Scanning SAXS–WAXS microscopy on osteoarthritis-affected bone – an age-related study
aIstituto di Cristallografia (IC-CNR), Via Amendola 122/O, I-70126 Bari, Italy, bPaul Scherrer Institut, 5232 Villigen PSI, Switzerland, cLaboratorio di Tecnologia Medica, Istituto Ortopedico Rizzoli, Via di Barbiano 1/10, 40136 Bologna, Italy, and dLaboratorio PROMETEO-RIT, Istituto Ortopedico Rizzoli, Via di Barbiano 1/10, 40136 Bologna, Italy
*Correspondence e-mail: firstname.lastname@example.org
Osteoarthritis (OA), among other bone pathologies, is expected to determine supramolecular changes at the level of the mineralized collagen fiber. In a proof-of-principle study, bone biopsies were collected from six coxarthritis-affected patients, aged 62–87 years, during hip prosthesis implant surgery, sliced down to 100 µm-thick tissues, and investigated using scanning small-angle and wide-angle X-ray scattering (SAXS and WAXS) transmission microscopy. A multi-modal imaging evaluation of the SAXS and WAXS data, combined with principal component and canonical correlation analyses, allowed the transformation of the raw data into microscopy images and inspection of the nanoscale structure of the mineralized collagen fibers across mm2 tissue areas. The combined scanning SAXS and WAXS microscopy is shown to be a suitable choice for characterizing and quantifying the nanostructural properties of collagen over extended areas. The results suggest the existence of a correlation between age and cross-linking-induced rigidity of collagen fibers.
Primary osteoarthritis (OA) of the hip, also named osteoarthrosis of the hip or coxarthrosis, is a chronic degenerative disorder of the hip joint, causing growing articular pain that can bring the patient to lifestyle limitations until surgical intervention is needed (Hoaglund & Steinbach, 2001). In general, OA of the joints like hip, knee and ankle affects a large part of the elderly population, mostly women, with about 50% of the patients already showing symptoms at the age of 65 (Zhang & Jordan, 2010; Lotz & Loeser, 2012). It is well known that in primary OA the cartilage collagen fibers can become more susceptible to mechanical injury, causing wear and degradation of the articular cartilage (Lotz & Loeser, 2012). At the same time, there is also a significant impact of OA on bone close to the articular cartilage, with important remodeling at the micro scale along the main loading direction of the tissue (Baker-LePain & Lane, 2012). It is important to know if OA affects also the collagen-fiber organization in the articular bone, with attention to the effect of age on bone quality. Any evidence of such a dependence on the age of the patient could lead in the future toward personalized pharmaceutical treatments (Martel-Pelletier et al., 2012) of primary OA, as well as to customized prosthetic devices for joint replacement (Luyten & Vanlauwe, 2012). As an example, therapies aiming at a remineralization of the bone may depend on the collagen structure within the bone being still intact, because the remineralized bone might be very brittle otherwise.
Fundamental questions still have to be addressed about (i) the influence of the patient's age on the quality of the collagen network and the mineral phase in bone, (ii) the existence of a relation between the collagen quality in bone and cartilage, (iii) the parameters promoting natural bone remodeling, and (iv) the influence of pharmaceutical treatments on both the collagen network and the mineral content of the bone.
Small- and wide-angle X-ray scattering (SAXS and WAXS) are established methods which allow the inspection of matter with nanometric or atomic resolution, respectively. Recently, it has been proposed to perform SAXS and WAXS scanning microscopy with a focused X-ray beam to raster-scan a bony tissue specimen (Paris, 2008; Dunlop & Fratzl, 2010; Bünger et al., 2010). This allows nanoscale information to be obtained across mm2 extended areas.
Such an approach was exploited by the authors of the present article as well, in combination with circularly polarized light (CPL) microscopy, to provide insights on the nanostructure and morphology of cortical and trabecular human bone affected by Paget or dwarfism pathology, at sub-osteon resolution (Giannini et al., 2012).
