Microarchitectural and Physical Changes During Fetal Growth in Human Vertebral Bone

Authors


  • The authors have no conflict of interest.

Abstract

The ossification process in human vertebra during the early stage of its formation was studied by X-ray diffraction (XRD) and X-ray microtomography (μCT) at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. Twenty-two samples taken from vertebral ossification centers of human fetal bone (gestational age ranging between 16 and 26 weeks) were investigated. The analysis of three-dimensional images at high spatial resolution (∼10 and ∼2 μm) allows a detailed quantitative description of bone microarchitecture. A denser trabecular network was found in fetal bone compared with that of adult bone. The images evidenced a global isotropic structure clearly composed of two regions: a central region (trabecular bone) and a peripheral region (immature bone). XRD experiments evidenced hydroxyapatite-like crystalline structure in the mineral phase at any fetal age after 16 weeks. Interestingly, the analysis of XRD patterns highlighted the evolution of crystalline structure of mineralized bone as a function of age involving the growth of the hydroxyapatite crystallites.

INTRODUCTION

BONE IS A LIVING tissue perpetually undergoing a metabolic process known as bone remodeling or bone turnover. The early developing stage of mineralized bone tissue in fetus (ossification) is a complex process in which an initial cartilage template is converted into bone.(1) Although such stages of skeleton development are, without a doubt, essential in determining the quality of the adult bone, several questions are still open about the processes related to the evolution and growth of bone tissue during ossification. Various techniques such as X-ray radiography,(2,3) histology,(4,5) and sonography,(6–8) have been used to probe the morphological development of bone during its formation. DXA(9,10) and quantitative computed tomography (QCT)(11) have been used to study the mineral density in developing vertebral bodies, showing an increase of bone mineral density (BMD) with fetal growth. Most of these techniques provide macroscopic information, but none of them are able to describe accurately the three-dimensional (3D) microscopic organization of trabecular components of bone or the mineral nature of calcified bone. Complementary techniques like X-ray diffraction (XRD),(12) electron or infrared microscopy,(13,14) and infrared spectroscopy(15) have been used to characterize the crystalline structure and the chemical composition of bone tissue. However, few studies have been performed on human samples,(16–18) and very rarely have they been performed on human vertebral fetal bone.(19)

The aim of this study was to give a deeper insight into the evolution process and microarchitectural and physical changes occurring in the early stages of skeleton formation in the human bone tissue. To this end we investigated the region of ossification centers in human fetal vertebra combining two techniques: X-ray microtomography and X-ray diffraction (XRD). These allowed the collection of morphological (micrometer scale) and crystallographic (nanometer and atomic scale) information.

Because of the scale of the bone structure, synchrotron radiation computed microtomography (SR-μCT) is extremely attractive for the 3D analysis of bone microarchitecture. Like conventional computed tomography, SR-μCT allows nondestructive 3D imaging of the internal structure of a sample.(20–22) The use of a third-generation X-ray synchrotron radiation source yields additional advantages: the high photon flux permits images up to very high spatial resolutions (<1 μm) and exceptional signal-to-noise ratios, while keeping reasonable acquisition times. Monochromatic X-ray beams are used, thus avoiding beam-hardening artifacts that are observed in images collected with conventional X-ray sources. As a consequence, accurate maps of the attenuation coefficient at the selected energy can be reconstructed. In addition, the possibility to finely tune the beam energy over a wide range permits the optimal choice of energy to improve the quality of images.(23)

The analysis of XRD patterns allows the description of the mineral structure of bone at an atomic and nanometer scale. Like microtomography, XRD also takes advantages of the high brilliance and the energy tunability of synchrotron radiation for collecting high-quality diffraction patterns. The use of a monochromatic X-ray beam gives simple line shape profile suitable for accurate line shape analysis and structural refinement. Moreover, the intense X-ray beam coupled to angle dispersed setups (bi-dimensional [2D] detector geometry) permits the collection of diffractograms with ample statistics in very limited time (from a few seconds to a few minutes).(24) Rapid data collection time is a substantial benefit compared with the time required with standard diffractometers based on single-detector scanning (several hours for each pattern). In addition, the reduction of collection time is advantageous for biological samples because it prevents an eventual degradation of the organic part during measurements. In this work, we used the advantages of SR XRD to collect high-quality diffractograms from bone samples to highlight the effect of age on the crystalline structure of mineralized bone.

