Infrared Analysis of the Mineral and Matrix in Bones of Osteonectin-Null Mice and Their Wildtype Controls


  • The authors have no conflict of interest


Osteonectin function in bone was investigated by infrared analysis of bones from osteonectin-null (KO) and wildtype mice (four each at 11, 17, and 36 weeks). An increase in mineral content and crystallinity in newly formed KO bone and collagen maturity at all sites was found using FTIR microspectroscopy and imaging; consistent with osteonectin's postulated role in regulating bone formation and remodeling.

Mineral and matrix properties of tibias of osteonectin-null mice and their age- and background-matched wildtype controls were compared using Fourier-transform infrared microspectroscopy (FTIRM) and infrared imaging (FTIRI) at 10- and 7-mm spatial resolution, respectively. The bones came from animals that were 11, 17, and 36 weeks of age. Individual FTIRM spectra were acquired from 20 × 20 μm areas, whereas 4096 simultaneous FTIRI spectra were acquired from 400 × 400 μm areas. The FTIRM data for mineral-to-matrix, mineral crystallinity, and collagen maturity were highly correlated with the FTIRI data in similar regions. In general, the osteonectin-null mice bones had higher mineral contents and greater crystallinity (crystal size and perfection) than the age-matched wildtype controls. Specifically, the mineral content of the newly forming periosteal bone was increased in the osteonectin-null mice; the crystallinity of the cortical bone was decreased in all but the oldest animals, relative to the wildtype. The most significant finding, however, was increased collagen maturity in both the cortical and trabecular bone of the osteonectin-null mice. These spectroscopic data are consistent with a mechanism of decreased bone formation and remodeling.


OSTEONECTIN, ALSO KNOWN as SPARC (secreted protein acidic and rich in cysteine) or BM-40, is a phosphorylated glycoprotein that is expressed predominately in bone (by osteoblasts) and in other tissues undergoing remodeling.(1) This “matricellular” protein, based on its numerous activities in a variety of organisms and its multiple sites of expression, is considered to be a pleiotropic factor(2) whose functions include antiadhesion, differentiation, and matrix organization. Matricellular proteins are secreted regulatory macromolecules that are not structural components of the extracellular matrix (ECM) but that mediate interactions between the ECM and cells. Osteonectin also stimulates angiogenesis in vivo and in vitro(3) and the production of matrix metalloproteinases.(4, 5) In vitro, osteonectin has been reported to bind to hydroxyapatite (HA; bone mineral), stimulate HA formation, inhibit HA growth, or have no effect, depending on concentration.(6–8) In the absence of cells, osteonectin chelates calcium, undergoing a conformational change.(9, 10) This is probably relevant for its interaction with HA.(8) Despite all these potential roles, and its abundance in bone, the specific function of osteonectin in mineralized tissues remains unknown.

Previously, several groups(11–13) reported that young osteonectin-null mice (KO) had no skeletal phenotype, but did have accelerated dermal wound healing and developed cataracts. However, in a more detailed study, we found(14) that osteonectin-KO mice develop a low-turnover osteopenia that becomes progressively worse as the animals age. In this study, at 11 weeks, the osteonectin-KO mice had a decreased radiographic density, a decreased trabecular bone volume relative to the wildtype (WT), and decreased bone formation rates.(14) By 17 weeks, the KO mice were severely osteopenic, and they had 50% the trabecular bone volume of the WT controls. By 36 weeks of age, the trabecular bone volume of osteonectin-KO mice was only 30% that of their WT controls. Interestingly, while WT and osteonectin-KO mice had similar cortical bone thickness, the mechanical properties of their long bones were different.(14) A three-point bending test that measures cortical bone properties showed that femora from WT mice increased in stiffness from 11 to 17 weeks of age and then showed a decrease in stiffness between 17 and 36 weeks of age. In contrast, the stiffness of femora from osteonectin-null mice did not change from 11 to 36 weeks of age. Thus, significant differences in stiffness were only observed at 17 weeks of age, a time of peak bone mass in mice.(15) At this time, osteonectin-KO femora had only 50% of the stiffness of WT controls.

