Collagen Pre-strain Discontinuity at the Cartilage - Bone Interface

The bone-cartilage unit (BCU) is a universal feature in diarthrodial joints, which is mechanically-graded and subjected to shear and compressive strains. Changes in the BCU have been linked to osteoarthritis progression. Here we report existence of a physiological internal strain gradient (pre-strain) across the BCU at the ultrastructural scale of the extracellular matrix constituents, specifically the collagen fibril. We use X-ray scattering that probes changes in the axial periodicity of fibril-level D-stagger of tropocollagen molecules in the matrix fibrils, as a measure of microscopic pre-strain. We find that mineralized collagen nanofibrils in the calcified plate are in tensile pre-strain relative to the underlying trabecular bone. This behaviour contrasts with the previously accepted notion that fibrillar pre-strain (or D-stagger) in collagenous tissues always reduces with mineralization, via reduced hydration and associated swelling pressure. Within the calcified part of the BCU, a finer-scale gradient in pre-strain (0.6% increase over ∼50μm) is observed. The increased fibrillar pre-strain is linked to prior research reporting large tissue-level residual strains under compression. The findings may have biomechanical adaptative significance: higher in-built molecular level resilience/damage resistance to physiological compression, and disruption of the molecular-level pre-strains during remodelling of the BCI may be potential factors in osteoarthritis-based degeneration.

plot, showing peaks due to collagen D-period, particularly clear at the 3 rd and 5 th order after 86 azimuthally integration (0-360°) around the diffraction pattern for the q-radial range 0.1-1.2nm -1 . 87 Right top plot is 3 rd order collagen peak after background subtraction. (H) I(χ) profiles after corrected 88 intensity with Gaussian model fitting to the peak. Gaussian fitting was used to infer parameters (fibril 89 orientation and ρ).

