1.

Introduction

Bone is a composite tissue that consists of mineral, organic matrix, water, and cells. Bone provides mechanical strength and protection to the body, and maintains mineral ion homeostasis. Always in a dynamic state, bone is deposited by cells known as osteoblasts, and removed by multinucleated cells known as osteoclasts. This formation and resorption process is often referred to as “remodeling.” When either of these processes is impaired or excessive, bone no longer has optimal mechanical properties. The best-known example of this imbalance between bone formation and bone resorption occurs in osteoporosis,^{1 2 3} a disease that affects both women and men, and sometimes children,⁴ and is characterized by an increase in porosity resulting in an increased risk of fracture. After a first fracture, the risk of sustaining a second fracture is extremely high.³ Osteoporosis may occur because osteoclasts are too active or because osteoblasts are inactive. The purpose of this review is to describe how infrared spectroscopy is helping us understand the changes in osteoporotic bone before and after therapeutic intervention, and to illustrate how other bone diseases can be studied with infrared (IR).

Infrared analysis was first applied to bone more than half a century ago.^{5 6} It was initially used to confirm the x-ray diffraction result that bone mineral was analogous to the naturally occurring geologic mineral, hydroxyapatite (HA) (Fig. 1).⁵ As instrumental sophistication increased, IR was used to determine the presence and nature of IR active bone impurities (HPO ₄,CO ₃), which substitute in HA lattice positions.^{6 7 8} IR and Fourier-transform IR (FTIR) were similarly used to demonstrate changes in HA composition caused by diet,^{9 10 11} space flight,¹² and chronic hypoperfusion.¹³ IR spectroscopy was used in 1966 to describe the changes in the crystal size and perfection of the bone apatite particles (% crystallinity).¹⁴ The increases in size and perfection were correlated with the sharpening of the c-axis x-ray diffraction peak. It should be noted that based on x-ray diffraction one cannot distinguish absolute increases in crystallite size from formation of a stoichiometrically more perfect lattice structure. The change in x-ray diffraction pattern was initially described as a transition from an amorphous (noncrystalline) calcium phosphate to a poorly crystalline hydroxyapatite,¹⁴ but later was recognized as the maturation of nanoparticles of hydroxyapatite.¹⁵ FTIR spectroscopy has frequently been used to characterize the mineralized matrix formed in cell and organ cultures (e.g., Refs. 16 17 18 19 20 21), and recently it has been used to demonstrate the properties of the tissue formed on bone-inductive materials.^{22 23 24} IR spectroscopy was also used to identify the presence of bone-like mineral in pathologic (abnormal) mineral deposits.^{25 26 27 28}

Figure 1

Structural formula for the hydroxyapatite lattice showing substituents that might be found in each lattice site.

The more recent instrumental advance in which a FTIR spectrometer is coupled with an optical microscope revolutionized the FTIR study of bone by enabling spectra to be acquired at anatomically discrete sites. This was extremely important because bone is a heterogeneous tissue (Fig. 2) and spatial variation had previously been lost due to the need to homogenize the bone before analysis. With FTIR microspectroscopy, spectra could be obtained at sites along an osteon (Fig. 3)²⁹ or during the process of new bone deposition.³⁰

Figure 2

Three-dimensional view of half a mouse femur based on micro-computerized tomography. Bone components discussed in the text are illustrated, including cortical bone with its periosteal and endosteal surfaces, the growth plate, and individual trabeculae.

Figure 3

Osteonal bone. A FTIR image illustrating the variation in mineral: matrix ratio along with a photomicrograph showing the concentric rings around the blood vessel in the center of the osteon, and individual spectra taken 20 μm apart along the osteon.

Addition of a focal plane array detector to the infrared microscope was another major advance in the IR analysis of bone as it allowed about a thousand fold increase in the speed of data acquisition. With the concomitant improvements in computer processing capabilities, hyperspectral images consisting of a complete mid IR spectrum at each x, y location in a thin section (see Fig. 3) can be readily acquired. In these images each pixel corresponds to a point in the x-y plane of the sample, with the z axis corresponding to the magnitude of the desired IR parameter (univariate or multivariate) at that site. Fourier-transform infrared microspectroscopic imaging was rapidly applied to the description of anatomic changes in bone properties,³¹ analyses of mineralization in cell cultures,³² and characterization of bone formation on implant materials.^{33 34} It has been most extensively applied to evaluation of diseased bone, which will be the main subject of this review.