To determine the underlying basis for these temporal differences in bone volume and biomechanical properties and to provide further insight into the bone composition in the osteonectin-KO mice, this study used infrared microspectroscopy and infrared imaging to characterize changes in both the mineral and matrix of the bones of osteonectin-KO and age- and background-matched WT animals previously analyzed. These spectroscopic techniques provide unique spatial information on mineral content, crystal size and perfection, and collagen maturity. Similar infrared spectroscopic analysis had previously been used to identify changes in mineral and matrix composition in mice with biglycan,(16) osteocalcin,(17) matrix gla protein,(18) and osteopontin(19) null mutations, as well as type X collagen transgenics(20) and naturally occurring collagen mutations.(21) These analyses allow further comparison of the impact of noncollagen matrix components on the mineral and matrix properties of bone.


Experimental animals

Four tibias per group were obtained from homozygous osteonectin-KO and WT animals, bred in a mixed genetic background of 129SVXC57BL/6 as previously described.(5, 14) The KO and WT animals, whose genotype was verified by southern analysis,(14) were 11, 17, and 36 weeks of age at time of death. The bones came from the same animals that had been used for the histomorphometry and mechanical studies(14) and were in the same background as the mutant animals previously reported to have no skeletal phenotype.(11–13) All animal protocols were approved by the Institutional Animal Care and Use Committee of Saint Francis Hospital and Medical Center. Tibias were cleaned of soft tissue, split longitudinally, and stored in 90% ethanol before embedding in Spurr's medium.(22) Longitudinal sections of the proximal tibias were cut on a Jung-K microtome at 2–3 μm thickness and mounted on barium fluoride infrared windows (Spectral Systems, Hopewell Junction, NY, USA). Static and dynamic histomorphometric analyses were performed as described.(14)

Fourier-transform infrared microspectroscopy and Fourier-transform infrared imaging

Sections (5–7) from all tibias were examined by Fourier-transform infrared microspectroscopy (FTIRM) using a BioRad FTS infrared microscope equipped with a mercury-cadmium-telluride (MCT) detector operating under nitrogen purge, with an automated x-y stage drive (BioRad, Cambridge, MA, USA). For the cortical bone of the proximal metaphysis in each section, spectra were collected from areas having dimensions of 20 × 20 μm in the bone immediately adjacent to the periosteum, in the cortical bone immediately adjacent to the endosteum, and in the central region of the cortical bone. For trabecular bone, spectra were obtained the same way from the trabeculae in the center of the metaphysis and distal to this area. Interferrograms (256) were collected under constant nitrogen purge, co-added, and Fourier-transformed as described elsewhere.(22)

The tibias of the 11- and 17-week-old animals were also examined by Fourier-transform infrared imaging (FTIRI) using the BioRad Sting-Ray system (BioRad). This system uses a 64 × 64 array detector to provide 4096 spectra simultaneously from 400 × 400-μm areas at ∼7 μm spatial resolution. At least five areas per bone were analyzed under nitrogen purge conditions. Areas containing only trabecular bone or only cortical bone were selected. Spectra were collected from overlapping areas along the entire length of the proximal tibias.

Data analysis

In the case of both FTIRM and FTIRI, the contribution of the embedding media was subtracted from all the raw spectra.(22) Spectra were base-lined, and mineral-to-matrix ratio (integrated areas of the phosphate v1, v3 band at 900–1200 cm−1 to the amide I band at 1585–1720 cm−1) and carbonate-to-phosphate ratio (integrated area of the carbonate v2 band at 860–890 cm−1 band to the above-mentioned phosphate v1, v3 band) were calculated as previously described.(23) For the FTIRM data, the phosphate and amide I bands were curve-fitted using a second derivative-based method,(24) and results were reported as percentage of total area under the curve. Mineral crystallinity, which reflects the apatite crystal size and perfection, was reported based on the ratio of the percent areas of subbands at 1030 and 1020 cm−1 and on the percent area of subbands at 1060, 1076, and 1103 cm−1. Collagen maturity (related to the relative proportion of pyridinium and reducible cross-links) was calculated from the ratio of the areas of subbands at 1660 and 1690 cm−1.(25)

Mineral-to-matrix ratios were calculated similarly from the FTIRI data using computer algorithms to correct for embedding media contributions, baseline the spectra, and calculate the relevant areas for each set of 4096 spectra. The Sting-Ray detector has a cut-off above 890 cm−1, precluding direct determination of carbonate-to-phosphate ratios. For FTIRI data, the mineral crystallinity parameter and the collagen maturity parameter were calculated based on peak intensity rather than on peak area ratios, because it is not possible to curve-fit 4096 spectra in each spectral image file to obtain area ratios.(23, 25) Images were produced from the raw spectral data using a combination of Bio-Rad Win IR-Pro, Microsoft Excel (Microsoft Corp., Bellevue, WA, USA), and Microcalc Origin (Origin Lab Corp., Northampton, MA, USA) software. Histograms describing the pixel distribution for each parameter were also calculated and mean and standard deviations for the pixel distributions were determined using JMP software (Statistical Software Institute, Inc., Cary, NC, USA).