91
transitional/superficial to deep zone cartilage [9], lying adjacent to the underlying mineralized tissue 120 with a putatively lower D-period. The calcified plate is subject to significant tissue-level and microscale 121 strains in loading [14]. However, the corresponding ultrastructural D-period variation in the calcified 122 plate (relative to the articular cartilage and trabecular bone) has not been measured. 123 124 This present study aims to measure the gradient in pre-strain in collagen D-period and related 125 ultrastructural properties in the BCI, focusing on the calcified plate, to better characterise this 126 biomechanical interface critical for healthy joint function. A bovine model of the bone-cartilage 127 interface from metacarpophalangeal joints is used [9,18]. High brilliance synchrotron X-ray SAXS 128 mapping, in the fast fly-scan modes, allows relatively rapid data acquisition across micro-and 129 macroscale tissue regions. The high X-ray flux at synchrotrons enable localized measurements with a 130 micro-beam, necessary to detect micron-scale variation of nanoscale ECM parameters. We carry out 131 SAXS mapping with multiple (three) spatial resolutions 1) across the entire articular cartilage, calcified 132 plate, and trabecular bone, and 2) focused on the articular cartilage and calcified plate, and 3) a high-133 resolution mapping of the calcified plate region. Our results will identify whether potentially 134 mechanically important equilibrium or pre-strain gradients exist at the ECM level, helping establish 135 baselines for deviations in osteoarthritis and other musculoskeletal degeneration.   Bovine cartilage was selected for use in the experiment, as it is both widely available, has less variability 142 as compared to human cartilage samples and is the next best biomechanical approximation of human 143 cartilage [29]. The intact joints were washed with biological detergent followed by immersion in 144 disinfectant (Chemgene) for 15 minutes (this procedure did not affect the internal cartilage and bone 145 tissue as the joints were intact). Joints were then opened using sterile disposable scalpel blades (Swann 146 Morton) in a biohazard safety cabinet. The metacarpal bone was removed, and two metacarpal condyles 147 were isolated from each joint using a high high-speed saw. Extracting intact bone-cartilage cores 148 required a custom sample-preparation process; a method inspired by the work of Aspden and co-149 workers was used [30]. First, each condyle was placed in a custom-designed 3D-printed holder with an 150 axis of rotation matching that of the condyle flexion axis. A bench drill (Axminster), with a diamond 151 coated coring drill bit was used to produce cores of 2mm diameter, under constant irrigation. The 152 condyle was rotated around the sample-holder axis so that, at each extraction point on the condyle, the 153 drill-face was incident locally normal to the condylar surface; this ensured flat cylindrical samples. The 154 drill was used to core to a 5 mm-depth at five locations on the proximal surface of each condyle. To 155 complete the extraction, the intact condyle was removed from the 3D-printed holder, and the reverse 156 face of the condylar section was abraded continuously with a high-speed Dremel rotary tool equipped 157 with a burring-bit, until each core detached. Both drilling systems were used with constant phosphate 158 buffered saline solution (PBS) (Sigma-Aldrich, Poole, UK) irrigation, with the drilling advancing 159 slowly to minimize heat and mechanical damage. Finally, the extracted cores were cut to an equal length 160 of 5mm by removing excess TB, using a Buehler IsoMet low-speed saw and a 3D printed sample holder 161 under constant irrigation. Final cores, as shown in Figure 1C Harwell Science and Innovation Campus, Didcot, UK). The beam size at the sample was measured to 168 be 20μm, beam energy of 14keV and the sample to detector distance at 5.8m. The samples were 169 mounted in film cassette windows (Figure 1C), which were sealed using a bilayer Kapton film 170 arrangement with enclosed PBS, to ensure hydration over the duration of the SAXS scan. SAXS 171 scanning was performed with a Pilatus 2M detector [32] (pixel size 172 μm; resolution 1475 × 1679 172 pixels) with an exposure time of 1s for each SAXS measurement. SAXS scans were carried out in "fly-173 scan" mode where the sample-stage moves continuously during the measurement; this ensures rapid data collection whilst still maintaining spatial registration. Three different SAXS scan settings were 175 used: 176 Regular scan: For n=6 samples, a rectangular 2D area scan of 2.6mm (vertical (y)) × 0.8mm (horizontal 177 (x)) was performed with 40 μm increments in both x-and y-direction across the BCU, which included 178 the full AC and CP tissue but only a part (1.36mm) of the ~4 mm long TB region. 179 Detailed scan: For a single sample, to obtained detailed SAXS data at the bone-cartilage interface (BCI) 180 region, a square 2D area scan of 0.4mmx0.4mm at the interface was performed with 5μm increments in 181 the x and y-direction. 182 Full-length scan: For a single sample, to obtain the SAXS data for the full-scale length of the extracted 183 BCU core shown in Figure 1C (extending from AC through the full length of TB), a rectangular 2D 184 area scan of 5mm (vertical (y); distance from joint surface) × 0.38mm (horizontal (x); distance parallel 185 to joint surface) was performed, with 20 μm increments in both x and y-direction.  Figure 1D), were reduced to 2D arrays of 1D intensity profiles. The radial I(q) and azimuthal 196 (angular) I(χ) files were extracted as a function of scattering vector (q) or azimuthal angle (χ) depending 197 on integration mode (azimuthal and radial, respectively) shown in Figure 1E, and saved as text files. 