There are limitations to IR imaging and microspectroscopy that should be noted. As discussed later, thin (2–6 μm) sections are required, and the samples cannot be examined in a hydrated form because of water interference with the spectra. While reflectance spectra can be used to examine thick specimens, there are reflectance artifacts that still remain to be overcome before surface reflectance can be applied. Perhaps the greatest limitation is that to do an IR analysis of bone a biopsy is required, thus this is an invasive technique.

A typical spectrum of ground bone is shown in Fig. 4. Only the frequencies relevant to the analyses of bone composition are shown. There are phosphate vibrations, carbonate vibrations, representing the extent and nature of carbonate substitution in the HA lattice, and strong protein peaks (Amide I, Amide II, Amide A). The latter arise from peptide bond C=O stretch, mixed N–H in plane bending and C=O stretch and N–H stretching, respectively. These protein bands arise mainly from collagen which accounts for almost 95% of the organic matrix. The other 5% of the matrix contains noncollagenous extracellular matrix proteins whose functions are reviewed elsewhere,³⁵ lipids, and cell components.

Figure 4

FTIR spectrum of ground adult mouse bone showing frequencies of interest.

A quantitative assessment of the mineral content of the bone being analyzed can be made by calculating the ratio of the integrated phosphate v1, v3, and amide I peaks. Mixtures of collagen and synthetic HA have been analyzed both by gravimetric and FTIR analyses and shown to be linearly related validating this ratio.³⁶ Carbonate to phosphate ratios can similarly be calculated from their respective integrated areas, and these ratios can be used to follow the incorporation of carbonate into the bone mineral.³⁷ Deconvolution of the carbonate bands can be used to determine whether the carbonate is substituting for phosphate or hydroxide.^{6 7} The validation of the carbonate: phosphate ratio was done by chemical measurements of carbonate content.^{38 39}

2.

Methods for FTIR Analyses of Bone

Bone, by its very nature, is a hard material, and hence, it is extremely difficult to prepare the thin sections required for microscopic examination in the visible spectral region. Sections thicker than 5–6 μm do not transmit sufficient light, hence, IR analysis of bone routinely requires samples 2–5 μm in thickness. To circumvent the problem of obtaining thin sections, pathologists routinely embed bone in a resin that is harder than the bone itself. This embedded bone mixture can then be sectioned on a microtome with a diamond or tungsten knife. While this facilitates specimen preparation, the process induces complications in the IR analysis of bone. The embedding materials all have IR active modes, and thus spectral subtraction must be used to remove these contributions.⁴⁰ While spectral subtraction of embedding media is fairly straight forward for single spectra [there is a unique band for poly(methylmethacrylate) (PMMA), the most common embedding material, at 1729 cm⁻¹] it is more of a challenge when subtracting PMMA from hyperspectral cubes containing thousands of spectra, and specialized algorithms have been developed for this purpose. Specifically, using ISYS software (Spectral Dimensions, Olney, MD), spectra in the region of interest are truncated, baseline corrected, and “normalized” to a pixel containing only PMMA, making each PMMA intensity in the cube equal to 1.0. The normalized PMMA spectrum is then subtracted from each of the normalized spectra, creating spectra at each pixel that are either 0 if the region contained only PMMA, or no evidence of the PMMA band.

The bone sites utilized for data collection are dependent on the question being asked. For example, when temporal changes in bone development are of interest, scans going from the outer bone forming surface (periosteum) to the inner bone resorbing surface (endosteum) are most useful (Fig. 2). If comparisons are made between a diseased and a healthy bone, different areas (cortex, trabeculae) may be examined.

Parameters that are measured in addition to the earlier-mentioned mineral: matrix ratio and carbonate: phosphate ratio are crystallinity,^{38 41} collagen maturity,⁴² and acid phosphate content.⁴³ Several of these parameters, as described later, have been validated by independent methods. Mineral to matrix ratio is correlated to the ash content when synthetic mixtures of hydroxyapatite and collagen are analyzed by both methods (Fig. 5).³⁶ As reviewed elsewhere,^{2 36} increasing mineralization leads to stronger bones, although when there is excessive mineralization bones can become brittle.