The aim of the present work has been to collect and analyze SAXS and WAXS scanning microscopies from biopsies of OA-affected bone tissue to investigate possible correlations between age and structural changes at the level of collagen type-I fibers and nanocrystalline hydroxyapatite. Across the investigated samples from a small population of six patients, a correlation between age and the stiffness of the mineralized collagen fibers, probably cross-linking induced, was found. Indeed, intermolecular cross-linking is a well known natural mechanism preventing the long rod-like collagen molecules from sliding past each other under stress (Bailey et al., 1980). Other authors have found that an excess of cross-links tends to over-stiffen the fibers (Bailey, 2001). Furthermore, for other pathologies or bone conditions, a clear correlation between aging and cross-linking overproduction has been shown (Zimmermann et al., 2011; Saito & Marumo, 2010). The article is organized as follows: a description of the preparation of the bone biopsies and of the setup used to collect SAXS and WAXS data is given in the Experimental section; the Methods section contains all the steps needed to transform raw data in microscopies; the Results and discussion section summarizes the main findings of the work.
The bone specimens in this study were harvested from the femoral necks of six patients, 62–87 years old, during total hip-arthroplasty surgery for primary coxarthrosis. A written informed consent was obtained from the patients. A brief description of the patients in terms of age and gender is reported in Fig. 1.
Slices of bone 3 mm in height were obtained in the lower area of the femoral neck, always from the same medial sector in proximity to the great trochanter. Once retrieved, the samples were preserved in 4% formalin until being investigated.
Each sample was embedded in methylmethacrylate with its distal surface supported on the container, according to Boyde's method (Boyde et al., 1984; Boyde & Riggs, 1990). In brief, after washing with bidistilled water to remove all the residual marrow debris, the samples were degreased and dehydrated in vacuum with the following series of alcohols: 80% ethanol overnight, 3 steps in 95% ethanol of 1 h each, 5 steps in methanol of 1 h each at 310 K. Infiltration occurred with four passages in methylmethacrylate, first destabilized by chromatography with aluminium oxide and finally embedded in mixtures of methylmethacrylate and di-n-butylphthalate at 25%(w/v) at 298 K. The reaction mixture catalyst was benzoyl peroxide at 3.5%(v/v).
From the embedded blocks, serial 100 ± 5 µm slices were obtained (Boyde et al., 1984; Boyde & Riggs, 1990) with a 300 µm-thick wafering blade (Leica 1600). The slices were compressed between two glass slides while maintaining the correct orientation. Each slice was visualized in a circularly polarized light microscope (NIKON AZ100 with magnification from 20× to 160×). The CPL data were used to determine the regions of interest for the scanning SAXS/WAXS microscopy investigations.
SAXS–WAXS experiments were performed at the cSAXS beamline (Bunk et al., 2009) of the Swiss Light Source (SLS), Paul Scherrer Institut, Villigen, Switzerland, in accordance with all relevant regulations including the Swiss ordinance 814.912 on the contained use of organisms. A monochromatic X-ray beam (E = 13.6 keV) was focused down to about 30 × 30 µm by a bent monochromator crystal and a bent mirror. Each biopsy, 100 ± 5 µm thick, once removed from the glass slides, was placed perpendicular to the direct beam and raster scanned through the beam spot, with the detector measuring the total intensity coming from the currently illuminated spot on the sample, integrated over the exposure time and across the sample thickness. Step sizes of 25 µm in both the horizontal and the vertical direction, and exposure times of 80 ms, were used. The embedding polymer protects the bone biopsy from external contamination as well as from artificial bending, i.e. ensures that the X-ray spot illuminates a flat and rigid surface.
To speed up acquisition, data were recorded in a continuous line-scan mode where the sample is moved at constant speed along a line of the two-dimensional raster scan, with the Pilatus 2M detector (Kraft et al., 2009) continuously recording data. SAXS and WAXS data were collected at sample–detector distances of 7092 and 108 mm, respectively. The SAXS and WAXS data sets were transformed into microscopy images as explained in the Methods section.
