1.

Introduction

It is well known that the susceptibility to bone fracture increases with age;¹ indeed, much work has focused on the mechanisms of bone deterioration responsible for this elevated fracture risk.² Bone mass or bone mineral density (BMD) has traditionally been used as a clinical predictor of such fracture risk. However, there is mounting evidence that BMD alone may not be the sole factor responsible for the aging-induced fracture risk.^{1 3 4} For example, work by Hui etal.¹ reported a roughly tenfold increase in fracture risk with aging in a sample group in which the BMD remained practically unchanged. Consequently, in an effort to develop a more complete understanding, a number of studies have investigated the influence of several factors (e.g., anatomical location, mineral density, porosity, collagen quality, mineral physicochemistry, etc.) on the evolution of the strength and toughness properties of bone with age in both human and animal subjects.^{5 6 7 8 9 10 11 12 13 14 15 16 17 18}

Understanding the mechanical properties of bone will require studies that probe over multiple dimensions owing to the complex, hierarchical structure of bone.^{19 20 21} At the macrostructural level, bone is distinguished into cortical (compact) and cancellous (trabecular) bone; most long bones are composed of a cortical shell with a cancellous interior. At microstructural length scales, cortical bone is organized into 200–300 μm diameter secondary osteons,²¹ which are composed of large vascular channels (50–90 μm diameter) surrounded by circumferential lamellar rings (3–7 μm thick), with so-called “cement lines” at the outer boundary.²⁰ These secondary osteons are the end result of the remodeling process that repairs damage in vivo and are believed to be involved in the toughening of bone.²² At the nanostructural level, the lamellae are composed of organic type-I mineralized collagen fibers (up to 15 μm in length, 50–70 nm in diameter, and formed by regular arrangement of subnanostructural collagen molecules) bound and impregnated with inorganic carbonated apatite nanocrystals (tens of nanometers in length and width, 2–3 nm in thickness).¹⁹

Over the past two decades, a number of vibrational spectroscopy techniques, including infrared (IR) absorption and reflectance (e.g., Refs. 23 24 25 26 27 28 29 30 31 32), Raman scattering (e.g., Refs. 16 and 33 34 35 36 37), and inelastic neutron scattering (e.g., Refs. 38 and 39), have been used to study both the organic and inorganic components of bone. Carden and Morris⁴⁰ report on a recent and comprehensive review of Raman and IR studies in this area. Spectroscopic studies have targeted issues that include mineralization (e.g., Refs. 25 and 39), disease (e.g., Refs. 26 and 28), aging (e.g., Refs. 16, 23, 24, 29, 31, and 33), and mechanical deformation (e.g., Refs. 35). Although IR-based techniques are relatively more well established, Raman spectroscopy, with its finer (∼1 μm in a Raman microprobe) spatial resolution, is increasingly being used.⁴⁰ Sample fluorescence, which can obscure the Raman scattering signal, can be a problem in performing in situ studies of bone and other biological samples with visible laser excitation. Near-IR excitation (see Ref. 40 and references cited therein) and sophisticated background subtraction methodologies³⁵ have been used successfully to partially avoid this problem in bone samples.

Surprisingly though, there are relatively few studies that have used vibrational spectroscopy to specifically study age-related changes in the bone matrix. Fourier transform infrared spectroscopy (FTIR) and Raman studies focused on the mineral component have revealed some small changes in the mineral content, crystallinity and chemical nature.^{16 23 24 33} Paschalis etal.³¹ reported age-related changes in the amide I band in bovine bone (it is unclear whether trabecular/cortical bone was used), which were associated with an increase in the nonreducible (trivalent) pyridinoline (Pyr) cross links and a decrease in the reducible (divalent) dehydrodihydroxylysinonorleucine (deH-DHLNL) cross links. Very etal.²⁹ investigated small age-related changes in the amide bands in the FTIR spectrum of human trabecular bone and found that, with increasing age, the amide peaks shifted to slightly lower energies and the relative intensities of the high energy components of the amide I band decreased with age.

However, most spectroscopic studies have not correlated the subnanostructural information gained from the vibrational spectra directly to the mechanical properties, and in particular the fracture resistance, of the bone specimen under study. Here, we show that the use of ultraviolet excitation eliminates fluorescence interference and allows nondestructive in situ measurements of Raman spectra from human cortical bone. Due to resonance effects, peaks associated with the bone organic component (particularly the amide I band) are selectively enhanced. Significant changes in the shape of the amide bands are found, which are correlated both with age of the bone donor and with the fracture properties of the specific bone specimen. The null hypothesis offered here is that the amide I band can be resonance-enhanced through the use of deep-UV excitation and that its shape is affected by aging.

2.

Materials and Methods

3.

Results

3.1.