Statistical analysis

For the FTIRM data, mean and SD for values from each site for each mouse age and phenotype were calculated, and KOs were compared with WTs using the Student's t-test with Bonferonni correction (GraphPad; ISI Software, Philadelphia, PA, USA), with n defined as the number of animals in each group. A p ≤ 0.05 value was taken as being statistically significant. For the FTIRI data, comparisons were made of the mean values in bone areas for each animal, with comparable statistical tests. Linear regression analysis was performed to compare results of both WTs and KOs obtained using both imaging modalities with the same software.


Although osteonectin-KO mice had decreased cortical bone formation rate, they did not have decreased cortical bone thickness (Table 1). However, infrared spectroscopic analysis of the bones of osteonectin-KO mice revealed significant changes in mineral content, mineral quality (crystallinity), and matrix quality (collagen maturity) relative to the WT mice at each age examined. The greatest differences were in actively forming and actively remodeling areas, e.g., the cortical bone adjacent to the periosteum and the cortical bone adjacent to the endosteum. The mineral content, expressed as mineral to matrix ratio for each of the four sites and all three ages of bone analyzed by FTIRM, is shown in Fig. 1. In the WT mice, the mineral content of the bone adjacent to the periosteum increased from 11 to 17 weeks and then decreased by 36 weeks (Fig. 1A). Surprisingly, the mineral content of cortical bone adjacent to the periosteum was higher in 11-week-old KO mice (Fig. 1A). In WT mice, as in other species, the mineral content of the cortical bone adjacent to the endosteum decreased with time.(26) Eleven-week-old osteonectin-KO mice had significantly less mineral content at this endosteal site; however, the mineral content did not vary with age in these mice (Fig. 1C). The mineral content of the center of the cortical bone, which is less actively remodeled, was similar in 11- and 17-week-old WT and osteonectin-KO mice. However, at 36 weeks of age, mineral content of the center cortical bone decreased in the WT mice but remained the same in mutant animals (Fig. 1B). Together, these data indicate that, although the osteonectin-KO mutation did not alter cortical bone thickness, it did affect the mineral content of this tissue. These data also support the concept from the trabecular bone histomorphometry that bone is not appropriately remodeled in osteonectin-KO mice. Trabecular bone volume was greatly reduced in osteonectin-KO mice, and trabecular mineral content was consistently reduced in 11- and 17-week-old KO mice, although these differences were not statistically significant (Fig. 1D). At 36 weeks of age, the osteonectin-null mice had only one or two visible trabeculae, and thus the trabecular mineral to matrix ratio could not be reliably determined. Similar patterns of variation in mineral content were noted in the infrared images, which vividly illustrate the spectral differences between WT and osteonectin-KO bone. Typical infrared images of the mineral to matrix ratio in cortical bone of KO and WT animals at 11 weeks are shown in Fig. 2a. The mineral to matrix ratio in the KO is lower and more uniform than that in the WT.

Table Table 1.. Static and Dynamic Histomorphometric Analysis of Cortical Bone From Tibias of Wildtype (WT) and Osteonectin-KO (KO) Mice (Mean ± SEM)
original image
Figure FIG. 1..

Mineral content, expressed as spectroscopically determined mineral to matrix ratios in the tibias of 11-, 17-, and 36-week-old osteonectin-KO and WT animals. Data represents mean ± SD for 5–15 spectra per site in n = 4 animals. *Significant differences between WT and KO at same site. The sites shown are (A) cortical bone adjacent to the periosteum, (B) bone in the center of the cortex, (C) cortical bone adjacent to the endosteum, and (D) trabecular bone.

Figure FIG. 2..

FTIR imaging of 11-week-old osteonectin-null (KO) and wildtype (WT) animals cortical bone. (a) Mineral-to-matrix ratios. (b) Infrared images of the mineral crystallinity (1030/1020 intensity ratios). (c) Infrared images of collagen maturity in 11-week-old animals. Each image represents an area 400 × 400 μm. The numerical scales shown represent the range of intensity ratios for each parameter and are the same for WT and KO. Background (PMMA only) is assigned a value of zero.