Figure 5

Development of spectra structure correlation. Spectrally determined mineral matrix ratio correlates with conventionally determined ash weight.

The crystallinity parameter (1030/1020 peak area ratio) was validated by comparison with x-ray diffraction analysis of HA particle size.³⁸ Investigation of the collagen maturity parameter which is related to the presence of stable collagen crosslinks, is currently in progress using type I collagen peptides whose composition was determined by HPLC. Whereas with FTIR microscopic data, the mineral parameters can be calculated by resolving the peaks in the complex ν₁, ν₃ contour using curve-fitting, two-dimensional-IR, spectral deconvolution or other methods),⁴⁴ for FTIR images, the large number of spectra which would have to be analyzed precludes the use of these algorithms, since each spectrum has to be individually examined. The need to analyze thousands of spectra in an image led to the use of intensities at particular frequency positions, rather than peak areas to represent the concentration of a particular component. While these intensities are linearly correlated with the same data obtained by curve-fitting based on a pixel by pixel comparison,⁴⁵ the range of intensity values is narrower than the range of peak area values, making the intensity ratio less sensitive because of a loss of dynamic range.

The composite picture of each of these parameters, whether obtained from imaging or point-by-point microscopic data, gives an image of the composition, crystallinity, and amount of bone mineral. We believe that one or more of these IR parameters are major predictors of “bone quality” or the tendency of bone to fracture. Other factors influencing bone quality are bone geometry, bone microarchitecture (especially connectivity), the presence of microcracks, and the density of the bones. Bone quality is altered with age, as mineral content increases, crystallinity reach maximal values, and collagen matures.⁴⁶ Our laboratory is trying to correlate spectroscopic changes in bone IR parameters with changes in mechanical properties, either tested in the same sample or in adjacent tissue. There are limited data indicating that crystal size matters and that a broad distribution of mineral: matrix and crystallinity values is found in healthier bones.⁴⁶ There is also a strong indication that crystal size is related to mechanical strength, as shown in a study by Turner in which the crystal thickness was inversely related to strength in fluoride treated animals.⁴⁷ Since HA crystal thickness decreases with age as crystal length increases, this is in accord with our preliminary data that mechanical strength is greater when average crystallinity is greater. Recently Raman spectroscopy has also shown an inverse correlation between crystallinity and elastic deformation in a rat aging model.⁴⁸

3.

Bone Quality

The quality of bone refers to its optimal ability to perform its functions, as described in terms of bone strength. Bone strength in turn is determined in part by both geometric and material properties,⁴⁹ bone mineral density,⁵⁰ bone connectivity (whether trabecular bone struts are connected),⁵¹ and genetics.⁵² We have proposed the features of bone characterized by IR parameters also contribute to bone strength, and are demonstrating that they show variation in both health and disease.

3.2.

Osteoporotic versus Normal Bone

The first IR analyses of bones from osteoporotic patients^{54 55} as reviewed elsewhere⁵⁶ were inconclusive, because tissues had been homogenized and some contained sites of recent fractures. IR microspectroscopy and imaging facilitated the description of the changes occurring in bones of humans and animals with osteoporosis.

Using biopsies from patients who had sustained osteoporotic fractures, but were at a distance from the fracture site, our studies consistently showed a decrease in mineral content, an increase in crystallinity, and an increase in collagen maturity in both the trabecular and cortical bones of the patients as contrasted with age-matched controls.^{57 58} More significantly the age-dependent temporal changes in these parameters were not detected in the untreated osteoporotic patients. Figure 6 compares the mineral content and crystallinity in trabecular bone of one normal and one untreated osteoporotic female. Typical of the osteoporotic cases studied to date, the trabeculae in the osteoporotic case are thinner than those in the normal; the mineral/matrix ratio in the osteoporotic patient is significantly reduced, while the bone crystallinity is increased. These all may contribute to the increased fragility of osteoporotic bone.