Bone is a composite material whose major components are cross-linked collagen type I, i.e. fibrous protein, and nanocrystalline carbonated hydroxyapatite (HA), i.e. mineral, which form mineralized fibers assembled in concentric lamellae (Weiner & Wagner, 1998; Rho et al., 2002; Olszta et al., 2007; Dunlop & Fratzl, 2010). Each collagen fiber consists of a staggered arrangement of collagen molecules with embedded HA nanocrystals. Bundles of mineralized collagen fibrils form collagen fibers.
Fig. 2(a) shows a schematic drawing of the setup that was adopted in the present experiment to collect the SAXS and WAXS data. In Figs. 2(b) and 2(d), typical WAXS and SAXS two-dimensional frames are displayed. Fig. 2(c) depicts the crystal unit cell for the HA structure that produces the WAXS pattern (Kay et al., 1964), which has been shown here to remind the reader that WAXS data contain information at atomic length scale, like atomic positions, cell symmetry and size.
SAXS is sensitive in general both to nanoscale mineral structures in the size range 1.5–50 nm (Weiner & Wagner, 1998; Rho et al., 2002; Olszta et al., 2007; Dunlop & Fratzl, 2010) and to mesoscale collagen fiber in the size range 65–120 nm (Bigi et al., 1982, Wilkinson & Hukins, 1999; Orgel et al., 2001; Suhonen et al., 2005). The large SAXS sample–detector distance of 7092 mm chosen here allowed us to better resolve mesoscopic length scales. The mesoscale collagen fiber SAXS diffraction pattern is composed of two distinct signals: the meridional and equatorial type-I collagen reflections. The meridional scattering reflections are due to the electron density periodicity along the collagen fiber axis (∼65 nm), the equatorial scattering ones to the packing of the collagen fibrils into fibers (≥100 nm).
Figs. 2(e) and 2(f) are cartoons of the relative positions of the collagen meridional and equatorial SAXS reflections with respect to the fiber orientation in real space, as expected for a tissue containing type-I collagen fibers. Along with each cartoon, typical corresponding meridional and equatorial SAXS profiles have been reported, to show the link between models and experimental signals. The meridional reflections observed in the SAXS profile of Fig. 2(e), located at Δq1 = 0.0090–0.0119 Å−1 (d1 = 52.9–69.6 nm) and Δq2 = 0.0274–0.0309Å−1 (d2 = 20.3–22.9 nm), are diffraction replicas of the d ≃ 64–67 nm periodicities, arising from the packing of the collagen molecules along the fiber axis (Wilkinson & Hukins, 1999; Gentleman et al., 2003). The equatorial profile in Fig. 2(f) shows only one broad peak, at about q = 0.0054 Å−1, which corresponds to a length in direct space of about 115 nm, related to the spacing of the collagen fibrils assembled into bundles (Weiner & Wagner, 1998; Suhonen et al., 2005). The higher the degree of packing of fibrils in bundles, the more pronounced and resolved the equatorial reflections are, i.e. the greater the inflexibility of the fibrils is, the higher the intensity of the meridional reflections is expected to be.
Lateral structural inhomogeneity in bone tissue is reflected in different collagen SAXS diffraction signals, both meridional and equatorial, as well as in different WAXS signals, across the investigated area. The information encoded in a single SAXS or WAXS two-dimensional frame is averaged over the X-ray-illuminated scattering volume of 30 × 30 × 100 µm in the present case. Inhomogeneity on this or larger length scales can be detected.