Curve-fitting of the phosphate band revealed several consistent trends in KO and WT bones. The broad phosphate bands are composed of more than 11 subbands, which can be assigned to phosphate vibrations in specific environments.(24) Detailed analyses of the underlying bands indicate relative amounts of nonstoichiometric apatite, carbonate, acid phosphate substitution, and crystallinity. Data for cortical bone in WT and KO mice are summarized in Table 2. Specifically, subbands directly related to crystallite size (1076 cm−1), inversely related to size (1060 and 1106 cm−1), and related to carbonate (1045 cm−1) and HPO4 (1123 cm−1) substitution (24, 27) indicated that in the younger (11- and 17-week-old) animals, the mineral in the KO was not as crystalline as that in the WT, whereas in the older (36-week-old) KO, crystals were larger and more perfect. Similarly, in the KO, as contrasted with the WT, the 1030/1020 peak area ratio, which is indicative of crystallinity (as measured by X-ray diffraction),(24) was decreased at 11 and 17 weeks and increased at 36 weeks. Figure 3 shows the curve-fit data for the 11- and 17-week-old trabecular bones. While at 11 weeks, significant differences were seen in the percentage area of the subbands at 1045, 1076, and 1123 cm−1, as well as in the 1030/1020 peak area ratio, by 17 weeks, the crystallinity in the knockout remained reduced, while the percentage area of the 1045 cm−1 was now decreased in the KO. These data show that KO tissue differs from WT in its carbonate for phosphate substitution, and it has decreased crystallinity.

Table Table 2.. Curve-Fit Parameters in Osteonectin-Null (KO) and Wildtype (WT) Mouse Tibias Presented as Percentage Area or Peak Area Ratios (Mean ± SD)
original image
Figure FIG. 3..

Curve-fit phosphate parameters for FTIR microspectroscopic data obtained from trabecular bone of osteonectin-KO and WT mice at 11 and 17 weeks of age. Mean ± SD for n = 4 bones per genotype. *p < 0.05 for same parameter in WT vs. KO. 1030/1020 ratios were multiplied by 10 for presentation ratios. The 1660/1690 subband ratio indicates collagen maturity. The phosphate subbands are those directly related to crystallite size (999 and 1076 cm−1), those inversely related to size (1060 and 1106 cm−1), and those related to carbonate (1045 cm−1) and HPO4 (1123 cm−1) substitution. The 1030/1020 peak area ratio is indicative of crystallinity.

Infrared imaging, in which crystallinity was assessed based on the 1030/1020 peak intensity ratio, showed similar age dependent findings (Fig. 2b). The same images as shown for mineral content in Fig. 2a are presented here, depicting crystallinity, and they show that at 11 weeks, crystallinity of the KO bone is lower than that of the WT. These data agree with the finding from trabecular bone histomorphometry that osteonectin-KO mice have decreased matrix apposition rate and decreased matrix turnover, which is most apparent in older animals.

While FTIRM and FTIRI give equivalent data for mineral to matrix distribution (linearly related to ash weight),(28) the absolute values of crystallinity calculated from curve-fit peak area ratios and peak intensity ratios were different. To determine whether these values were correlated, the crystallinity values calculated from intensity ratios for each 400 μm × 400 mm area were plotted against the mean peak area ratio calculated from 5–10 spectra obtained from 20 × 20 μm areas in that 400 μm × 400 μm area field. WT and KO values were both included. The data were correlated with an r = 0.8, p = 0.06 (Fig. 4).

Figure FIG. 4..

Correlation between intensity ratios obtained by FTIR imaging and curve-fit peak area ratios for FTIR microspectroscopic data. Values on x axis represent mean values determined in the same section within the same 400 μm × 400 μm field in data from all genotypes in 11- and 17-week-old animals.

The amide I band (1660/1690 subband ratio) was similarly analyzed both by curve fitting and calculation of peak intensity ratios from FTIRM and FTIRI data, respectively. The curve fit data for the collagen is included in Table 2, whereas Fig. 2c presents the imaging data for the sites shown in Figs. 2a and 2b, showing the increase in collagen maturity in KO trabecular and cortical bone at each age analyzed as contrasted with the WT at equivalent ages.