Figure 6

Image of trabecular bone in normal and osteoporotic females. Right figure shows mineral: matrix ratio and left figure shows crystallinity.

Comparable changes in IR parameters were found in ovariectomized monkeys^{58 59 60} indicating the excellence of this model for studies of the human disease. Ovariectomized dogs failed to show the changes in mineral: matrix ratio and mineral crystallinity,⁶¹ although they did manifest treatment-dependent changes. Similarly, in an unpublished pilot study of ovariectomized sheep (n=5) there were no significant changes in imaging parameters, however sheep subjected to metabolic acidosis (MA) did have changes in mineral to matrix ratio similar to those seen in osteoporotic humans [Fig. 7(a)]. The carbonate; phosphate distribution in the MA sheep was also different from that in the controls [Fig. 7(b)]. The pixel histograms describing the distributions of these parameters in the image shown are also presented.

Figure 7

Comparison of (a) the mineral: matrix ratio and (b) carbonate: phosphate ratio in vertebrae of a sheep with metabolic acidosis (left) contrasted with a normal age-matched control sheep (right). Pixel histograms show the distribution of parameters in the biopsy. Courtesy of Paul West, Hospital for Special Surgery.

Manipulation of genes in mice has become routine, and genetic polymorphisms associated with osteoporosis in humans can be examined in rodent models. While rodent models do not develop osteoporosis, they do lose bone, and hence, can be used for evaluating therapies, or validating spectroscopic parameters. Using congenic mice, Blank et al. was able to relate their mechanical strength to mineral: matrix ratio, and to show a direct correlation between this ratio and crystallinity.⁶²

Overexpression of the vitamin D receptor in mice resulted in a phenotype in which calcium absorption was disturbed. There while bones were thicker and the distribution of peak calcium content as determined by backscatter electron imaging was increased and sharpened, there were, however, no changes in mineral content, mineral composition, crystallinity, particle size, or density as determined by a number of physicochemical techniques including IR.⁶³ Male and female mice in which the noncollagenous protein osteonectin expression is knocked out sustain rapid loss of trabeculae,⁴⁵ and IR microspectroscopy demonstrated an increase in collagen maturity, a decrease in mineral content, and an increase in crystal size, reminiscent of human changes in osteoporosis. Mice which lack osteocalcin expression, have thickened bones, and fail to resorb their bone. FTIR microspectroscopy demonstrated that the crystals were smaller although there were more crystals present, indicative of defective bone modeling in these animals.⁶⁴ Mice lacking the matrix protein, osteopontin, have increased mineral content, and unlike the osteocalcin deficient mice, increased crystal size,⁶⁵ implicating this protein both in regulation of crystal size and in regulation of remodeling. There are numerous other models which have been created to mimic human genetic defects associated with both decreased and increased bone density [e.g., Refs. 66 67 68 69 70], but their mineral properties, have not yet been reported.

3.3.

Therapies for Osteoporosis

A large number of pharmaceutical interventions are currently being evaluated for the prevention and treatment of osteoporosis.⁷¹ Because there are surrogate markers of bone quality, fewer and fewer biopsies are being collected for histomorphometric evaluation of bone turnover, thus there are fewer biopsies that can be used to quantify bone mineral properties. This has caused more and more reliance on the animal models discussed earlier. We have, however, shown that hormone replacement therapy in women just reaching their menopause improves bone quality,⁷² Similarly, we have shown, in agreement with published data on mineral density,⁷³ that the selective estrogen receptor modulator, raloxifene, is not detrimental to bone quality (manuscript in preparation). Estrogen itself was demonstrated by IR imaging, as summarized in Fig. 8, to facilitate the rate of healing in a rat fracture injury model.⁷⁴

Figure 8

Effect of estrogen (solid circles) on time-dependent changes in healing fracture callus in an ovariectomized rat is accelerated, and rapidly reaches the values seen in control bone (open circles) at a location distant from the fracture site in the same animals. Values are mean±S.D. for five rats. Courtesy of H. OuYang, thesis, Rutgers University, Department of Chemistry.