Fig. 3 shows four representative two-dimensional/one-dimensional WAXS (Figs. 3a–3d) and SAXS (Figs. 3e–3h) frames/patterns simultaneously measured at different sample positions. Figs. 3(a) and 3(e) present the WAXS and SAXS frames/profiles acquired in an area that does not contain any bone, but only the polymer which embeds the biopsy. The SAXS data vary from an extremely clear and pronounced collagen meridional and equatorial scattering, with up to six–seven meridional diffraction orders (Fig. 3f), to an almost isotropically distributed SAXS intensity (Fig. 3g), where neither the meridional nor the equatorial peaks are revealed. The intermediate situation of Fig. 3(h) shows a case where only the first meridional order is distinguishable from the background and the equatorial scattering is broadly diffused. We selected these four WAXS/SAXS frames/patterns using the principal component analysis (PCA) approach (Abdi & Williams, 2010) as explained below. The intensity anisotropy observed in the SAXS frames depends on the orientation of the collagen fibers in the X-ray-illuminated volume seen in a projection plane perpendicular to the X-ray beam. If the collagen fibers are highly oriented in this plane then an oriented scattering pattern like the one shown in Fig. 3(f) will result. If fibers run in different directions within the X-ray-illuminated area in this plane, or if they are oriented mainly along the X-ray beam, then more isotropic scattering patterns will be the result, as in Fig. 3(g).
The WAXS frames/patterns displayed in Figs. 3(b) and 3(d) are typical of data from HA nanocrystals (Giannini et al., 2012) and allow us to attribute the SAXS scattering in Figs. 3(f) and 3(h) to mineralized collagen fiber, as expected. Conversely, the WAXS data in Fig. 3(c) do not correspond to the scattering from the mineral bone component, which allows us to understand why the relevant SAXS data in Fig. 3(g) did not contain any reflection related to fiber structure. Most probably, the WAXS/SAXS data in Figs. 3(c) and 3(g) arise from other types of tissues, different from type-I collagen fiber.
Fig. 4(a) is the composite image of a typical scanning SAXS measurement, with each pixel containing a two-dimensional SAXS frame, encoding the strength of the scattering signal from the laterally inhomogeneous sample. The yellow area corresponds to the bone tissue and the blue one to the polymer the biopsy is embedded in. The entire image corresponds to a 7 × 7 mm area of the investigated sample, comprising about 80 000 SAXS frames.
The same area was also studied by WAXS, producing maps analogous to Fig. 4(a) (not shown). Considering the large number of SAXS and WAXS two-dimensional frames, data evaluation leading to scanning microscopy images with SAXS or WAXS contrast [like those shown in Figs. 4(d)–4(f)] relied on statistical methods and was carried out in the following steps: (1) folding the two-dimensional frames into one-dimensional patterns; (2) length scale definition; (3) the multi-modal X-ray scattering imaging method; (4) principal component analysis; and (5) canonical correlation analysis. A flow chart is added in the supporting information1 (Fig. S1 ) to summarize the logic underneath the analysis.
Let us follow step by step the entire procedure. All two-dimensional SAXS and WAXS frames were azimuthally integrated into one-dimensional profiles (two-dimensional–one-dimensional folding). Typical one-dimensional SAXS and WAXS profiles are presented in Figs. 4(b) and 4(c), respectively: the SAXS profile showing meridional diffraction peaks, replicas of electron density periodicity along the collagen fiber axis; the WAXS profile being indexed as the diffraction pattern of the hydroxyapatite (HA) crystalline phase, as expected for the mineral bone tissue component (Giannini et al., 2012).
From the SAXS/WAXS profiles (Figs. 4b and 4c), four peaks, two for the SAXS and two for the WAXS data, were identified as signals to be monitored through the entire raster scan (length scale definition). The peaks correspond to precise length scales: Δq1 = 0.0090–0.0119 Å−1 (d1 = 52.9–69.6 nm) and Δq2 = 0.0274–0.0309 Å−1 (d2 = 20.3–22.9 nm), extracted from the SAXS data, corresponding to the first and third meridional reflections of collagen (see Fig. 4b and inset) and carrying information on the periodicity along the fiber axis; Δq3 = 1.8051–1.9061 Å−1 (d3 = 0.330–0.348 nm) and Δq4 = 2.0358−2.1050 Å−1 (d4 = 0.309–0.299 nm) related to the 002 and 210 crystallographic reflections, respectively, of the HA WAXS pattern (see Fig. 4c and corresponding inset) and carrying information on the nanocrystalline hydroxyapatite mineral component of bone. The spacing of the third meridional reflection (d2) is just a higher-order replica of the first meridional reflection of collagen (d1), i.e. d2 = 3d1, and they both monitor the same direction and packing along the fiber axis.
