAXS, amino-acid spacing for WAXS, macro length for sample camera) of the unloaded sample and X denotes the characteristic length calculated from the loaded diffraction pattern. X 0 for SAXS and WAXS was calculated as an average of measurements taken twice at 5 positions along the tendons axis, for n = 3 unloaded tendons. Unloaded D-period for SAXS was computed as 674.5 Å ± 0.07%. Unloaded intra amino-acid spacing from WAXS was calculated as 2.798 Å ± 0.09%. Fig. 3. Stress-strain curves from in situ tensile loading of tendon fascicles probed by X-ray diffraction, and loading modulus relation to damage.

Loading moduli of fibril, molecule and tissue
Strains measured at the three length scales probed were plotted against stress. Stress calculations assumed tight lateral packing of molecules and fibrils. Stress-strain curves for molecules and fibrils show linear increasing trends for all samples (Fig. 3a), confirming the elastic nature of the material, allowing loading moduli to be calculated from the curve slopes. Fig. 3a confirms strain partitioning at the three structural levels: molecular elongation is lower Fig. 4. Molecular elongation and rearrangement contributions to fibril deformation. Red bars = average slope of data sets from corresponding condition. Red bar thickness represents deviation of slopes (thicker bar = higher deviation, a.u.). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) than that of fibrils, which in turn is lower than whole tissue elongation. This confirms additional mechanisms involved during loading of tendon samples, with a rearrangement of collagen molecules accounting for the additional strain measured in fibrils, and ground substance deformation and fibril superstructure rearrangement contributing to higher strains measured in the whole tissue sample [8]. Studies of cross-link deficient tendons have shown an increase in this effect, indicating the cross-link locations as the areas which are deformed during straining [19].
Tissue-level stiffness was not significantly affected by preapplied damage (Fig. 3b). Mean fibril loading modulus, decreased significantly (p < 0:05) from 1300 to 900 MPa with pre-applied damage (Fig. 3b). Conversely, average molecular Loading modulus significantly (p < 0:05) increased from 2690 to 3395 MPa with damage. This divergent behavior indicates permanent changes to the structural organisation of the tissue following damage application. Qualitatively, molecular stiffening observed might be explained by a change in triple helix geometry, as reported for mechanically overloaded tendon collagen, which denatures with load application. Previous studies on overloaded collagen fibrils showed a reduction in enthalpy of denaturation, and proposed that the collagen molecules were uncoiled into a stable denatured state, consistent with our observations [20]. This may reduce axial mobility of neighboring amino-acids and thus reducing the increase in intra-amino-acid spacing as load is applied [21]. The loss of stiffness at the fibril level implies ulterior deformation mechanisms in fibril straining, not limited to affine deformation from triple helix straining [19,22].

Damage alters tendon deformation modalities
Combining simultaneous WAXS and SAXS data, the contributions of deformation modes to the overall deformation of the tissue were calculated. Assuming fibril elongation to be an additive effect of an affine molecular stretch and molecular rearrangements, then: where E D is the elongation (nm) at the fibrillar D-period level, E M is the elongation of intra-amino-acid spacing along a molecule (nm), and M includes all molecular rearrangement effects. Hence, overall stretch of the fibril D-period can be described by: where D 0 is the fibril D-period in unloaded state and m is the strain at the molecular level. From Eq. 3, using WAXS and SAXS data, undamaged samples qualitatively showed a more uniform partitioning of D-period deformation between molecular elongation and rearrangement, with an average decrease in elongation and increase in rearrangements with stress ( Fig. 4a and b). Molecular elongation decreased rapidly in damaged samples (Fig. 4c), with rearrangements conversely increasing (Fig. 4d), apparent at low stresses (<6 MPa). In both groups, molecular elongation converged to approximately 0.38% of D-period strain at high stresses.

Damage causes changes in structure
Changes in the molecular arrangement, caused by sliding of laterally and axially adjacent molecules, cause a variation in the ratio of the overlap region to the total D-period. These changes account for the higher straining of fibrillar D-period compared to only an affine molecular deformation.
Assuming the electron density along a fibril is described by a square function with a period of 67 nm, having a low period (gap) of 0.54D and a high period (overlap) of 0.46D, a variation in gap and overlap would result in a change in the relative integrated intensities of successive SAXS peaks [23,9]. This can be described as: where n is the peak order, O is the overlap length, and D is the Dperiod length. Qualitative analysis of O/D values plotted against stress indicates a steeper average decrease for damaged samples compared to undamaged samples (Fig. 4e). Since the D-period has been shown to decrease with damage (Fig. 2a), this effect must be due to the overlap region shrinking faster in damaged samples. This region corresponds to the location of cross-links between terminals of laterally adjacent collagen molecules within fibrils, and an increased deformation as a result of applied stress may implicate cross-link damage as the location of tensile damage. In artificiallyinduced hyper-cross linking, the opposite effect is observed [13], suggesting that damage accumulation decreases cumulative crosslinking strength.

Limitations of experimental method and analysis
To further understand the mechanisms of damage in collagenous connective tissues, the ability to probe lateral packing of fibrils and helical arrangement of molecules simultaneously would allow for a more complete appraisal of structural changes with damage, by measuring triple helix pitch and fibril helical characteristics [24,5,25].

Conclusions
Understanding collagen damage at the nanoscale is key to preventing injury from accumulated micro-damage, as well as to develop strategies for encouraging healing. X-ray diffraction allows us to probe biological collagenous tissues at the nanoscale. By performing in situ mechanical loading it is possible to probe the mechanical properties at various scales within the collagenous superstructure. Simultaneous data collection is key to studying the molecular, fibrillar and tissue behaviour under load, and exploring the relationship between these three structural levels.
Here, we focus on the effects of mechanical damage induced by tensile elongation on the mechanical properties of rat tail tendons at two scales: the fibril and the molecular scales. We show a significant decrease in both fibril D-period and molecular intra-amino acid spacing as a result of applied damage, which is indicative of damage mechanisms at both scales. We have shown significant loss of fibril level stiffness and significant increase of stiffness at the molecule level following damage, indicating distinct mechanisms at the two levels. The overlap region of the fibrillar D-period exhibits a faster reduction in length in damaged samples than the gap region, indicative of potential damage being localised here, where laterally adjacent collagen molecules are cross-linked. This is consistent with the observed inhomogeneous partitioning of D-period strain between molecular elongation and rearrangements observed in damaged samples, once again indicating damage localised where the mechanisms allowing molecules to move relative to one another are found.