Nanostructure of the neurocentral growth plate: Insight from scanning small angle X-ray scattering, atomic force microscopy and scanning electron microscopy

Bone 39 (2006) 530 – 541 www.elsevier.com/locate/bone

Nanostructure of the neurocentral growth plate: Insight from scanning small angle X-ray scattering, atomic force microscopy and scanning electron microscopy Mathias Hauge Bünger a,c,d , Morten Foss c,d,⁎, Kurt Erlacher c,e,1 , Mads Bruun Hovgaard c,d , Jacques Chevallier c,d , Bente Langdahl a , Cody Bünger b,c , Henrik Birkedal c,e , Flemming Besenbacher c,d , Jan Skov Pedersen c,e a

Department of Endocrinology and Metabolism C, Aarhus University Hospital, Tage Hansens gade 2, DK-8000 Aarhus, Denmark b Orthopaedic Research Laboratory, Aarhus University Hospital, Nørrebrogade 44, DK-8000 Aarhus, Denmark c Interdisciplinary Nanoscience Center (iNANO), University of Aarhus, Ny Munkegade, DK-8000 Aarhus, Denmark d Department of Physics and Astronomy, University of Aarhus, Ny Munkegade, DK-8000 Aarhus, Denmark e Department of Chemistry, University of Aarhus, 140 Langelandsgade, DK-8000 Aarhus, Denmark Received 28 September 2005; revised 13 March 2006; accepted 16 March 2006 Available online 12 June 2006

Abstract In this study, the experimental techniques scanning electron microscopy (SEM) including energy-dispersive X-ray analysis, atomic force microscopy (AFM) and scanning small angle X-ray scattering (SAXS) have been exploited to characterize the organization of large molecules and nanocrystallites in and around the neurocentral growth plate (NGP) of a pig vertebrae L4. The techniques offer unique complementary information on the nano- to micrometer length scale and provide new insight in the changes in the matrix structure during endochondral bone formation. AFM and SEM imaging of the NGP reveal a fibrous network likely to consist of collagen type II and proteoglycans. High-resolution AFM imaging shows that the fibers have a diameter of approximately 100 nm and periodic features along the fibers with a periodicity of 50–70 nm. This is consistent with the SAXS analysis that yields a cross-sectional diameter of the fibers in the range of 90 to 112 nm and a predominant orientation in the longitudinal direction of the NGP. Furthermore, we find inhomogeneities around 7 nm in the NGP by SAXS analysis. Moving towards the bone in the direction perpendicular to the growth plate, a systematic change in apparent thickness is observed, while the large-scale structural features remain constant. In the region of bone, the apparent thickness equals the mean mineral thickness and increases from 2 nm to approximately 3.5 nm as a function distance from the NGP. The mineral particles are organized as plates in a rather compact network structure. We have demonstrated that SEM, AFM and SAXS are valuable tools for the investigation of the organization of large molecules and nanocrystallites in the NGP and adjacent trabecular bone. Our findings will be an important basis for future work into identifying the defects on nanometer length scale responsible for idiopathic scoliosis and other growth-plate-related diseases. © 2006 Elsevier Inc. All rights reserved. Keywords: Endochondral ossification; Biomineralization; Growth plate; Small angle X-ray scattering; Atomic force microscopy

Introduction

⁎ Corresponding author. The Interdisciplinary Nanoscience Center (iNANO), University of Aarhus, Building 1521-311, Ny Munkegade, 8000 Aarhus C, Denmark. Fax: +45 89 42 36 90. E-mail address: [email protected] (M. Foss). 1 Present address: Bruker AXS Inc., 5465 E. Cheryl Parkway, Madison, WI 53711, USA. 8756-3282/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.bone.2006.03.013

In vertebrates the longitudinal growth of bone during childhood originates from the growth plate. Traditionally, all growth plates are divided into four different morphological zones: the resting zone, the proliferative zone, the hypertrophic zone and the zone of provisional mineralization. The central resting zone contains stem cells that supply the outer layers of the growth plate with proliferating chondrocytes [1,2]. In the proliferative

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

zone, the chondrocytes replicate into cellular columns and grow outward into the hypertropic zone, where they mature and eventually undergo apoptosis [1,2]. Simultaneous with these cellular events, the composition and structure of the extracellular matrix changes from cartilage to mineralized bone [1]. Here, we study this gradual change from cartilage to bone in one neurocentral growth plate (NGP) using a series of experimental techniques providing structural information on the nanometer length scale. The NGP separates the pedicle from the central part of the vertebral body (Fig. 1A). The NGP is bipolar with respect to endochondral bone formation [3–5], and the NGP is believed to be especially important for the formation of the pedicle [6]. In humans, the NGPs are replaced by bone between the age of 11 and 16 years [4,5]. In the central part of the growth plate, the matrix primarily consists of water, collagen type II, proteoglycans and about 10% of collagen type I [7,8]. In the hypertrophic zone, also collagen type IX, X and XI are found [9,10]. Collagen type IX and XI are attached to the surface of collagen type II and in part control the diameter of the collagen type II fibrils and bind them together in a fibrillar network [11–14]; a detailed understanding of the role of collagen type X is still

531

missing [15]. The hypertrophic chondrocytes produce several non-collagenous proteins, known to affect bone formation including alkaline phosphatase, osteocalcin, osteopontin and bone sialoprotein [10]. The provisional mineralization occurs within this organic matrix rich in collagen type II between the columns of hypertrophic chondrocytes [16]. In this region, the mineral crystals show no preferred orientation and are not organized with respect to, nor formed within, the collagen fibrils [16]. Later on, blood vessel invasion occurs and the provisional mineral deposits are resorbed by osteoclasts and replaced by a new mineralizable matrix produced by osteoblasts [1] mainly consisting of collagen type I. The mineralization process creating carbonated hydroxyapatite (HA) crystals is influenced by noncollagenous proteins [17–21], and several studies have evidenced that nucleation preferentially takes place within gaps of the assembled collagen fibrils, referred to as hole zones [22–25]. The newly formed HA crystals are believed initially to grow preferentially along their long axis following the orientation of the collagen fibrils followed by growth in the crystallite thickness [24]. The growth of the HA crystals is thought to be directed by the structure of the collagen molecule and non-collagenous proteins [17,24]. As the mineralization process proceeds, some

Fig. 1. (A) A computerized tomography (CT) image of a pig vertebrae L2: P = pedicle, CPVB = central part of the vertebral body, NGP = neurocentral growth plate. The NGP crosses the vertebral bone and is approximately 150–200 μm wide and 20 mm long. (B–F) Scanning electron microscopy pictures of the different morphological zones in a gold-coated growth plate. The image contrast has been adjusted in Adobe Photoshop®. (B) An overview of the growth plate at low magnification. The high magnification images roughly correspond to the different morphological zones of the growth plate. (C) Mineralized bone, (D) outer hypertrophic zone, (E) inner hypertrophic zone and (F) resting zone.