Once the length scales of interest had been selected, the final quantitative microscopy images were composed by means of the multi-modal X-ray scattering imaging method (Bunk et al., 2009). Each SAXS/WAXS two-dimensional frame was divided into 16 angular sectors, each 22.5° wide (see also Fig. S1 ), and only the diffraction rings selected in the previous step were studied from the entire two-dimensional frame. The presence/absence of these diffraction rings, as well as the intensity distribution along the azimuth, was studied, and relevant quantitative information was extracted and transformed into microscopy images (shown in Figs. 4d–4f). The evaluation of the background below each of the monitored peaks was performed by linear interpolation through three–four selected points on the tails of the WAXS peaks and interpolation of an empirical 10−2.6 background for the first-order and 10−1.5 for the third-order collagen peaks in the SAXS data. For the HA 210 peak no background subtraction was applied since it overlaps with a neighboring peak. This leads to an isotropic background in the sample regions where the polymer used for embedding is prominent. The following information was encoded in these images: (i) the local presence (color) or absence (black) of collagen/HA tissue components per pixel, (ii) the orientation of the collagen fibers/HA nanocrystals designated by the direction of the relevant sector within the colored wheel, with white assigned to un-oriented material, and (iii) the degree of orientation of the collagen fibers/HA nanocrystals, determined as the brightness of the colors. The HA orientation is displayed both along the nanocrystal elongation direction, i.e. the  crystallographic direction, and perpendicular to it, along the  direction. A comparison of Figs. 4(d) and 4(e) immediately confirms that the long axis of the HA nanocrystals is parallel to the collagen fiber axis, as expected from the literature (Paris, 2008; Dunlop & Fratzl, 2010). The orthogonality of the  and  HA crystal axis orientations in Figs. 4(e) and 4(f) confirms the reliability of the automated orientation extraction.
As a further step in the statistical analysis of the two-dimensional SAXS/WAXS scanning maps, four representative one-dimensional WAXS patterns were extracted from the whole data set, for each sample, by using a PCA approach, searching for the lowest correlated signals (Abdi & Williams, 2010). The two-dimensional/one-dimensional WAXS frames/patterns have already been shown in Figs. 3(a)–3(d) and coupled to the two-dimensional/one-dimensional SAXS frames/patterns of Figs. 3(e)–3(h). It is worth specifying that the PCA analysis was applied to one-dimensional WAXS patterns, instead of one-dimensional SAXS, because they contain more quantitative information, i.e. more diffraction peaks, readily accessible to a robust model-dependent analysis.
In the last analysis step, considering that each SAXS two-dimensional frame was previously divided into 16 angular sectors, each 22.5° wide, meridional and equatorial scattering profiles were easily identified for each SAXS frame in Figs. 3(e)–3(h). From them, three main one-dimensional SAXS profiles were taken as model variables in a canonical correlation analysis (CCA) (Laudadio et al., 2005; Ladisa et al., 2007). They are depicted in Fig. 5(a) for sample #5004: meridional (magenta), equatorial (yellow) and background profiles (black) have been vertically shifted for the sake of clarity. The CCA is a statistical method which quantifies the relationship between two random variables: the x variable (models) representative here of the meridional, equatorial and background profiles; and the y variable consisting here of SAXS intensity profiles obtained from the original two-dimensional frames once folded into 16 profiles, one for each 22.5°-wide angular segment (see also Fig. S1 ). The CCA allows us to quantify the degree of linear dependence between x and y, providing the so-called correlation coefficient, a scalar quantity with a value between 0 and 1. Through CCA, the contribution of the three major principal signals in Fig. 5(a) was determined for each point, and for each sector, of the scanning two-dimensional SAXS image of Fig. 5(a) returning 16 nosological maps. In the maps the pixel color is defined by using the CMYK color map: the cyan component is set to zero; magenta represents correlation with the collagen meridional scattering model; yellow, the correlation with the collagen equatorial scattering model; black, the correlation with the background model (Ghersetich et al., 1994). The sector shown in Fig. 5(b) was selected, among the 16 available, as one of the nosological maps not dominated by either the equatorial or the meridional signals but with a significant content of information in terms of relative abundance of both.