Previously we showed that osteonectin-KO mice develop osteopenia because of decreased trabecular bone remodeling and decreased trabecular bone formation rate.(14) In the current studies, a detailed analysis of the mineral and matrix components of cortical and trabecular bone from osteonectin-KO mice provides additional insight into the function of osteonectin in bone. These studies show that in the osteonectin-KO mice, the mineral content of the newly formed periosteal bone is increased, the crystallinity of the cortical bone is decreased in all but the oldest animals, and the collagen maturity is increased.

This study provides much needed data on cortical bone in osteonectin-KO mice. In the cortices of these animals, bone formation rate is decreased, yet cortical thickness is not decreased (Table 1). The spectroscopic data most likely reflect the persistence of both existing mineral and of the more highly crossed linked matrix. Failure to resorb the bone would result in increased mineral content in cortical bone, explaining the difference between the mineral content in the younger (11 weeks) and older (17 and 36 weeks) mice (Fig. 1). Indeed, the osteonectin-null mutation resulted in decreased osteoblast and osteoclast numbers and surface, supporting the concept of decreased bone resorption as well as decreased bone formation. More recent studies indicate that in the absence of osteonectin, osteoblastic cells are more susceptible to stress and achieve a less mature phenotype than osteonectin-expressing cells.(29) These data are consistent with the histomorphometric findings of decreased osteoblast number and bone formation rate. In addition, because osteoblast-derived factors support osteoclast formation and maturation, these data also help explain the decreased osteoclast numbers and surface observed in mutant mice in vivo.

Crystallinity in osteonectin-KO cortical and trabecular bones was decreased relative to the WT in the younger animals, indicative of the impaired bone formation rate, but was increased relative to the WT in older animals, presumably because of the failure of the bones to be remodeled (Table 2). This could explain the increased brittleness of the bones of older animals. More importantly, and perhaps providing a greater insight into the reason for the altered phenotype, was the increase in collagen maturity. The increased collagen maturity is analogous to what is seen in states of osteopenia with low bone remodeling.(25) If bone formation is delayed, and remodeling accelerated, there is a greater opportunity for mature collagen cross-links to form and degrade. In contrast, in states of decreased remodeling, as in the osteonectin-KO animals, collagen ages without being degraded and replaced, and the mature collagen cross-links would persist. The skin of osteonectin-KO mice has been shown to have thinner collagen fibrils, and significantly less collagen content.(30) Because osteonectin can regulate collagen fibrillogenesis,(8) these changes may also reflect the importance of osteonectin for proper collagen deposition.(31) Furthermore, improper collagen cross-linking, could in turn, affect mineral deposition.(32)

Whole bone mechanical properties, after correcting for variations in geometry, have been shown in rodents to vary directly with mineral content (bone mineral density [BMD]),(33–35) with crystallite size,(36) and with collagen maturity.(37, 38) In the present study, the greatest differences in mineral content were noted in the 36-week-old KO mice; greatest decrease in crystallinity was noted in the 11- and 17-week-old animals, and increased collagen maturity was noted at all ages examined. Previously, we showed that the mechanical properties of cortical bone in the femora from osteonectin-KO mice are impaired,(14) resulting in increased fragility despite a normal cortical area. It is possible that the lack of appropriate matrix composition and decreased crystallinity as demonstrated in the tibial cortex contribute to the impaired biomechanical properties of the cortical bone.

The osteonectin-KO mouse is the first example of a bone-matrix protein-deficient animal studied by infrared microspectroscopy in which matrix properties as well as mineral properties are significantly compromised. In the osteocalcin knockout, the mineral content was increased at all sites, particularly in the trabecular bone; however, the crystallinity was not significantly altered, implying a role for this protein in remodeling.(17) In the osteopontin null animals,(18) the mineral content of the cortical bone was increased, as was the crystallinity, but there were few changes in metaphyseal bone. These data were consistent with osteopontin having both a physical chemical effect as an inhibitor of mineral crystal proliferation, but also a cellular effect as an osteoclast recruiter. The biglycan knockout(16) had less mineral in newly forming bone sites, but the crystals were larger, again consistent with the role of biglycan as a bone mineral nucleator.(39) The osteonectin-KO mice resembles each of these to some extent, because the crystals in the younger bone are smaller/less perfect, but in the mature bone are larger, in agreement with the view that the protein is regulating bone formation and remodeling. Investigations of double-knockouts may reveal how some of these matrix proteins interact.


This study was supported by National Institutes of Health Grants DE04141 (ALB), AR44877 (AMD), and AR21707 (EC). The authors would like thank L Spevak and P Carroll for their assistance with the data analyses.