MATERIALS AND METHODS

Experimental subjects

Twenty-two lumbar vertebrae taken from human fetuses were provided by the department of Pathology of “Edouard Herriot” Hospital, Lyon (France). They were collected from preterm human stillborns with gestational age ranging from 16 to 26 weeks. Fetal age was determined from several data: last menstruation date, first echography, morphologic examination by the anatomo-pathologist, in particular using body weight, femur, and foot length. There was nothing in the history of these fetuses to suggest any problem that might affect bone mineralization. The vertebrae were stored under pure ethyl alcohol to prevent degradation. For comparison, a sample taken from a young adult human bone (33 years old) was also examined.

Figure 1 shows a microscopic picture of a section cut perpendicularly to the vertebra body. It shows an inner part, the ossification center (region B), consisting of calcified tissue, embedded in a surrounding medium composed of cartilaginous amorphous tissue (region A).(9) Samples extracted from the ossification center of the vertebrae were respectively prepared for SR-μCT imaging and XRD measurements.

Figure FIG. 1..

Microscopic image of a fragment of vertebra section (23 weeks old). The (A) cartilaginous and (B) mineralized regions have been probed by XRD, and the diffractograms are reported in the Fig. 3.

X-ray microtomography

Data acquisition:

Twelve cubic bone samples of 2 and 7 mm3 were embedded in methyl-methacrylate and imaged on the 3D SR-μCT setup developed on the topography and high-resolution diffraction beamline (ID19) at the European Synchrotron Radiation Facility (ESRF), in Grenoble, France. A detailed description of the device is reported elsewhere.(22) The principle of the system is based on a 3D parallel tomographic acquisition setup: for different angles of view, 2D projections of the sample are recorded and processed to get 3D images. The transmitted X-ray beam passing through the sample is recorded using a 2D detector based on a 1024 × 1024 FRELON (Fast REadout LOw Noise) CCD camera developed by the ESRF Detector group (pixel size 19 μm, dynamic range 14 bits).(25) The whole detection system includes a fluorescent screen converting X-rays to light, light optics magnifying the screen image, and the FRELON camera. Different choices of optics were designed to achieve different pixel sizes on the recorded image, which currently vary from 10.13 to 0.4 μm. Because the available field of view is a function of the spatial resolution, the larger samples (n = 6, 7 mm3) were imaged using a pixel size of 10.13 μm and the smaller (n = 6, 2 mm3) with a pixel size of 1.8 μm. The monochromatic X-ray energy was set to 20 KeV. For each sample, 900 projection images (1 s of integration time each) were acquired over a range of 180° with an angular step of 0.2°. Images without sample and dark current images were also collected to normalize the data for the decay of the synchrotron radiation during the scan and to correct flat field distortions.

Image processing:

The acquisitions were processed using a 3D-filtered backprojection algorithm to get the 3D reconstructed images.(26) The reconstruction voxels (volume element) were cubic with an edge length of 10.13 or 1.8 μm according to the chosen spatial resolution.

Image analysis tools were used to extract structural parameters quantifying bone architecture. Six regions of interest (ROIs) were selected in each of the 3D images at 10.13 and 1.8 μm. Because of the high contrast of the images, the separation of bone signal from that of the background was obtained by a simple thresholding. The following morphometric parameters derived from those used in histomorphometry were computed: bone volume to total volume (BV/TV) ratio, trabecular thickness (TbTh), and trabecular separation (TbSp). To avoid geometrical assumption on the bone model (plate or rod), we used direct parameters instead of conventional Parfitt parameters.(27) The trabecular thickness and separation were directly estimated from the 3D images according to the method proposed by Hildebrand and Rüegsegger.(28) The algorithm does not require any special constraints regarding the size and geometry of bone volumes.