Other studies reported determination of the effects of therapies on mineral quality in simian and rodent models. In a mouse model of brittle bone disease, a condition in man and mouse caused by a variety of mutations in the collagen gene, the bisphosphonate alendronate reduces fracture risk, without causing a statistically significant change in mineral crystallinity.^{75 76} Similarly a study of dogs treated with both alendronate, and another bisphosphonate, risedronate, did not show significant changes in mineral properties as determined by IR of ground bones.⁷⁷ FTIR imaging of biopsies from osteoporotic patients or animal models treated with the bisphosphonates currently in clinical use are still needed.

Studies in monkeys have shown that parathyroid hormone increases new bone formation,⁷⁸ and that a bone resorption inhibitor, nandrolone decanoate, also improves bone quality by stimulating new bone formation, evidenced by increased acid phosphate, and carbonate content, and lower mineral: matrix ratios.⁵⁸ Ibandronate, another bisphosphonate,⁷⁹ and strontium ranelate,⁸⁰ drugs in clinical trial have been evaluated in monkeys, and the properties of those bones in the near future should be evaluated by IR methods.

4.

Other Bone Diseases

In addition to osteoporosis, a great deal of information, beyond what could be acquired by standard histochemical techniques, is being learned from FTIR microspectroscopic studies of other bone diseases. In osteoarthritis, there is some debate as to whether the disease starts in bone and progresses to the overlying cartilage, or vice versa. FTIR analyses of human tissues⁸¹ showed the mineral: matrix ratio was highest in normal cortical bone of the ulna, and lower in the sclerotic area of the affected site, which in turn was higher than values in loose bodies or trabecular bone of the tibia. This led the investigators to suggest that demineralization was considerably higher in osteoarthritis of the knee joint and remineralization was likely to occur under the acidic conditions in the articular region in this disease.

Osteogenesis imperfecta or brittle bone disease, as mentioned earlier, is a rare condition characterized by defects in the collagen gene and increased risk of fracture. FTIR analyses have demonstrated that the mineral: matrix ratio in the mouse model of a moderate-to-severe phenotype is reduced,⁷⁵ most likely because of a deficit in collagen content, and that the mineral crystals may be outside the collagen fibrils, and in general, are smaller than those in age-matched controls.⁸² Other than a crystallographic evaluation of mineral crystal properties in humans with this disease,⁸² and a recent Raman evaluation of a “knock-in” mutation,⁸³ there is little data on the effect of the various collagen mutations on mineral crystal size and perfection. Such analyses are extremely important as collagen provides the backbone on which bone mineral crystals deposit.

Osteopetrosis is another rare disease, this one due to abnormalities in the way osteoclasts (bone resorbing cells) develop. In this disease the marrow cavity is absent, and the bone is like rock because the remodeling process is impaired. Bone crystals in the rat model of this disease are decreased in size⁸⁴ while mineral content is increased. In a single human case, the same increase in mineral: matrix ratio and decrease in crystallinity was noted by infrared imaging.⁸⁵

The last example is a more frequently observed disease, osteomalacia. This disease is caused by deficiencies in vitamin D. This vitamin is also the hormone that is responsible for regulating serum calcium levels, by regulating intestinal and kidney adsorption, and by contributing to the regulation of bone remodeling. In children, the disease is known as rickets, and is characterized by softened bones. In adults, the disease is manifest as an increased amount of unmineralized bone (osteoid). We used IR imaging to examine the mineralized matrix in patients with “normal” bone and patients diagnosed with osteomalacia. In the cortical bone, there were no differences in IR parameters when the unmineralized osteoid was excluded from analysis by spectral masking. In the trabecular bone, there was a general decrease in mineral content, a significant increase in matrix maturity, but no difference in crystallinity.⁸⁶ These differences could not be assessed by standard histomorphometry.

5.

Conclusions

The application of infrared microspectroscopy and infrared microscopic imaging to the characterization of bone diseases and evaluation of the effects of therapies on the bone properties in these diseases is in its infancy. While IR has had a long history in the bone field, it is these newer applications that will have the greatest impact on the medical profession and on the population at large. The next generation of technological advances will include more sophisticated means for data processing, including cluster and factor analyses, and detectors for imaging instruments which will allow access to the v₄ phosphate bands which currently cannot be examined with imaging spectrometers.