532

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

crystals fuse by coplanar fusion to form larger plates and collagen becomes filled up with parallel HA sheets [24,26,27]. In the past, most structural knowledge about collagen and mineral interaction in bone has been obtained using transmission electron microscopy (TEM) and scanning electron microscopy (SEM) [24,26–28]. Only a limited number of studies have employed other experimental techniques such as nuclear magnetic resonance (NMR), scanning small angle X-ray scattering (sSAXS) and atomic force microscopy (AFM) [28–32]. While SEM and AFM are strong imaging techniques with nano- to mesoscopic resolution in direct space, sSAXS is a reciprocal space technique based on the scattering of X-rays. It offers integral data of morphology over the same length scales as AFM and SEM and hereby ideal complementary information to these techniques. The sSAXS technique makes it possible to investigate nanoscale structures in hierarchical structures such as bone with a lateral resolution determined by the X-ray beam size (∼100 μm with laboratory sources, <5 μm with synchrotron radiation). It has previously been used to gain information about size, shape and orientation of mineral particles in bone and at the bone/cartilage interface [30,33–36]. The aim of the present study was to explore the possibilities of a series of complementary techniques like sSAXS, AFM, SEM and SEM–energy-dispersive X-ray microanalysis (SEM–EDAX) in the characterization of the nano- and mesoscopic organization of molecules and crystallites in a representative growth plate in a pig. We show by this illustrative single-sample study that sSAXS can provide new and deep insights into bone and for the first time cartilage nanostructure. It is important to understand the formation and architecture on the nanometer scale of healthy bone for the future investigations of the possible microscopic changes related to diseases originating from growth plates. Materials and methods Sample preparation From a 6-month-old healthy Danish Landrace pig, the same transversal midbody section of pig vertebrae L4 was used for the sSAXS and SEM–EDAX studies of the NGP, while other parallel sections were used for SEM and AFM imaging. The sections were found to be representative for the neurocentral growth plate of 6-month-old pigs by histological analyses of parallel sections from the investigated vertebrae and of sections from 5 other 6-month-old animals. Only a single sample was studied in detail by sSAXS, AFM and SEM– EDAX with the purpose of illustrating the information that can be obtained on the bone, bone–cartilage interface and cartilage nanostructure by combining these techniques. The bone sections were cut with a water cooled diamond saw (Exakt®, Apparatebau, Nordenstedt, Germany) with a thickness of 190 μm and 1600 μm for sSAXS/SEM–EDAX and SEM/AFM studies, respectively. The miscut angle on the sample with respect to the transversal plane was judged to be below 15°. Subsequently, the section was subdivided by a scalpel into smaller samples of approximately 10 × 10 mm and the sample that included the growth plate was selected for further studies. To remove debris from sawing, the sample was gently washed in deionized water for approximately 10 s. The samples were dehydrated in a graded series of ethanol (70–99%) and then freeze-dried to increase the contrast in electron density in the growth plate and to allow SEM analysis on the same sample. A cross-sectional CT image of the lumbar spine (Fig. 1A) of a pig was kindly made available by Dr. Zou and Dr. Li (Department of Orthopedics, Aarhus University Hospital, Denmark) [37]. The study was conducted in accordance with the Helsinki II declaration on research animal ethics and approved by the local ethical committee.

Scanning electron microscopy and EDAX SEM and SEM–EDAX for elemental analysis was performed using a Maxim CamScan electron microscope. High-resolution SEM imaging of the growth plate and the surrounding bone was performed after coating the sample with an approximately 23-nm-thick layer of gold. Imaging was performed by detection of secondary electrons. SEM–EDAX was done on the same uncoated sample used for the sSAXS study. The measurements were performed under low vacuum using an electron energy of 20 keV. SEM imaging for EDAX was performed by detection of back scattered electrons (BSE). EDAX spectra showed that the main elements were C, P, S and Ca. Maps and lines of the distribution of these four elements were obtained by measuring the intensity of the Kα X-ray emission peaks of the target elements, while scanning over the sample surface.

Atomic force microscopy (AFM) Atomic force microscopy of the growth plate and adjacent bone was conducted using a Nanoscope IIIa Veeco® AFM (Digital Instruments, Santa Barbara, California, USA). All images were taken in AFM tapping mode under ambient conditions using MacLevers TII cantilevers (Molecular Imaging, Tempe, Arizona, USA) with typical radius of curvature smaller than 10 nm. In order to locate the position of interest, an optical light microscope was attached to the AFM.

sSAXS The sSAXS investigations were performed on a modified SAXS Nanostar camera (Bruker-AXS GmBH, Germany) [38]. The SAXS instrument uses a roting copper anode X-ray generator (45 kV/90 mA; Cu-Kα, λ = 0.154 nm). The X-ray beam was focused through a pair of cross-coupled Göbel mirrors and collimated by a three-pinhole system with diameters of 750, 100 and 200 μm placed as previously described [38]. The first two pinholes define the divergence and the beam profile, the third one acts as an anti-scattering pinhole which helps in reducing background scattering. The scattered photons were detected with a Bruker AXS HI-STAR area detector. A beam stop (Ø = 3.0 mm) was placed in front of the detector. The sampledetector distance was 662 mm, which gave simultaneous access to a range of scattering vector moduli q of about 0.01 to 0.35 Å−1. The scattering vector modulus q is given by the wavelength λ of the X-rays and the scattering angle 2θ by q = 4πsinθ/λ. The bone sample was mounted with the plane of the sample perpendicular to the X-ray beam in a sample holder, which could be moved perpendicular to the beam with a precision better than 0.1 μm in the horizontal (x) and vertical direction (y). The diameter of the X-ray beam at the sample position was slightly larger than 100 μm, which defines the nominal lateral resolution. Consequently, all data correspond to the average over a specimen volume laterally defined by the diameter of the X-ray beam. In all experiments, the step width in the x and y directions was 50 μm, which allowed detecting structural changes occurring on a length scale smaller than the nominal resolution. In order to determine the positions of interest, a survey 2D transmission scan and a survey 2D SAXS intensity scan of the entire sample, approximately 10 × 10 mm, were conducted with a measuring time of 1 s per point (Figs. 2A and B). The transmission scan was carried out by placing a uniform strongly scattering specimen (glassy carbon) right after the bone sample. In this case, the integrated scattering intensity is proportional to the transmission on that specific position of the bone. Following the initial scan, a randomly selected region containing both bone and growth plate was chosen for more detailed investigations. This region consisted of 299 data points and measured 650 × 1150 μm (Fig. 2C). The region was sufficiently large to map the intrinsic variations along the growth plate. In order to improve the statistics, the measurement time was increased to 100 s per point.