XRD

Data acquisition:

Samples for XRD measurements were drawn in the mineralized part of ossification centers (see Fig. 1). The bone was crumbled by hand in agate mortar to obtain fine powder. According to Very and Baud,(29) this procedure should not affect the crystalline status of mineralized bone and allows an homogeneous dispersion of fine particles well suited to good-quality XRD patterns. For each sample, 10-20 mg of powder was enclosed in cylindrical borosilicate capillaries (diameter = 0.5 mm) that were sealed under N2 to prevent degradation of organic parts. This procedure took a few minutes. Each capillary was placed in ultrasonic bath for 10 minutes to compact the powder.

The experiments were carried out on the general Italian beam line for diffraction and absorption (GILDA) at the ESRF.(30) The monochromatic X-ray energy (λ = 1.0331 Å with ΔE/E = 10−4) was selected using a Si (311) double crystal monochromator. XRD patterns were collected using the angular disperse setup employing a 2D imaging plate (IP; 200 × 400 mm2) camera.(31) The beam size on the sample was 1.5 (horizontal) × 1 (vertical) mm2. Samples contained in capillaries were mounted horizontally. In such geometry the capillary diameter and sample to IP distance mainly determine the instrumental line breadth.(24) The IP was positioned perpendicularly to the incoming beam, at about 257 mm from the sample. The images were digitized using a BAS2500 reader with 100 × 100 mm2 pixel size and a dynamic range of 16 bits/pixel. Collection time for each sample was fixed at 15 minutes, and the capillaries were kept rotating during the acquisition of the XRD patterns to improve the grain statistics. The instrumental setup parameters (sample to IP distance, X-ray wavelength, instrumental peak broadening) were calibrated by collecting and refining diffraction patterns of a LaB6 reference sample (NIST-SRM 660).

Data analysis:

Figure 2 represents a typical digitized image of a 2D diffraction pattern collected on a fetal bone sample (23 weeks old). The asymmetrical geometry allows a wider angular range. 2D images were processed using the FIT2D package(32) to obtain powder diffraction intensities as a function of scattering angle, 2θ, corrected for geometrical and polarization effects. Diffractograms equivalent to a 2θ scan were obtained by integrating the intensity of circular arcs in an angular region (width ∼30°) of the 2D image (see Fig. 2).

Figure FIG. 2..

Typical 2D XRD pattern collected on a fetal bone sample. The diffraction rings corresponding to the (100) and (002) reflections of OHAp are indicated by arrows. The diffractograms are then obtained by integrating the intensity on circular arcs in an angular region (width ∼30°) as illustrated on the image.

The shape of diffraction peaks reflects the crystallization state of a substance: larger and broader shapes of the peaks indicates poorer crystallization state with smaller crystallites (i.e., the coherently diffracting regions in the sample) and large number of defects.(33) Sharper intense diffraction peaks indicate well-crystallized substances having larger, defect-free, crystallites. For the sake of simplicity, we assumed that the peak breadth was due only to the crystallite size effect. Within this approximation, the Sherrer formula relates the average crystallite size perpendicular to the (hkl) planes, Dhkl, to the full line breadth at half maximum, corrected for instrumental broadening, Whkl:

equation image

where λ is the X-ray wavelength, θhkl is the diffraction angle, and K is a constant depending on the crystal shape and shape distribution. K ranges between 0.8 and 1.39.

In this study, the absolute value of crystal size is not crucial; what is more interesting is the evolution of the crystal size with aging, so we used the approximated value K = 0.94.(33)

According to several published works,(34,35) we estimated the crystallite size along the c axis, analyzing the (002) Bragg peak (2θ ≈ 17.3°) because it is intense and well isolated. The peak breadth (W002) was obtained by fitting the (002) reflection with a Pseudo-Voigt function.