198 The Processing perspective in DAWN was used for the batch reduction of the data files [33]. As well 199 as a mask file specific for the Pilatus detector, which excluded the dead-regions (seen as white strips in 200 Figure 1D) between charge-coupled device (CCD) plates or the beam-stop. 201 I(q) processing for Fibril D-period: Figure 1G shows an I(q) plot from our data with DAWN showing 202 the radial meridional peaks, particularly clear at the 3 rd and 5 th order diffraction peaks (wavevector range 203 0.1 to 1.2 nm -1 ). For collagen-specific analysis, the 3 rd order meridional peak was used because it is 204 clearly visible in both (uncalcified) cartilage, the calcified plate and in trabecular bone (in contrast to 205 the 5 th order peak which is clearly visible in articular cartilage [9] but less clear in calcified tissue). After 206 azimuthal integration on DAWN, the I(q) data sets produced for each scan point were read by custom 207 Python scripts to perform the peak analysis for the desired parameters, as described below: 208 1) Total SAXS intensity: The total SAXS intensity was calculated from the area under the I(q) curve. 209 Total SAXS intensity, primarily from diffuse scattering, is related to the amount of interfacial area 210 between nanoscale inclusions like mineral platelets and surrounding matrix in bone [34], or between 211 fibrils and the extrafibrillar matrix. Here, we use it to a) distinguish between the bone-cartilage core and 212 surrounding fluid and b) to distinguish between the calcified and uncalcified tissue regions in the scan, 213 as calcified tissue regions have much higher diffuse SAXS intensity arising from the mineral phase. To estimate D-period, initially a parametric peak fitting approach (starting with Gaussians) was 230 used, following prior work in our group [9,41] and using the lmfit nonlinear least squares fitting package 231 in Python [42]. However, we found that there was a pronounced rightward asymmetry in the peak 232 profile, which was not visible in prior work [9,41]. The possible structural reasons for this effect (arising 233 from fibre diffraction), and its implications for the measured D-period will be explored and justified in 234 detail in the Discussion. Here, we note only that the peak asymmetry is visible due to the recent upgrade 235 of X-ray optics at the I22 beamline (in 2019) which led to much higher resolution in the peak profile 236 which could not be resolved in earlier work [9,18]. Due to the asymmetric peak nature, we use a 237 nonparametric method-of-moments approach, treating the I(q) profiles (with the diffuse linear 238 background intensity subtracted) as continuous distributions around the central wave vector (q) values, 239 to find the parameters centre(μ) and sigma(σ). The first moment (qFM=μ), that is the expectation value 240 μ, gave the position of the peak (q0 in our prior notation[9]) and the standard deviation gave σ (the width 241 of the peak). The following equations were used to calculate the first moment, the second moment and 242 the standard deviation. To represent the SAXS scans as 2D images, parameters (like D-period, total SAXS intensity and 254 collagen peak intensity) were displayed as a colour-scale bit-map, where the colour value was related 255 to the value of the calculated parameters through a colour-map. 256 I(χ) processing for fibril orientation and degree of fibrillar alignment: 1D azimuthal intensity profiles 257 I(χ) at each scan point were generated to obtain parameters characterising the angular collagen fibril 258 distribution -the average fibril orientation χ0, and degree of fibrillar alignment (ρ: random order→ρ=0, 259 uniaxial alignment → ρ large). In a similar manner to the I(q) analysis, the background diffuse 260 scattering needs to be corrected for. A three-ring subtraction method was used to obtain corrected 261 intensity value, by radially averaging the intensity just outside of the peak and subtracting this average 262 from the radial average on the peak, as described previously by us in total SAXS intensity was used to differentiate between articular cartilage (low total SAXS intensity) 297 and calcified plate (high total SAXS intensity due to mineral scattering). The dense cortical calcified 298 plate and the spongy trabecular bone regions were differentiated based on the vertical y-coordinate, as 299 the cylindrical sample geometry leads to a simplified 1D (vertical y) variation in the sample tissue 300 structure, and a visually clear demarcation between the calcified plate and the trabecular bone. For zones 301 within in the articular cartilage region, the degree of fibrillar alignment ρ and direction of fibril 302 orientation χ0 were used as differentiators: points with low ρ (< 2.1) with randomly aligned fibrils were 303 assigned to the transitional zone (TZ). Points above the transitional zone and with χ0 around 0° or 180° 304 (horizontal) was assigned to the superficial zone (SZ), and points below with χ0 around 90° were 305 assigned to the deep zone (DZ). The justification for this assignment is the well-known arcade-like 306 In the complementary approach 2), for each sample-scan, every SAXS pattern in the scan was assigned 321 to zones SZ, TZ, DZ, CP or TB based on the protocol described above, and reduced to obtain the 322 relevant SAXS parameters. Statistical comparison of each SAXS-parameter was carried out by 1-way 323 ANOVA, with the zone as the variable. After Anova, Tukey's honestly significant difference (HSD) 324 pair-wise tests were carried out between data in different zones. Statistical test-results were reported on 325 a per-sample basis and histograms of the SAXS parameters were also plotted. 326