sSAXS data analysis The SAXS signal is due to variations in the electron density of the sample on the nanometer length scale, and the method thus yields structural information on that length scale. In bone, the variation in electron density is mainly due to differences in electron density between inorganic minerals and organic molecules. Thus, bone can for the purpose of SAXS data interpretation be considered as a two-phase material. In this good approximation, the SAXS technique offers unique information about

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

533

Fig. 2. X-ray transmission and sSAXS images of the vertebral cross-section containing the growth plate. The pixel size in A, B and C are 50 μm. (A) Survey X-ray transmission scan. High transmission (light colors) corresponds to low electron density, while low transmission (dark colors) results from high electron density. The latter comes from high mineral density. (B) Survey sSAXS intensity scan. The plot shows the integral SAXS intensity in the q range from 0.01 to 0.35 Å−1. High SAXS intensities (light colors) result from large variation in electron density. (C) High-resolution sSAXS scan of the selected region of interest marked in B. The color coding is the same as in B. (D) Examples of raw 2D SAXS images from various regions shown in C. Horizontal trabeculae cause a stronger scattering in the vertical direction because the elongated mineral particles are aligned horizontally along the collagen fibers in the direction of the trabeculae. q and χ are the scattering vector and azimuthal angle, respectively.

mineral plate thickness, orientation, shape and overall dimensions. In the unmineralized region of the growth plate, the contrast originates from the difference in electron densities between organic molecules and air. Here, knowledge about molecule size and substructure can be obtained. The 2D SAXS patterns (e.g. Fig. 2D) were analyzed following the approach of Rinnerthaler and Fratzl with regard to the thickness parameter T predominant orientation and degree of orientation of the bone mineral particles [33–36]. In addition, the shape and an overall size of particles were determined using additional data processing procedures applying a generalized Debye–Bueche expression to bone/cartilage for the first time (vide infra).

the particles perpendicularly to the beam [33], and we will henceforth refer this projected degree of orientation as PDO. The PDO is found by dividing the anisotropic part of the Iq(v) distribution with the total intensity. pffiffiffiffiffiffi 2 2pðImax  Ibg Þr PDO ¼ pffiffiffiffiffiffi : 2 2pðImax  Ibg Þr þ 360Ibg

Anisotropy I(χ): projected orientation

Angular dependence I(q): size and shape

The presence of oriented structures in a sample, for example, oriented mineral plates in bone, results in anisotropic or nonspherical SAXS 2D patterns (e.g. Fig. 2D). Orientation of structures within the plane of the section thus causes an azimuthal, χ, dependence of the intensity. This azimuthal dependence of the intensity, which is directly related to the predominant orientation of structures, can be determined from the radially integrated intensity, Iq(χ). In the region of bone, it typically results in Iq(χ) distributions with two symmetrical peaks separated by 180°. In an automated procedure using a home-written analysis program, each Iq(χ) curve was fitted with two Gaussian curves that had the same height and width and were separated by 180°, Ibg + Imax[g(χ; ψ) + g(χ; ψ + 180°)] with g(χ; ψ) = exp[−(χ − ψ)2 / (2σ)2]. From the fits, the maximum Imax, the width σ and the position ψ of the Gaussian curves as well as the background intensity Ibg were determined. The predominant orientation of structures in real space corresponds to ψ + 90°. In addition, a parameter that describes the degree of orientation of structures within the probed sample volume can be estimated. Both the predominant orientation and degree of orientation should be regarded as the 2D projection of

Information about the particle shape α, characteristic thickness T and overall length scale Rg is obtained from the q dependence of the SAXS intensity averaged over the azimuthal angle I(q). At a given scattering vector modulus q, a typical length scale in the investigated material ξ ≈ π/q is probed. Information about size and shape of particles on different length scales is thus found in different domains of the I(q) distribution. The overall length scale of the particles Rg and the shape parameter α is determined from data at small values of q, while the thickness parameter T is determined at large values of q (Fig. 3). In the small-q range, the Guinier approximation gives, for finite size objects, a universal behavior I(q) ∝ 1 − q2Rg2/3, where the Guinier radius, Rg, is related to the overall size of the object (see e.g. [39]). For plates, like mineralites in bone, Rg gives the average plate dimensions (Fig. 4). For very long cylindrical objects like collagen molecules, the Guinier is related to ffiffiffiffiffiffiffiffi pbehavior the cross-sectional dimensions and R ¼ 2Rg , where R is the cylinder radius. The Guinier behavior is observed for q<1/Rg. For the data presented herein, this region is almost only observed in the growth plate because the plate size of the mineralites in the bone is large and the information needed to determine their cross-

534

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

sectional dimensions thus occurs at scattering vectors moduli below the smallest one accessed in the experiment. For q values larger than those in the Guinier region, a powerlaw behavior: I(q) ∝ q−α is observed for simple geometric objects like rods/cylinders/needles and plates. The exponent α is related to the shape of the objects generating the SAXS signal; it is from hereon called the shape parameter. For example, needleshaped and plate-like particles show α = 1 and α = 2, respectively. The exponent of the power law is thus directly related to the dimensionality of the particles considered. For non-integer values of α, the particles are termed fractals of dimensionality df = α [40]. The higher the value of df, the more dense and compact is the structure of the aggregate. If the particle interface has a graded profile or if part of the Guinier region is probed, an apparent exponent larger than α = 4 can be obtained [41]. In the high-q range, the scattering intensity follows the Porod behavior for all types of particles with a sharp surface. In this regime, the intensity is given by I(q) = Pq−4, P being the Porod constant [39], which for a two-phase system is proportional to the total interface R l area. The thickness parameter is obtained from T ¼ 4 0 q2 IðqÞdq=ðpPÞ [40]. For two-phase systems with sharp interfaces (like mineral particles embedded in collagen), T describes the volume to surface ratio of the particles without any assumptions about shape [33]. Specifically, if the particles are plate-like, the T parameter is a measure of the mean thickness (Fig. 4) [42]. The SAXS 2D intensity frames were averaged over the azimuthal angle in the q range from 0.00966 Å−1 to 0.3484 Å−1 in q steps of 0.0014 Å−1. For determination of the thickness parameter T, we used an automated procedure using homewritten least-squares software. In the Porod region, q > 0.20 Å−1 (Fig. 3), the Porod constant R l P was determined by fitting I (q) = Pq−4. The integral 0 q2 IðqÞdq was evaluated numerically in the q range qmin–qmax. The contributions outside this range were calculated analytically by extrapolating in the following way: I(q) = I(qmin) for q < qmin and I(q) = Pq−4 for q > qmax. Thereafter, T could be calculated [33].