Another quantitative structural information that can be derived from the analysis of XRD patterns is the c/a ratio. In the hexagonal lattice, the angular position of a diffraction peak is related to the cell edges (a, c) and to the reflection indexes (h, k, l) by the equation:

equation image

For selected pairs of diffraction peaks, it is possible to get the c/a ratio as a function of their angular positions. For example, for (002) and (222) reflections, the c/a ratio is:

equation image

and for (002) and (310) reflections, it is:

equation image

The values of the c/a ratios reported in Fig. 9 have been calculated by averaging, for each sample, the c/a ratios obtained from several pairs of diffraction peaks, chosen as the most resolved and/or least affected by superimposition with other reflections, namely, (002) (310), (213), and (004).

Figure FIG. 9..

Average values of c/a ratios as a function of age calculated from the XRD patterns (see text). The linear fit of the data (R = 0.931, p = 0.0023) is also reported, pointing out the growth as a function of age.

RESULTS

Figure 3 reports diffraction patterns collected from the external and internal regions of the vertebra slice (23 weeks old; regions A and B, respectively, in Fig. 1). The diffraction pattern of the external region (Fig. 1A) presents broad and smooth halos that are typical of amorphous structures. On the contrary, the diffractogram collected from the inner region (Fig. 1B) clearly presents sharper reflexes (Bragg peaks), showing the presence of crystalline structures. Thus, the vertebra structure appears formed of a core of mineralized bone surrounded by cartilaginous noncalcified material. This inner part is the main subject of this study because it has high contrast for SR-μCT and defined crystalline structure suitable for XRD analysis.

Figure FIG. 3..

Diffractograms collected from (A) external and (B) inner regions of the vertebra section shown in Fig. 1. Diffraction intensity (I) is shown in arbitrary units (a.u.); the two curves are vertically shifted for clarity to avoid superimposition. The arrow points out a reflection peak around 7° in both diffractograms. In the diffractograms of the mineral region (B), this peak clearly corresponds to the (100) reflection of OHAp.

X-ray microtomography

Figure 4 presents a 3D display of a fetal vertebra sample (24 weeks old) and a 2D slice extracted from the reconstructed volume (voxel size: 10.13 μm). The images depict a very dense trabecular network that looks relatively isotropic when observed in 3D space. Two different structures of bone are clearly contrasted, corresponding respectively to an internal (trabecular) and a surrounding region. From a qualitative point of view, the external surface of the vertebra appears very thin and tight while the internal region is made of thicker components. The blurring visible in this periphery indicates that bone structures are not completely resolved in this area at the 10-μm scale. The structural bone organization is better appreciated in the images recorded at higher spatial resolution (1.8 μm). In addition, differences in gray levels can be observed within trabecular bone (Fig. 5, left side). Because of the monochromaticity of the X-ray beam, the resulting 3D images are maps of the linear attenuation coefficient within the sample.(36) Thus, the differences in gray levels correspond to differences in absorption related to the amount of mineral within bone tissue. Black contrasts correspond to background and resin absorption, while brighter contrast corresponds to mineralized bone. It is particularly interesting to notice the presence of very bright contrast zones corresponding to mineralization fronts.

Figure FIG. 4..

Illustration of the microarchitecture in a fetal vertebra (24 weeks old) imaged at 10.13 μm: 3D display (left side) and 2D slice (right side). The scale is the same on the two images. The white hashed circle drawn on the left side image roughly delimits the inner and the peripheral regions.

Figure FIG. 5..

Illustration of the microarchitecture in a fetal vertebra (24 weeks old) imaged at 1.8 μm: 2D slices through inner bone region (left side) and peripheral regions (right side; same scale on the two images). The inner region (left side) presents thicker structures compared with the peripheral region.

To quantify these qualitative observations, morphological parameters were computed on 3D ROIs extracted from each reconstructed volume in the internal and external bone region. The 10-μm scale (voxel size = 10.13 μm) was used to quantify trabecular bone architecture in the internal bone region. At the same scale, bone structures in the external regions would not be correctly quantified because of partial volume effects. Thus, ROIs were accurately selected in the images at the micrometer scale (voxel size = 1.8 μm) to quantify the peripheral bone regions.