327
The approach 1) of paired t-test analysis is reported in the main text, and the intra-sample approach 2) 328 in the Supplementary Data. This is because the main scientific question is whether there are (on average) 329 differences in SAXS-derived ultrastrucutural parameters across zones in the bone-cartilage unit. In the following, we present quantitative 2D mapping and zonal analysis of ultrastructural SAXS-based 339 parameters across the bone-cartilage unit at different length-scales. Figure 2A  squares; error bars: standard deviation across samples. 384 385

Significant differences in fibrillar-parameters between zones: 386
To extract the mean values of the SAXS-derived parameters, and estimate statistical differences 387 between zones, first histograms are plotted on a per-zone basis (Supplementary Figure 2) Table 2). 399 400

Qualitative differences in meridional SAXS peak profile between zones: 401
Since the zonal variation in D-period is small (e.g., ~65.2 nm to 66.0 nm in Figure 4) as is often the 402 case in collagenous tissues [18,47], it is a reasonable question as to how significant these changes are, 403 and whether our new non-parametric method of peak estimation may be artefactually influencing the 404 difference. First, we note that the magnitude of the effect (~1%) is comparable to the fibril-level strains 405 experience in bone under deformation to fracture [43]. To further clarify this, we plot I(q) profiles, with 406 and without diffuse SAXS background subtraction, from each of the 5 different zones in an example 407 BCU (Figure 7) to see how evident these peak-profile differences are in the original data itself. It is 408 clear, especially from the background subtracted data (Figure 7B), that there are distinct changes in 409 peak position and shape across the zones; most notably, the right-shifted I(q) for TB demonstrates the 410 lowered D-period seen in the colour-plots in Figure 2 and 4 and quantified in Figures 3 and 5. Figure  411 7B also shows that in AC, the total SAXS intensity due to the collagen peak alone (excluding the diffuse 412 SAXS scattering) is maximum in the DZ, and least in the SZ. This effect may arise due to a combination 413 of varying collagen concentration as well as degree of fibrillar ordering (higher ordering corresponds 414 to higher SAXS peak intensity [9,47]). Interestingly, the collagen peak height is also larger in CP than 415 in the TB region.

High resolution SAXS mapping at the interface shows finer-scale gradients: 432
To map out the ultrastructural variation of the fibril parameters at the calcified interface itself, Figure  433 6-7 shows the high-resolution colour maps with 5 µm spacing. Laterally-averaged line-profiles of the 434 same data are shown in Figure 8, to more clearly shown the spatial variation with distance from the 435 articular cartilage/calcified plate interface. The location at the top of the images corresponds to the 436 transition from unmineralized deep zone articular cartilage to the calcified plate, evidenced by the lower 437 total SAXS intensity coupled with the high collagen peak intensity. As indicated by the arrows in Figure  438 Figure 8A-B), the increase in D-period is accompanied by a decrease in ρ by about 456 20% (dashed lines a1 and b1). On the right-hand side in Figure 8A-B (also indicated by the region 457 "SCB" in Figure 7), there is a transition to lower ρ (dashed line b3), followed by further decline as the 458 TB region is approached (dashed line b4). In contrast to the AC/CP transition behaviour, on the right-459 hand side (CP/TB transition), the D-period decreases with reduction in ρ (dash lines a3 and b3). On 460 the left-hand side in Figure 8C, orientation of the fibrils is predominantly perpendicular to the AC/CP 461 interface (χ0 = 90°), which is a well-known "anchoring" type of architecture for the uncalcified AC to 462 adhere to the SCB. The variance in orientation is less toward AC/CP interface (evidenced by a more 463 homogeneous colour distribution as shown in Figure 7B). On moving toward the TB-region, we see a 464 trend to wider ranges of fibril orientations χ0, coupled with overall lower degree of alignment ρ, which 465 is low in TB. The higher resolution of the SAXS configuration compared to previous configurations [9,18] means 523 some features like the peak asymmetry are visible where before they were hidden by beam divergence. 524 We believe the finite fibre diameter is influencing this asymmetry via a fibre-diffraction effect [50], 525 conceptually sketched in Figure 9B. Fibrils with axial electron-density periodicity of D and radii R 526 exhibit a set of parallel ellipsoidal layer-line reflections (spacing 2π/D), with finite width (wp: p: 527 parallel) proportional to 1/R perpendicular to the fibril axis [50]. Thinner fibrils will have greater wp 528 (e.g. case 3 (thin fibril) vs case 1 (thick fibril) in Figure 9B); the axial width wa is kept constant in this 529 sketch. The polar azimuthal (circular) average of flat ellipsoids parallel to the qx axis thus leads to a 530 rightward skew asymmetry of the integrated peak (simulated data shown in Fig 9B, centre), as the tails 531 of the ellipsoids contribute to larger wavevector values in the azimuthal (χ) integration. As a result, the 532 measured D-period (via weighted moments) underestimates the true D-period (as q3=6π/D), with an 533 increasing skew correlated to a reduced measured D-period. Figure 5B shows that (especially in 534 cartilage) there is pronounced rightward skew of the I(q) peak. As a result, the estimated D-period via 535 our nonparametric method is likely an underestimate in these regions, and the pre-strain level in both 536 the AC-and CP-regions are even larger than estimated here, and the differences with respect to the 537 underlying trabecular bone maintained. 538