The SAXS/WAXS microscopy images provide a full description of the local abundance of collagen fibrils and their HA mineralization.
Fig. 1 summarizes some of the experimental findings: comparing the third-order collagen meridional signal, shown in the third column, with the map of the HA(002) crystals, shown in the fourth column, it is evident that, for all samples, the colors in these images and thereby the orientation of collagen and HA  coincide. This confirms both prior findings, namely that the hydroxyapatite  crystal direction is aligned with collagen fibers (Paris, 2008; Dunlop & Fratzl, 2010) and the robustness of the data analysis of these two independent measurements.
For verification purposes the collagen first meridional diffraction order SAXS map and the HA  WAXS map are also reported in the supporting information in Fig. S2 . Here, the colors, i.e. orientation, of the first and third collagen map coincide, as expected, since they are caused by the same repeat distance along fibers (d ≃ 65 nm = d1 = d2/3). In the crystallographic structure of HA, the  and  crystal axes are perpendicular to each other. Comparing the corresponding maps in Fig. S2 , their orientation – encoded as color – is perpendicular to each other, demonstrating that we are correctly probing two perpendicular signals from crystals having identical orientation.
The CCA of the SAXS signals reveals the abundance of collagen equatorial scattering (sensitive to the packing of mineralized collagen fibrils in bundles) and collagen meridional scattering (sensitive to the about 65 nm repeat distance occurring along the axis of each fibril/fiber). These two signals are coded as yellow and magenta in the images shown in the second column of Fig. 1. By visual inspection one might get the impression that the CCA images of the samples from younger patients (#5046 and #5108) have a slightly more yellow color than the images for the samples from older patients, which tend more towards magenta. In order to quantify this tendency towards magenta, i.e. increased meridional scattering, we computed across the entire CCA map the ratio between the correlation coefficients of the meridional and equatorial scattering, here called the MECR value, which is indicated in Figs. 1 and S2 . Each MECR value may vary by about ±0.02 with respect to those segments similar to that represented in Fig. 1, i.e. not dominated by any of the equatorial or meridional signals. The derived values are 0.65–0.71 for patients in the age range from 62 to 71 and increase with age up to a maximum of 0.78 for the oldest patient of 87 years. By calculating the correlation coefficient of MECR as a function of age from the data in Fig. 1, an R2 value equal to 0.77 is found, which is statistically significant at the 5% level (meaning that MECR and age are correlated at a 95% level), pointing towards a linear dependence between MECR and age. Actually the MECR values reported in Fig. 1 also concide with the median evaluated for each sample on all the 16 sectors. The median is, among the statistical measures of central tendency, the one that most fulfills the property of robustness, with respect to mode or mean, in the presence of anomalous data (Kenney & Keeping, 1962) – which are expected here, at least in the angular sector across the beam stopper.