Architectural parameters were calculated in all 3D ROIs. The availability of images in 3D space allows to avoid assumptions on the geometry of bone structures and to estimate model-independent parameters. This is a relevant advantage of the technique compared with histomorphometry, because the observation of μCT images clearly shows that the typical parallel plate model conventionally used is not appropriate at this stage of formation. The results of the calculations (mean values, SD, and range of architecture parameters) for the six 3D ROIs both in the external and internal bone regions are presented in Table 1. The mean value of partial bone volume, BV/TV, calculated over all samples at the 10-μm scale is 40.07 ± 10.08% and is in agreement with visual observations. A significant linear increase of BV/TV with aging is also found (BV/TV = 0.032 × weeks − 0.232; p < 0.05; Fig. 6). Concerning the trabecular thickness, the mean value in the inner bone region is about 102.30 ± 11.25 μm, and no significant variation with aging is found. In the external region, the thickness of the microstructures observed in the 3D images is found to be around 9.64 ± 0.17 μm.

Figure FIG. 6..

Evolution of partial bone volume (BV/TV) measured from trabecular bone images of fetal vertebra at 10.13 μm as a function of gestational age. The line represent a linear fit to the data showing a significant increase (R = 0.964, p = 0.0019).

Table Table 1.. Minimum (Min), Maximum (Max), and Mean Values (Mean) of Architectural Parameters ± SD Calculated Directly in Internal and External Bone ROIs
original image

XRD

Figure 7 illustrates the diffractograms of synthetic hydroxyapatite (OHAp), Ca10(PO4)6(OH)2 (Riedel-de Haën AG, Seelze, Germany), and of two bone samples from an adult and a fetal vertebra (16 weeks old). Figure 7 also indicates the position of the hydroxyapatite Bragg peaks calculated according to Posner and Diorio(37) (space group P63/m, 176, a = b = 9.415 Å, c = 6.879 Å). Several diffraction peaks may be identified in diffractograms of both adult and fetal bone, showing the presence of a crystalline phase in all samples. Moreover, human bone and synthetic OHAp diffractograms present large similarities. This indicates that the main component of mineralized human bone in adults,(38) as well as in fetuses, is structurally very similar to OHAp, and that crystallized OHAp mineral is already present in the younger fetal bone sample (16 weeks old). At the same time, the diffractograms also exhibit the presence of structural differences between the samples: diffraction peaks are sharper in synthetic hydroxyapatite with respect to human bone samples. Fetal bone samples present broadest peaks with an intense background because of the presence of amorphous phases. Our data do not allow us to characterize and/or quantify these amorphous phases because, in the spectra, there is also a large diffuse scattering contribution coming from the glass capillary sample holder. However, we can affirm that amorphous phases are present because spectra collected on the region of the calcified bone slice (Fig. 1), which was not measured in the glass capillary, shows the presence of amorphous phases (Fig. 3).

Figure FIG. 7..

Diffraction patterns obtained from synthetic (a) OHAp, (b) adult bone, and (c) fetus (16 weeks old). The intensity (I) of the different curves is expressed in arbitrary units (a.u.) and are shifted to avoid superimposition. The two inserts highlight regions of the diffractograms to put in evidence the differences. The ticks below point out the position reflections calculated for standard OHAp structure (space group P63/m, 176).

In the diffraction patterns reported in Fig. 3, a weak peak is observed at about 2θ° ≈ 7°. It is noticeable that this peak has a similar intensity in the patterns collected both from region A (cartilage) and B (calcified tissue) of Fig. 1, even though the two diffractograms are very different in the high-angle domain. In region B, this peak corresponds to the (100) reflection of OHAp (see a diffractrogram of OHAP in Fig. 7, curve a). In region A, this peak cannot be interpreted in the same way because the other more intense reflections of OHAp are not observed in the diffractogram.