Fig. 4. Left: schematic drawing of mineral crystallite with indication of the measured quantities: the thickness parameter T and the average plate dimension R, which is related to the radius of gyration extracted at small q. Right: schematic drawing of the proposed structured fibrils seen in the growth zone. The measured radius of gyration is a measure of the cross-sectional dimensions.

Information about the overall length scale of the particles, Rg and the shape parameter α was derived from the small-q region (Fig. 3). However, the beamstop shadows the scattering in the small-q region even outside the completely covered region. Therefore, data below q = 0.014 Å−1 were corrected by dividing the SAXS intensity curves with a sensitivity curve derived from the scattering of glassy carbon (see [38] for details). For determination of Rg, we used the generalized Debye–Bueche expression: " !#a=2 2 2 I ð qÞ ¼ 1 þ q 3aR2g This expression is at low q equivalent to the I(q) ∝ 1 − q2Rg2 / 3 Guinier approximation, and it crosses over to the q−α as q is increased. When applied to data from spherical homogeneous particles or from the Guinier region originating from the crosssectional dimension of elongated particles, values for α of 5–6 are found. The experimental data from 0.015 to 0.06 Å−1 were fitted and α and Rg derived; in large parts of the bone, Rg could not be precisely determined because the smallest q accessed in the experiment was not low enough for the large structures present. The standard deviation of α was 2–11%. The uncertainty of T was dominated by the error on the determination of P, which was on the order of 3%. A rough estimate of the error on the PDO is approximately 10%. Data points in the empty marrow regions were excluded using a filter variable based on the integrated SAXS intensity and the transmission intensity. The remaining data analysis and visualization were done using Matematica5®. Results

Fig. 3. The SAXS intensity corrected for absoption plotted against scattering vector q, for four representative positions: Growth plate centre (blue), 50 μm from the growth plate middle (red), 100 μm from the growth plate middle (black), and trabecular bone (green). Correction for shadowing by the beamstop has been made in the q-region below 0.02 Å−1.

SEM Fig. 1 presents a selection of representative SEM images of the NGP taken in the different morphological regions of the

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

growth plate. At low magnification, the unmineralized part of the growth plate can be viewed as the dark area in the center of the growth plate with a width of approximately 100 μm (Fig. 1B). The unmineralized part of the growth plate comprises the resting zone, the proliferative zone and inner part of the hypertrophic zone. Figs. 1E and F are taken within the unmineralized central part of the growth plate. In the inner central part (Fig. 1F), fibrous structures of approximately 300–400 nm in diameter are observed. The orientation of the fibrous structures is along the direction of the growth plate, and the structures in this region appear to cross the plane of the section. The fibrous structures consist of individual fibers closely aligned with respect to each other. The thickness of these fibers is approximately 100 nm. Toward the edge of the unmineralized part of the growth plate, the fibers appear to be aligned perpendicular to the direction of the growth plate (Fig. 1E). Furthermore, it is easier to identify the individual fibers that form a fine mesh of filaments, which is more porous than in the NGP center. Figs. 1C and D are obtained in the mineralized part of the growth plate. Fig. 1D is from the outer hypertophic zone, including the region of provisional calcification, while Fig. 1C

535

is from the region of young primary bone. In the hypertrophic zone, the fibrous filaments seen in the unmineralized part of the NGP are less visible and the structure seems denser. In the region of primary bone, fibrous structures are not as evident and the surface appears more flat and solid with less porosity (Fig. 1C) in agreement with other SEM studies on trabecular bone [43]. SEM–EDAX elemental analysis Fig. 5A is a BSE-SEM image of the NGP and bone, covering the same region as investigated with sSAXS. Provisional calcification is seen as a bright white zone on both sides of the center of the NGP. Lacunas from hypertrophic chondrocytes are seen extending from the NGP center. Figs. 5B and C show the distribution of C, S, Ca and P obtained by SEM–EDAX. The investigations show, as expected, a high density of carbon and sulfur in the growth plate in agreement with the presence of collagens and proteoglycans, whereas calcium and phosphor principally are found in the trabecular bone (Figs. 5B and C). Ca and P are expected to be concentrated in the hydroxyapatite (HA) [44].

Fig. 5. (A) BSE-SEM image of the growth plate and the neighboring region of bone (the same region as Fig. 2C and Fig. 7). (B) SEM–EDAX line scan (orange in A): content of C, S, Ca and S. (C) SEM–EDAX maps of content of Ca, P, S and Ca and S. Carbon and sulfur are especially present in the growth plate, whereas calcium and phosphorous are predominantly found in the bone. Furthermore, sulfur is found within the lacunas of the hypertrophic chondrocytes.

536

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

Interestingly, the high density of S in the NGP center extends into the lacunas of the hyperthropic chondrocytes in the region of provisional calcification (Figs. 5C—Ca and S). Atomic force microscopy AFM images of the central part of the growth plate reveal fibrous structures arranged either in a fibrous mesh without overall preferred orientations as seen in Fig. 6A or as more densely packed fibers highly aligned in the direction perpendicularly to the plane of the section (Fig. 6C). In the areas without overall preferred orientations (Fig. 6A), some fibers are seen to lie separate, without close association to other fibers. However, most fibers are found to bundle up into larger fiber bundles composed of fibers closely aligned with each other. The individual fibers are approximately 100 nm in thickness in agreement with the SEM observations. Detailed scans at higher magnification reveal the internal surface of the fibers (Fig. 6B). The fibers have a periodicity of approximately 50–70 nm (Fig. 6B). Furthermore, in some regions, smaller globular structures located on the surface of the fibers are observed. In places where the fibers are perpendicular to the surface, it appears as if the fibers have been cut and have curled up (Fig. 6C). sSAXS The 10 × 10 mm2 X-ray transmission image (radiography) of the vertebral bone sample including the NGP is shown in Fig. 2A. The transmission is low in regions with trabecular bone and high in the growth plate and even higher in the bone marrow cavities. Since the attenuation of the transmitted intensity is completely dominated by absorption, it reflects the electron density of the sample volume probed by the X-ray beam. Thus, the electron density is high in the trabecular bone and low in the NGP. Fig. 2B shows the corresponding SAXS intensity image. Interestingly, we observe high SAXS intensities not only in the bone, but also in the growth plate. High SAXS intensities are present where there is a large variation in electron density on the nanometer length scale [39] (typically 1–100 nm). The combination of a high transmission and a high SAXS intensity indicate that the NGP mainly consists of unmineralized tissue and that there are large variations in the electron density, and thus of the structure, on the nanometer length scale. The area of the high-resolution scan is illustrated in Fig. 2C, while Fig. 2D gives examples of individual SAXS measurements. We observe a strong anisotropy of the 2D SAXS patterns when measuring on the mature trabecular bone (Fig. 2D), illustrating the preferred orientation of the crystallites. The azimuthally averaged intensity I(q) depends on the location in the sample (Fig. 3). The curves are very uniform within the region of trabecular bone, with a cross-over around q = 0.1 Å−1 separating the q−4 Porod behavior at high values of q from a q−2 to q−3 behavior at lower q values. Within the region of the growth plate, the cross-over is less pronounced and α equals 5 to 6 (see Fig. 7), which is consistent with the presence the Guinier region originating from elongated particles or a Guinier region for spherical particles. Fig. 7A shows a map of the shape parameter α. In the region of bone, the shape parameter is between 2 and