In elderly individuals, the level of joint immobilization is expected to increase with age. The loss of movement is also concomitant with a decrease in the quantity of water accumulated in the body and with a reduction of glycosaminoglycans, especially hyaluronic acid, in most tissues. These are effects typical of aging (Ghersetich et al., 1994). In other pathologies, it was found that aging produces an increase of the collagen cross-links (Zimmermann et al., 2011; Saito & Marumo, 2010). These increased levels of cross-linking with age stiffen the collagen fibrils, thereby affecting the plasticity of the bone (Zimmermann et al., 2011). How would these changes be manifested at a supra/submolecular level? As already explained, the collagen meridional and equatorial diffraction signals are related to the order along the collagen fibers (meridional) and to the order of the fibril in bundles (equatorial). An increased cross-linking, promoting the stiffening of the fibers, is expected to cause an increased level of the meridional versus equatorial scattering signal, as observed in the data. In other words, more rigid fibers allow the scattering signal due to the order along the fiber axis to be more pronounced. In other connective tissues, an increase in the mechanical integrity of the collagen network with age (Wang et al., 2002), which causes much more rigid fibers, loss of elasticity and a consequent brittleness of the bones, was also found.
In conclusion, in spite of having statistically little relevance as collected on a small group of patients, this complex crystallographic study performed with scanning SAXS/WAXS microscopy allowed us to mark a structural variation at the meso/nanoscale with age in OA pathology. The enhancement in the collagen scattering with age was supposed to be related to an increased level of cross-links, inducing stiffening of the collagen fibrils.
The authors would like to thank the SEED project `X-ray synchrotron class rotating anode microsource for the structural micro imaging of nanomaterials and engineered biotissues (XMI-LAB)', IIT Protocol No. 21537 of 23 November 2009, and the FIRB 2009/2010 project `Rete integrata per la Nano Medicina (RINAME)', RBAP114AMK_006, and Caterina Chiarella for her support with this project. The authors acknowledge financial support from the European Commission under the Seventh Framework Programme by means of the grant agreement for the Integrated Infrastructure Initiative No. 262348 European Soft Matter Infrastructure (ESMI). Davide De Cristoforo (DDC) and Andreas Menzel are acknowledged for critical reading of the paper and DDC also for fruitful discussions on the OA pathology. Xavier Donath is acknowledged for technical assistance with the measurements at the cSAXS beamline.
Abdi, H. & Williams, L. J. (2010). Comput. Stat. 2, 433–459. CrossRef
Bailey, A. J. (2001). Mech. Ageing Dev. 122, 735–755. Web of Science CrossRef PubMed CAS
Bailey, A. J., Light, N. D. & Atkins, E. D. (1980). Nature, 288, 408–410. CrossRef CAS PubMed Web of Science
Baker-LePain, J. C. & Lane, N. E. (2012). Bone, 51, 197–203. Web of Science PubMed
Bigi, A., Ripamonti, A., Roveri, N., Compostella, L., Roncon, L. & Scivazappa, L. (1982). Int. J. Biol. Macromol. 43, 87–92.
Boyde, A., Bianco, P., Portigliatti Barbos, M. & Ascenzi, A. (1984). Metab. Bone Dis. Relat. Res. 5, 299–307. CrossRef CAS PubMed Web of Science
Boyde, A. & Riggs, C. M. (1990). Bone, 11, 35–39. CrossRef CAS PubMed Web of Science
Bünger, M. H., Oxlund, H., Hansen, T. K., Sørensen, S., Bibby, B. M., Thomsen, J. S., Langdahl, B. L., Besenbacher, F., Pedersen, J. S. & Birkedal, H. (2010). Calcif. Tissue Int. 86, 294–306. Web of Science PubMed
Bunk, O., Bech, M., Jensen, T. H., Feidenhans'l, R., Binderup, T., Menzel, A. & Pfeiffer, F. (2009). New J. Phys. 11, 123016. Web of Science CrossRef
Dunlop, J. W. & Fratzl, P. (2010). Annu. Rev. Mater. Res. 40, 1–24. Web of Science CrossRef CAS
Gentleman, E., Lay, A. N., Dickerson, D. A., Nauman, E. A., Livesay, G. A. & Dee, K. C. (2003). Biomaterials, 24, 3805–3813. Web of Science CrossRef PubMed CAS
Ghersetich, I., Lotti, T., Campanile, G., Grappone, C. & Dini, G. (1994). Int. J. Dermatol. 33, 119–122. CrossRef CAS PubMed Web of Science
Giannini, C., Siliqi, D., Bunk, O., Beraudi, A., Ladisa, M., Altamura, D., Stea, S. & Baruffaldi, F. (2012). Sci. Rep. 2, 435. Web of Science CrossRef PubMed
Hoaglund, F. T. & Steinbach, L. S. (2001). J. Am. Acad. Orthop. Surg. 9, 320–327. PubMed CAS
Kay, M. I., Young, R. A. & Posner, A. S. (1964). Nature, 204, 1050–1052. CrossRef PubMed CAS Web of Science
Kenney, J. F. & Keeping, E. S. (1962). Mathematics of Statistics, Part 1, 3rd ed., pp. 32–35, 52–54 and 211–212. Princeton: Van Nostrand.