In Fig. 8, the values of OHAp crystal size along the c axis (D002) are reported. The D002 value in fetal samples ranges between about 15 and 25 nm. In adult bone, we found D002 of about 40 nm, whereas in synthetic OHAp, the grain size is over 60 nm. Thus, bone crystallites are, on average, smaller in fetuses than in adults. Most interesting is that the size of these crystallites sensitively grows with aging; in fact, data are well described by a significant linear relationship: D002 = 0.75 + 1.489 × 10−4 weeks.

Figure FIG. 8..

OHAp crystallite size as a function of gestational age as calculated from the diffraction line breadth (see text). The line represents a linear fit to the data (R = 0.931, p = 0.0023), definitively probing the growth of OHAp crystallite size as a function of fetal age.

The ratio c/a reflects deformations of the OHAp unit cell and is related to the chemical composition of OHAp.(35) The ratio c/a = 0.731 obtained for synthetic hydroxyapatite matches well with that reported in literature.(37) In the adult bone samples, we found a larger ratio, c/a = 0.7325, in agreement with the value 0.7326 reported in Smith and Smith.(17) All the fetal bone samples, on the contrary, have smaller c/a ratios, ranging between 0.727 and 0.730. The linear regression on the c/a values indicates significant growth of c/a ratio with aging: c/a = 0.725 + 1.280 × 10−4 weeks (Fig. 9), with a slope significantly larger than the slope reported by Smith and Smith.(17)

DISCUSSION

X-ray microtomography

The ossification region of fetal vertebra has been studied from a morphological and crystallographic point of view combining SR μCT and XRD techniques.

3D images at a very high spatial resolution (10 and 2 μm) are reported here for the first time. These images show a quite different microarchitecture in fetus bone vertebra compared with that previously observed in adults.(22) The qualitative observation of images shows that the trabecular microarchitecture does not present the privileged orientations as observed in reconstructed images from adult human vertebra samples.(39) A hypothesis to explain the isotropy observed on the trabecular organization is the absence of biomechanical constraints on fetuses. The range of BV/TV values varying from 30% to 54% in fetuses indicates that the network is 2.5 or 3 times denser than in the young adult (BV/TV around 15%). In this stage of bone formation, a very rapid increase of BV/TV is also noticeable, showing a very short time scale of changes in fetus. This variation of BV/TV toward higher values is opposite to those reported in elderly adults with aging.(40)

Surprisingly, the mean value of trabecular thickness (∼102 μm) in the inner region is close to the trabecular thickness in adult vertebra (∼100 μm) reported in previous works.(39,41) This finding suggests that trabeculae reach their mature size at an early stage of their development. The mean trabecular separation (∼224 μm) is found to be much smaller than in the adult (∼500 μm), and it explains the higher density of the fetal bone network. Changes of trabecular thickness with aging in fetuses were negligible. The same effect was observed in adults.(22) The stability of trabecular thickness and increase of BV/TV in fetal bone are in agreement with the decreasing trend observed for trabecular spacing.

The mean trabecular thickness and separation values in the external region are about 10 times lower than in the inner region, suggesting that the very early bone structure is being formed from the precursor. This means that the peripheral shell may correspond to early deposited mineral. Thus, the 3D μCT images suggest that the ossification process in fetal vertebra is an expansion from a central nucleus (central region) by the formation of shells of new mineralized structures (external region). The very thin and tight components structure of bone in the surrounding region are quantified here for the first time from images recorded with a pixel size fixed at 1.8 μm. This feature stresses the importance of improving the spatial resolution of images to achieve quantitative parameters. Notice that although a spatial resolution down to 0.4 μm is possible, the 1.8-μm scale was chosen as a good compromise for a sufficient field of view (FOV) and to avoid partial volume effects. In fact, the increase of spatial resolution necessarily yields a decrease of the available FOV in the imaging system because of the fixed number of pixels in the CCD camera.

XRD

XRD results establish that the principal constituent of mineralized bone in the ossification centers of fetal vertebra is an OHAp-like crystalline phase similar to that of vertebral adult bone. However, a large contribution from amorphous phases is observed in the calcified fetal bone.