539
To confirm the above, we carried out a simulation-experimental test (Supplementary Information: 540 Fibre Diffraction Modelling), where the skew-effect was incorporated. Briefly, using fibre-diffraction 541 models of the type shown in Figure 9, model SAXS peak functions with varying levels of (fibre-542 diameter induced) skew were generated for a range of axial and parallel peak widths wa and wp. By 543 matching the experimentally observable skew and axial width parameters for a representative sample, 544 the true D-period can be estimated (Supplementary Figure 6). It is seen that while the cartilaginous 545 tissue D-period is elevated in comparison to the first-moment method, the difference between the high 546 D-period in the cartilaginous tissue and the lower D-period in the underlying trabecular bone is 547 maintained. Future work will include concurrently fitting the azimuthal and radial intensity profiles, to 548 independently determine wp as well as D, thus enabling a point-wise estimation of this effect. For the 549 current results, we note only that the difference between the D-period of the articular cartilage/calcified 550 plate (Type-II collagen fibrils) and the trabecular bone (Type-I collagen fibrils) is maintained. 551

552
The finer-structure gradient at the BCI interface, and the colour-contrasts in Figures 6-7  in regular biomechanical loading, with a material energy density of ½ E ε 2 where E is the tissue elastic 581 modulus of E~10-20 GPa [49], and ε the fibril pre-strain level of ~1.5%. Indeed, disruption of this ECM 582 material-level energy absorption mechanism due to tidemark duplication in OA could be a biomarker-583 level indicator of OA progression. 584 Turning to the limitations of our work, a basic technical drawback is the use of 2D SAXS mapping of 586 what is intrinsically a 3D cylindrical tissue object. Nevertheless, we mitigate these effects by choosing 587 a cylindrical sample geometry, with relatively tissue homogeneity along the X-ray beam direction, 588 coupled with a limited lateral width across the scan versus the height (high aspect ratio). Further, the 589 variations we are interested in are along the length of the sample, rather than through the thickness, and 590 are thus relatively insensitive to such 3D effects. For full-scale 3D reconstruction of the depth-591 dependent nanostructure, SAXS tensor tomography [36,53] will be essential. Secondly, we limit 592 ourselves to static scanning (no in situ mechanical loading) and leave spatial mapping of fibrillar 593 mechanics under loading to future work. Thirdly, we have limited our sampling to a specific condylar 594 anatomical site in the metacarpophalangeal joint and given that in-vivo strains vary across the joint, it 595 would be interesting to test if nanomechanical parameters like pre-strain shows a correlation with 596 varying physiological force levels across the joint. Lastly, we have not yet considered changes in OA 597 and ageing, to establish a baseline characterisation of the normal BCI in an animal model; a recent study 598 [11] has shown that nanoscale mineral particle thickness is significantly different in OA versus normal 599 human patients and investigating in situ mechanical response in aged human cartilage would no doubt 600 be of importance.