Fig. 6. (A) AFM image obtained in the unmineralized part of the growth plate. Abundant fibrous structures are identified in this part of the growth plate. The fibers are approximately 100 nm in thickness and seem to aggregate into larger bundles. The bundles of fibers show no preferred orientation, and it appears like they have been cut in some regions (white circle). Furthermore, smaller globular structures are observed in close association to the fibers. (B) AFM image obtained by scanning a subsection of the structures in A at higher resolution. The fibrous structures have a periodicity between 50 and 70 nm. (C) AFM image taken in the central part of the growth plate representing a different topography from the one found in A. The fibers in this subsection are highly aligned and do not form a fibrous mesh. The fibers cross the cutting plane and curl up when cut as indicated by the white circle.

3, which is consistent with mineral plates arranged in a dense network. In the growth plate, values between 5 and 6 are found, while values around 4 are located at the interface between the

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

537

Fig. 7. Results of high-resolution SAXS scans of the region shown in Fig. 5 and 2C in an area crossing the NGP. The pixel size is 50 μm. The background in (A), (B) and (D) maps the X-ray transmission, while the integral SAXS intensity in used for background color in B. (B) shows the predominant orientation of the particles given by the direction of the blue lines. The length of the blue lines is proportional to the PDO of the particles. One pixel width equals a PDO of 83%. (C) displays the thickness parameter T. The orientation of the line corresponds to the smallest particle dimension, while the length of the line gives the relative values of T. The pixel size equals 6 nm. (A) The color equals the magnitude of the shape parameter. Turquoise and green colors in the region of bone correspond to a shape parameter between 2 and 3, respectively. Blue is a shape parameter around 4, which is found in the bone marrow cavities and at the edge of the trabeculae and the growth plate. 5 and 6 are purple and red, respectively, and are found at the edge and center. (D) shows the overall size parameter Rg given by the gray scale. Only values were the SD were less than 30% of the magnitude of Rg are shown.

growth plate and bone and at positions in the upper-left corner and upper-right corner of the scan region. The latter corresponds to areas with high transmission, low SAXS scattering and low PDO (vide infra). We attribute these areas to artefacts most likely due to debris from the sawing of the sample during sample preparation. The predominant orientation and the PDO, extracted from the SAXS patterns of the selected region around the growth plate, are displayed in Fig. 7B. The predominant orientation is depicted in the direction of the large dimensions of the plate. In the region with bone, the predominant orientation of the mineral particles follows the direction of the trabeculae around the bone marrow cavity [23], whereas in the growth plate the particles are predominantly oriented in the longitudinal direction of the growth plate. The PDO appears to be almost constant along the center of the growth plate, while it is more inhomogeneous in the region of bone and at the edge of the growth plate. Note that all areas show a preferred orientation. Fig. 7C illustrates the thickness parameter T by the length of the lines. The orientation of these lines gives the predominant orientation of the shortest dimension of the mineral particles (i.e. rotated 90° relative to the predominant orientation showed in Fig. 7B). T is especially large in the growth plate, whereas small T values are found in the region of bone. In Fig. 7D, we show the overall size parameter Rg. It is only in the growth plate that we can determine the overall size since the dimensions of the mineral particles in the bone are large, requiring data to smaller q-values than what is accessible in our experiment. In the central 100 μm of the growth plate and the adjacent 50 μm region on each side, the overall size Rg are approximately constant (Fig. 7D), with values ranging from 32.1 ± 2.7 nm to 39.8 ± 3.8 nm. If the molecules are cylindrical

as shown by the AFM images, this Rg corresponds to an average cylinder diameter between 90.8 ± 7.6 and 112.6 ± 10.7 nm. Interestingly, we observe a significant change in the magnitude of T from up to 9 nm in the NGP center to approximately 3 nm when moving outward from the center of the growth plate, despite a constant overall size Rg of approximately 35 nm and a shape parameter around 5.5. In this case, the shape parameter shows the presence of a cross-section Guinier behavior and overall size can be interpreted as originating from large particles with a diameter around 70 nm, whereas the T originates from internal surfaces and interfaces in the particles [39,41]. To look for systematic changes in PDO and T with distance from the NGP center, we calculated the average value of these parameters over the dimension parallel to the growth plate. The resulting dependences of the PDO and T on the distance from the growth plate are shown in Figs. 8A and B, respectively. The center of the growth plate is estimated from the transmission and SAXS scans over larger regions and corresponds roughly to the morphological region of the resting and proliferative zones. Relative to the center of the growth plate, the highest average PDO of the particles is found in the region of trabecular bone approximately 500 μm from the center of the growth plate towards the pedicle with an average PDO of 31 ± 13% (mean ± the sample SD), (Fig. 8A). However, the PDO is also relatively high in the center of the growth plate, around 24%. Low PDO, below 10%, is present toward the central part of the vertebral body, 75 to 300 μm from the center of the growth plate with low degrees of orientation between 7 ± 11% and 11 ± 10%. At the pedicular side of the growth plate, the lowest average PDO of the particles is found at the edge of the growth plate 50 to 100 μm from the center. This region that corresponds to the hypertrophic zone has a mean PDO of 12 ± 15%. The growth