The weak peak observed at 2θ° ≈ 7° in region A of Fig. 3 could possibly be interpreted as the (200) peak of the ortocalcium phosphate (OCP) phase [Ca8H2(PO4)6 · 5 H2O]. In fact, the OCP (200) reflection is very close to the (100) peak of OHAp. If so, our finding may signal the presence of a consistent amount of OCP-like phases in the cartilage that can be considered as a precursor of the OHAp phase in the calcified bone tissue. The nature of the precursor of calcified bone tissue is still an open and discussed question(38): several calcium phosphate phases have been postulated, among them OCP-like and amorphous calcium phosphate (ACP). These phases are difficult to probe because of several factors: for example, the instability of OCP phase that rapidly hydrolyzes producing OHAp according to the reaction Ca8H2(PO4)6 · 5 H2O + 2Ca+ = Ca10(PO4)6 (OH2) + 4H+. Moreover the occurrence of interlayered mixtures of OCP and OHAp would give rise to superimposed and mixed diffraction peaks making the analysis of XRD patterns difficult. At this stage, our findings are qualitative and cannot solve the question about the nature of the mineral bone precursors. However, the reduced acquisition time and the good data statistics allowed by the IP setup is very promising to get deeper insight into this problem.

The analysis of XRD patterns suggests that bone development during the ossification process corresponds to structural changes of OHAp crystallites that become larger or may be less defective. In particular, the observed growth, during stages of rapid bone development, is in agreement with data reported in rat bone maturation(42) but has never been quantified on human fetal bone. This observation with aging in fetuses is different to that in children and adults where no changes in crystal size were observed,(34,43) although this had been a controversial issue.(16) The evolution occurs because the bone mineralization and growth in fetus proceed through the formation of new small crystallites from the precursor material.

The increase of c/a ratio with gestational age was also observed. Several parameters can affect the c/a ratio in OHAp. In particular, changes in composition have considerable effects.(42,44) These changes may be related to the ability of bone apatite phase to absorb and release ions according to its physico-chemical activity in human. For instance, the inclusion of CO3 groups, substituting the larger PO4 units, provokes the expansion of OHAp cell along the c axis and its compression in the ab plane.(38) At the same time, the increase in CO3 groups was shown to be associated to an increase of Ca/P during bone maturation in chick embryos.(45) The increase of c/a values found in human fetuses is probably associated to such an increase of Ca/P ratio throughout gestation. This interpretation comes in agreement with chemical analyses of calcification of lumbar vertebra during human fetal development, showing Ca/P ratios lower than in OHAp (Ca/p = 1.667) and a positive increase with aging.(19)

In conclusion, the early phases of the ossification process in normal human fetal vertebrae have been investigated by combining μCT and XRD techniques. The results accurately describing the microarchitecture and physical nature of bone apatite bring a close insight into the mechanisms of ossification. The availability of 3D information at adapted spatial resolution that a standard μCT device could not reach was a keystone to get quantitative parameters. The analysis of 3D μCT images showed that the ossification occurs from the central nucleus: new mineralized structures (central region) are formed subsequently expanding toward the extremities (external shell), by the conversion of a cartilage template into mineralized tissue. SR XRD provided diffraction patterns within the minute range and with excellent statistics well suited for extracting accurate characteristics of crystallites. The results showed that the mineralization process produces structural and compositional changes in the OHAp-like phase. Relevant parameters characterizing the evolution of mineral bone with aging from a morphological and crystallographic point of view were identified. Bone density, microarchitectural parameters (trabecular thickness and trabecular spacing), size, and composition of OHAp crystallites have all been quantified. A unique finding is the increase of crystal size with increasing fetal maturity.

Acknowledgements

We thank the ESRF for financial support, the ID19 group team and GILDA group team for help during experiments, and Dr G Boivin for reading the manuscript. GILDA is financed by the Italian Institutions CNR, INFM, and INFN.

Ancillary