Kraft, P., Bergamaschi, A., Bronnimann, C., Dinapoli, R., Eikenberry, E. F., Graafsma, H., Henrich, B., Johnson, I., Kobas, M., Mozzanica, A., Schleputz, C. M. & Schmitt, B. (2009). IEEE Trans. Nucl. Sci. 56, 758–764. Web of Science CrossRef CAS
Ladisa, M., Lamura, A. & Laudadio, T. (2007). EURASIP. J. Adv. Signal Process. article 19260.
Laudadio, T., Pels, P., De Lathauwer, L., Van Hecke, P. & Van Huffel, S. (2005). Magn. Reson. Med. 54, 1519–1529. Web of Science CrossRef PubMed
Lotz, M. & Loeser, R. F. (2012). Bone, 51, 241–248. Web of Science CrossRef CAS PubMed
Luyten, F. P. & Vanlauwe, J. (2012). Bone, 51, 289–296. Web of Science CrossRef CAS PubMed
Martel-Pelletier, J., Wildi, L. M. & Pelletier, J. P. (2012). Bone, 51, 297–311. Web of Science CAS PubMed
Olszta, M. J., Cheng, X., Jee, S. S., Kumar, R., Kim, Y. Y., Kaufman, M. J., Elliot, P., Douglas, E. P. & LaGowera, L. B. (2007). Mater. Sci. Eng. R, 58, 77–116. Web of Science CrossRef
Orgel, J. P. R. O., Miller, A., Irving, T. C., Fischetti, R. F., Hammersley, A. P. & Wess, T. J. (2001). Structure, 9, 1061–1069. Web of Science CrossRef PubMed CAS
Paris, O. (2008). Biointerphases, 3, FB16–FB26. Web of Science CrossRef PubMed
Rho, J. Y., Kuhn-Spearing, L. & Zioupos, P. (2002). Med. Eng. Phys. 20, 92–102. Web of Science CrossRef
Saito, M. & Marumo, K. (2010). Osteoporos Int. Feb. 21, 195–214. Web of Science CrossRef CAS
Suhonen, H., Fernández, M., Serimaa, R. & Suortti, P. (2005). Phys. Med. Biol. 50, 5401–5416. Web of Science CrossRef PubMed
Wang, X., Shen, X., Li, X. & Mauli Agrawal, C. (2002). Bone, 31, 1–7. Web of Science CrossRef PubMed
Weiner, S. & Wagner, H. D. (1998). Annu. Rev. Mater. Res. 28, 271–298. Web of Science CrossRef CAS
Wilkinson, S. J. & Hukins, D. W. L. (1999). Radiat. Phys. Chem. 56, 197–204. Web of Science CrossRef CAS
Zhang, Y. & Jordan, J. M. (2010). Clin. Geriatr. Med. 26, 355–369. Web of Science CrossRef PubMed
Zimmermann, E. A., Schaible, E., Bale, H., Barth, H. D., Tang, S. Y., Reichert, P., Busse, B., Alliston, T., Ager, J. W. & Ritchie, R. O. (2011). Proc. Natl Acad. Sci. USA, 30 14416–14421.
© International Union of Crystallography. Prior permission is not required to reproduce short quotations, tables and figures from this article, provided the original authors and source are cited. For more information, click here.