538

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

Fig. 8. Average PDO (A) and thickness parameter T (B) are shown as a function of distance from the growth plate center. The solid bars are mean values, while the error bars are the uncertainty around the mean. The position of the growth plate is marked with a gray background. (A) Low PDO is especially found close to the edge of the growth plate and over a wider region toward the vertebral body. (B) The highest values of T are found in the center of the growth zone with the T = 7 nm. T seems to increase from 2.2 to 3.5 nm as a function of distance from the growth plate on the pedicular side. Note that T only corresponds to mineral thickness in the region with bone.

plate is thus asymmetrical with respect to the PDO since a low PDO is found over a wider region on the side facing the midline of the vertebral body. In the inner 100 μm of the growth plate, high average T values of approximately 7 nm are found. In the 50 μm area adjacent to this inner 100 μm of the growth plate, which corresponds to the hypertrophic zone and the edge of the growth plate, the average T drops to 3.6 ± 0.9 nm in the region towards the vertebral body and 3.2 ± 0.6 nm towards the pedicle region. The lowest T values are observed symmetrically in the area between 150 and 200 μm from the center of the growth plate, which is the region of bone adjacent to the growth plate, with values of 2.2 nm. Towards the pedicle, the T values seem to increase from approximately 2.2 to 3.5 nm as a function of distance from the growth plate. This tendency is not seen towards the vertebral body. Discussion The present combination of SEM, AFM and sSAXS provides detailed knowledge about the structure and organization of the nanostructured components of a representative porcine growth plate.

The SEM images reveal that the center of the NGP is composed of elongated fibrous structures closely aligned to each other, while further away from the center, single fibrils are organized in a fibrillar network (Fig. 1). When combining these findings with the AFM data (Fig. 6), we suggest that the apparent variation in the SEM pictures is due to the fibers being organized differently with respect to each other in the different regions of the growth plate. The periodicity and thickness of the fibers are in good agreement with the current knowledge about the structure of collagen type II fibers [45]. Furthermore, the fibrillar network observed in the AFM images is similar to aggregates of collagen type II and proteoglycans in chondrocyte culture that has been observed by SEM in previous studies [46]. Our data on the fibrillar network in the center of the NGP are thus consistent with a picture of bundles of collagen type II, possible cross-linked by collagen type IX and proteoglycans. The high X-ray transmission of the NGP relative to the trabecular bone (Fig. 2A) is consistent with unmineralized tissue and with the low Ca and P concentrations found in our SEM–EDAX analysis. Furthermore, this is in accordance with the structure of the growth plate previously reported [1]. The intensity of the SAXS signal is high both in the growth plate and in the bone region. In bone, this originates from the presence of mineral nanocrystals in an

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

organic matrix, while in the growth plate it is most likely explained by the contrast between organic molecules and air since no Ca and P were found in the center of the growth plate. In the NGP, the presence of a q−5 to q−6 at low q and q−4 behavior at high q in the SAXS data is not straightforwardly explained with a simple two-phase model. One possible explanation is the presence of large structures with internal electron density variations [39]. Another possibility could be that the volume sampled by the X-ray beam includes both regions with large particles and regions with smaller particles similar to those present in the trabecular bone. Since the width of the NGP is approximately 200 μm, there are pixels in the SAXS measurements, which exclusively probe the NGP center and have a vanishing contribution from the trabecular bone. Furthermore, these pixels show the q−5 to q−6 at low q and q−4 behavior. Thus, we ascribe the contribution to the scattering at high q to internal electron density variations within large structures. For pixels at the edge of the NGP, there are contributions to the scattering from both the central part of the NGP and from the trabecular bone, and the resulting intensity variation is thus a weighted average of that from the two regions. The observed decrease in the magnitude of the shape parameter α at the edge of the growth plate therefore probably reflects the weighted amounts of fibrous tissue, calcified cartilage and mineralized bone. This is supported by the SAXS intensity curves obtained at the edge of the growth plate that show a cross-over or “shoulder” emerging between values of high and low q (Fig. 7). The cross-over is likely to be caused by mineral crystals that cause an increase in SAXS intensity. Furthermore, the cross-over coincides with high SAXS intensity at low q. This suggests that both minerals and fibrous tissue are present within the same sample volume. The values of α, the overall size parameter Rg and to a lesser degree T are remarkably uniform along the growth plate direction, and variations are mostly seen in the direction perpendicular to the NGP center. We find Rg to be 32–39 nm in the NGP. In principle, Rg could be related both to the observed periodicity along the collagen fibers, the diameter of the fibers and to other NGP molecules like aggregates of proteoglycans. Interestingly, if Rg is converted to a cross-sectional dimension of a cylinder, diameters of approximately 90–112 nm are found. Since the thickness of fibers found by AFM and SEM is within this length scale, we propose that the Rg of the NGP is related to the diameter of the collagen type II fibers. Despite the constant values of á and Rg, the (effective) thickness parameter T decreased from ∼9 to ∼3 nm from the center to the exterior of the NGP. This decrease in the nanometer length scale structures occurs without a significant change in the large scale structure since α and Rg remain constant. This suggests that the structure of the large scale objects is preserved and exists throughout the different morphological zones, while the substructure in (or between) these objects changes. Since the X-ray scattering intensity is due to any variation in the electron density of the sample, the change in internal structure could also be due to holes within a porous structure and not directly related to a particle size. We propose that one explanation be that the collagen type II bundles tighten up whereby the internal distances between the tropocollagen fibrils are reduced. This would most likely not affect the outer surface of the bundles significantly. These internal

539

changes may serve different purposes such as mechanical matching between areas of different stiffnesses as suggested by Zizak et al. [30]. The design of material and structural gradients for mechanical matching begins to emerge as a general principle in mechanically functional biological materials from several species [47]. The change in characteristic length scale may also be a prerequisite for the later provisional mineralization. However, the detailed molecular basis for the SAXS observations remains to be explored. In trabecular bone, we found that the predominant orientation of mineral plates is along the trabecular surface. Previous studies have revealed that the particle orientations obtained by SAXS correspond to the long axis of the HA crystals [34] and that the particles are oriented along the direction of the collagen fiber of the trabeculae [33]. Our current observations are therefore fully consistent with this established picture [30,33]. However, our current PDO values, 15–30%, in the bone region are lower than what has been reported by Zizak et al. [30]. The PDO is influenced by the alignment of structures on a larger length scale, e.g. fibrous orientations in the trabeculae. The influence of these larger structures on the PDO increases as the measured sample volume increases so the PDO depends on both the X-ray beam diameter and the sample thickness. For a smaller spot size, a higher PDO would therefore be expected and Zizak et al. indeed used a smaller beam size, 20 μm, by use of synchrotron radiation-based instrumentation. At the edge of the NGP, a very low PDO is observed. It may be anticipated that the PDO will be lower in the zone of provisional mineralization than in the region of mature bone since the minerals in the latter are spatially related to aligned fibers of collagen. A disordered arrangement of crystals has been observed by electron microscopy in the zone of provisional calcification [48]. Since the current results are projections, a low PDO could also result from structures aligned perpendicular to the plane of the sample. We determined the mineralites in bone to be plates arranged in a dense network with a mean thickness ranging from ∼2 to ∼3.5 nm. This is in perfect agreement with previous SAXS findings where mineral thicknesses of 2–4 nm have been found in human vertebral bone [30,33]. These mineralite thicknesses are, however, large compared with AFM measurements on HA crystals isolated from young bone, where thickness below 1 nm has been found [31,32]. In favor of the SAXS method is the large number of crystallites probed with the X-ray beam (in the present study ∼1012/point) and that the crystallites sizes are measured in situ with minimum sample preparation. A thickness of 3 nm corresponds to 2–5 unit cells of HA [49]. The smallest thickness parameter Twe observe is ∼2 nm corresponding to only twice the unit cell thickness of HA. It should be noted that in neither AFM nor SAXS probe crystallinity, an observed particle thickness is therefore not necessarily crystallite thickness. The fact that SAXS contrast originates from electron density differences, and hence the presence of mineralites, combined with the in situ nature of the SAXS experiment, makes it most probable that T represents the actual average crystal thickness. It is also observed that T increases with distance from the NGP. It is generally believed that in the vicinity of the growth plate where no bone remodeling has taken place, the distance from the growth plate reflects the age of the bone. Furthermore, it has previously been shown that degree of mineralization increases

540

M. Hauge Bünger et al. / Bone 39 (2006) 530–541

with the age of bone [50]. The observed increase in mineralite thicknesses may therefore be due to an increased mineralization. Direct imaging techniques such as light microscopy, SEM and AFM give very detailed information about the structure of bone on a sub-micrometer to millimeter length scale. However, it can be difficult to interpret images obtained at high magnification in a complex system like bone, and representativeness of the images is always an important issue. In contrast, sSAXS is an integral technique, where the data obtained are mean values over a larger sample volume providing nanostructural information of the various components in bone [51]. The methodological problems in sSAXS are mostly concerned with linking the information obtained with the molecular players in the sample responsible for the scattering contrast. The sections investigated in the present study were washed, dehydrated and freeze dried, all of which might introduce artefacts in the sample. Washing in deionized water could affect minerals in contact with the surface of the sample, while freeze-drying might potentially influence the organic matrix in the growth plate. However, freeze-drying was done to increase the contrast in the SAXS signal in the growth plate. Furthermore, freeze-drying the sample made it possible to do electron microscopy on the same sample as investigated with SAXS. During childhood, growth plates are in general mechanically weak points and common sites of fracture. The gradients in apparent particle thickness in the NGP and in the crystal thickness of the adjacent bone may play an important role for determining the strength of the bone–NGP interface. Identifying structural changes in the NGP at the nanometer length scale might provide further insights in diseases like idiopathic scoliosis with development of spinal curves [3]. Acknowledgments We thank Dr. Li Haisheng, Dr. Zou Xuenong and laboratory technicians Anette Milton and Anette Baatrup for the help with sample preparation and for providing the CT shown in Fig. 1A. We furthermore acknowledge the financial support from the Danish Research Council through the “Large Interdisciplinary Research Group—Nanoscience and Biocompatibility” (2052-010006). H.B. is a Steno Research Assistant Professor funded by the Danish Natural Science Research Council and the Danish Technical Research Council (contract 21-03-0433) and gratefully acknowledges this support. References [1] Kronenberg HM. Developmental regulation of the growth plate. Nature 2003;423:332–6. [2] Abad V, Meyers JL, Weise M, Gafni RI, Barnes KM, Nilsson O, et al. The role of the resting zone in growth plate chondrogenesis. Endocrinology 2002;143:1851–7. [3] Maat GJ, Matricali B, van Persijn van Meerten EL. Postnatal development and structure of the neurocentral junction. Its relevance for spinal surgery. Spine 1996;21:661–6. [4] Rajwani T, Bhargava R, Lambert R, Moreau M, Mahood J, Raso VJ, et al. Development of the neurocentral junction as seen on magnetic resonance images. Stud Health Technol Inform 2002;91:229–34.

[5] Rajwani T, Bhargava R, Moreau M, Mahood J, Raso VJ, Jiang H, et al. MRI characteristics of the neurocentral synchondrosis. Pediatr Radiol 2002;32:811–6. [6] Vital JM, Beguiristain JL, Algara C, Villas C, Lavignolle B, Grenier N, et al. The neurocentral vertebral cartilage: anatomy, physiology and physiopathology. Surg Radiol Anat 1989;11:323–8. [7] Eyre DR, Muir H. Quantitative analysis of types I and II collagens in human intervertebral discs at various ages. Biochim Biophys Acta 1977;492:29–42. [8] Wardale RJ, Duance VC. Quantification and immunolocalisation of porcine articular and growth plate cartilage collagens. J Cell Sci 1993;105:975–84. [9] Boskey AL. Mineral–matrix interactions in bone and cartilage. Clin Orthop 1992:244–74. [10] Orth MW. The regulation of growth plate cartilage turnover. J Anim Sci 1999; 77:183–9. [11] Mendler M, Eich-Bender SG, Vaughan L, Winterhalter KH, Bruckner P. Cartilage contains mixed fibrils of collagen types II, IX, and XI. J Cell Biol 1989;108:191–7. [12] van der Rest M, Mayne R. Type IX collagen proteoglycan from cartilage is covalently cross-linked to type II collagen. J Biol Chem 1988;263:1615–8. [13] Smith Jr GN, Brandt KD. Hypothesis: can type IX collagen glue together intersecting type II fibers in articular cartilage matrix? A proposed mechanism. J Rheumatol 1992;19:14–7. [14] Eyre DR, Apon S, Wu JJ, Ericsson LH, Walsh KA. Collagen type IX: evidence for covalent linkages to type II collagen in cartilage. FEBS Lett 1987;220:337–41. [15] Shen G. The role of type X collagen in facilitating and regulating endochondral ossification of articular cartilage. Orthod Craniofac Res 2005:11–7. [16] Cameron DA, Robinson RA. Electron microscopy of cartilage and bone matrix at the distal epiphyseal line of the femur in the newborn infant. J Biophys Biochem Cytol 1956;2:253–60. [17] Young MF. Bone matrix proteins: their function, regulation, and relationship to osteoporosis. Osteoporos Int 2003;14:S35–42. [18] Termine JD, Belcourt AB, Conn KM, Kleinman HK. Mineral and collagenbinding proteins of fetal calf bone. J Biol Chem 1981;256:10403–8. [19] Termine JD, Kleinman HK, Whitson SW, Conn KM, McGarvey ML, Martin GR. Osteonectin, a bone-specific protein linking mineral to collagen. Cell 1981;26:99–105. [20] Ducy P, Desbois C, Boyce B, Pinero G, Story B, Dunstan C, et al. Increased bone formation in osteocalcin-deficient mice. Nature 1996;382: 448–52. [21] Hoang QQ, Sicheri F, Howard AJ, Yang DS. Bone recognition mechanism of porcine osteocalcin from crystal structure. Nature 2003;425:977–80. [22] Landis WJ, Song MJ. Early mineral deposition in calcifying tendon characterized by high voltage electron microscopy and three-dimensional graphic imaging. J Struct Biol 1991;107:116–27. [23] Landis WJ, Song MJ, Leith A, McEwen L, McEwen BF. Mineral and organic matrix interaction in normally calcifying tendon visualized in three dimensions by high-voltage electron microscopic tomography and graphic image reconstruction. J Struct Biol 1993;110:39–54. [24] Landis WJ. An overview of vertebrate mineralization with emphasis on collagen–mineral interaction. Gravit Space Biol Bull 1999;12:15–26. [25] Petruska JA, Hodge AJ. A subunit model for the tropocollagen macromolecule. Proc Natl Acad Sci U S A 1964;51:871–6. [26] Su X, Sun K, Cui FZ, Landis WJ. Organization of apatite crystals in human woven bone. Bone 2003;32:150–62. [27] Weiner S, Traub W, Wagner HD. Lamellar bone: structure–function relations. J Struct Biol 1999;126:241–55. [28] Siperko LM, Landis WJ. Aspects of mineral structure in normally calcifying avian tendon. J Struct Biol 2001;135:313–20. [29] Potter K, Leapman RD, Basser PJ, Landis WJ. Cartilage calcification studied by proton nuclear magnetic resonance microscopy. J Bone Miner Res 2002;17: 652–60. [30] Zizak I, Roschger P, Paris O, Misof BM, Berzlanovich A, Bernstorff S, et al. Characteristics of mineral particles in the human bone/cartilage interface. J Struct Biol 2003;141:208–17. [31] Eppell SJ, Tong W, Katz JL, Kuhn L, Glimcher MJ. Shape and size of isolated bone mineralites measured using atomic force microscopy. J Orthop Res 2001;19:1027–34.

M. Hauge Bünger et al. / Bone 39 (2006) 530–541 [32] Tong W, Glimcher MJ, Katz JL, Kuhn L, Eppell SJ. Size and shape of mineralites in young bovine bone measured by atomic force microscopy. Calcif Tissue Int 2003;72:592–8. [33] Rinnerthaler S, Roschger P, Jakob HF, Nader A, Klaushofer K, Fratzl P. Scanning small angle X-ray scattering analysis of human bone sections. Calcif Tissue Int 1999;64:422–9. [34] Fratzl P, Schreiber S, Klaushofer K. Bone mineralization as studied by small-angle x-ray scattering. Connect Tissue Res 1996;34:247–54. [35] Fratzl P, Fratzl-Zelman N, Klaushofer K, Vogl G, Koller K. Nucleation and growth of mineral crystals in bone studied by small-angle X-ray scattering. Calcif Tissue Int 1991;48:407–13. [36] Fratzl P, Groschner M, Vogl G, Plenk Jr H, Eschberger J, Fratzl-Zelman N, et al. Mineral crystals in calcified tissues: a comparative study by SAXS. J Bone Miner Res 1992;7:329–34. [37] Li H, Zou X, Xue Q, Egund N, Lind M, Bunger C. Effects of auto- genous bone graft impaction and tricalcium phosphate on anterior interbody fusion in the porcine lumbar spine. Acta Orthop Scand 2004;75:456–63. [38] Pedersen JS. A flux- and background-optimized version of the NanoSTAR small-angle X-ray scattering camera for solution scattering. J Appl Crystallogr 2004;37:369–80. [39] Lindner P, Zemb T, editors. Neutrons, X-rays and light: scattering methods applied to soft condensed matter. Elsevier Science; 2002. [40] Poon WCK, Haw MD. Mesoscopic structure formation in colloidal aggre gation and gelation. Adv Colloid Interface Sci 1997;73:71–126. [41] Ruland W. Small-angle scattering of 2-phase systems—Determination and significance of systematic deviations from Porods law. J Appl Crystallogr 1971; 4:70.

541

[42] Fratzl P. Statistical-model of the habit and arrangement of mineral crystals in the collagen of bone. J Stat Phys 1994;77:125–43. [43] Fantner GE, Birkedal H, Kindt JH, Hassenkam T, Weaver JC, Cutroni JA, et al. Influence of the degradation of the organic matrix on the microscopic fracture behavior of trabecular bone. Bone 2004;35:1013–22. [44] Robey P, Boskey A. The biochemistry of bone. In: Marcus R, Feldman D, editors. Osteoporosis. California, San Diego: Academic Press; 1996. [45] Beck K, Brodsky B. Supercoiled protein motifs: the collagen triple-helix and the alpha-helical coiled coil. J Struct Biol 1998;122:17–29. [46] Lewis JL, Johnson SL, Oegema Jr TR. Interfibrillar collagen bonding exists in matrix produced by chondrocytes in culture: evidence by electron microscopy. Tissue Eng 2002;8:989–95. [47] Waite JH, Lichtenegger HC, Stucky GD, Hansma P. Exploring molecular and mechanical gradients in structural bioscaffolds. Biochemistry 2004; 43:7653–62. [48] Bonucci E. Crystal ghosts and biological mineralization: fancy spectres in an old castle, or neglected structures worthy of belief? J Bone Miner Metab 2002;20:249–65. [49] Mann S. Biomineralization. Principles and concepts in bioinorganic materials chemistry. Oxford: Oxford Univ. Press; 2002. [50] Meunier PJ, Boivin G. Bone mineral density reflects bone mass but also the degree of mineralization of bone: therapeutic implications. Bone 1997;21: 373–7. [51] Gupta HS, Roschger P, Zizak I, Fratzl-Zelman N, Nader A, Klaushofer K, et al. Mineralized microstructure of calcified avian tendons: a scanning small angle X-ray scattering study. Calcif Tissue Int 2003;72: 567–76.