UV Raman Spectroscopy

Figure 1(a) shows the fluorescence spectrum from 250 to 800 nm obtained from human cortical bone under 244 nm (deep-UV) excitation. Evidently, the fluorescence is strong throughout the visible region of the spectrum and well into the near-IR regime. However, with 244 nm excitation, Stokes Raman shifts between 500 and 2000 cm⁻¹ are observed between 247 and 256 nm. The bulk of the fluorescence occurs to the red of this wavelength range. Hence, as shown in Fig. 1(b), it is possible to obtain Raman spectra with a very low background that does not substantially interfere with the collection of the vibrational spectrum of interest. Data obtained separately from a similar human bone sample with 632.8 nm (visible) excitation are shown in Fig. 1(b) for comparison; due to strong fluorescence interference, only the phosphate ν₁ peak can be observed. Although we do not know if UV excites fluorescence more effectively than visible light and indeed, the short absorption length of the deep-UV light might cause the effective fluorescence signal to be smaller due to the smaller volume of material excited, deep-UV excitation was found to enable low-background Raman spectroscopy in spite of strong fluorescence.

Figure 1

(a) Photoluminescence spectrum of human bone excited with 244 nm laser light and detected with single spectrometer, photomultiplier, and lock-in amplifier. Strong fluorescence exists throughout the visible and into the near-IR regime. However, the fluorescence signal is low between the laser wavelength and 260 nm, enabling Raman shifts of up to 2500 cm⁻¹ to be observed with low background. (b) Comparision of Raman spectra obtained with 244 nm (bottom) and 632.8 (top) nm light. Strong fluorescence background obscures all but the strongest peak (phosphate ν₁) in the visible (632.8 nm) Raman spectrum.

Figure 2 compares vibrational spectra from bone samples (human and animal) obtained with UV Raman, near-IR Raman, IR reflectance, and inelastic neutron scattering. The near-IR spectrum obtained using 1064 nm excitation, from Smith and Rehman,⁴¹ is dominated by the phosphate ν₁ and other inorganic features [Fig. 2(b)]. In fact, the strength and narrow linewidth of this feature has enabled it to measure changes in mineral content of bone in the context of mechanical deformation.³⁵ The FTIR spectrum [Fig. 2(c)] is similar to that of the near-IR Raman spectrum, except that the carbonate features are dominant. Although a number of groups have previously reported spectra obtained with visible and near IR lasers (530–785 nm) in which fluorescence does not appear to be a problem, for the material studied here 632.8 nm excitation produced spectra in which only one peak (phosphate ν₁) was definitively observable, as shown in Fig. 1(b). Whether near-IR excitation might reduce the background further in our samples remains to be investigated. Clearly, the organic features (amides I, II, and III and the CH ₂ wag) are resonance enhanced and are typically more prominent than the inorganic features in the UV Raman spectrum. This spectrum resembles that obtained from inelastic neutron scattering [Fig. 2(d)], a technique which is more sensitive to vibrations with significant hydrogen motion.³⁹ Deep-UV excitation thus leads to very distinct, well-defined, resonance-enhanced Raman peaks from the organic matrix, with almost complete absence of fluorescence, and without the need for demineralization, making it an excellent candidate for nondestructive spectroscopic evaluation of bone and other mineralized tissues. Thus, the combination of visible Raman for careful measurement of the inorganic component and of the organic/inorganic ratio and UV Raman for detailed matrix analysis might be a powerful tool to probe such tissues.

Figure 2

Comparison of vibrational spectra of bone obtained with different spectroscopic techniques: (a) ultraviolet Raman spectrum of human bone, this work; (b) near infrared Raman spectrum of bovine bone;⁴¹ (c) IR reflectance of human bone, this work; (d) inelastic neutron scattering of ox femur.³⁰ Strong inorganic (phosphate and carbonate) and organic (amide I, II, and III and CH ₂ wag) are noted. Measurement conditions were reported as follows: (a) UV Raman, 244 nm, 500 mm spot size, <5 mW laser power, 100 second collection time, subtraction of small, linear background; (b) near-IR Raman, 1064 nm excitation, 650 mW laser power, 4 cm⁻¹ resolution, white light corrected; (c) FTIR obtained in reflectance with 4 cm⁻¹ resolution, using same sample as in (a); (d) 2.5 powdered sample, 12 h collection time.

3.2.

Spectral Analysis

Figure 3(a) shows data in the organic portion of the spectrum (∼1100–1800 cm⁻¹) for three representative specimens. To the authors’ knowledge, these are the first Raman spectra from mineralized biological tissues such as bone that utilized a deep-ultraviolet laser excitation source. The spectra were all normalized to the height of the CH ₂ wag peak, the position (and height) of which was observed to be practically invariant with age and across donors/specimens. In particular, three organic bands are strongly resonance enhanced and definitively identified here—amide III (primarily from the in-phase combination of NH in-plane bend and CN stretch, ∼1245–1260 cm⁻¹), CH ₂ wag (∼1454–1461 cm⁻¹) and amide I (primarily from the C=O stretch, ∼1626–1656 cm⁻¹). The amide II band (primarily from the out-of-phase combination of NH in-plane bend and CN stretch) that is sometimes seen in IR spectra at ∼1540–1580 cm⁻¹ (Ref. 40) was rather weak in the present data (Fig. 3), presumably as it was not as strongly resonance enhanced.

Figure 3

UV Raman spectra of human humerus bone from donors of three different ages. The relative height of the amide I feature compared to the CH ₂ wag is larger in the sample from the older donor, and there is a positional shift of this band to higher energies with aging.

Large changes in peak shape were observed for the amide I band, particularly in the spectra from the very elderly donors (85–99 years). When the band is considered as a whole, it appears that its center frequency shifts by as much as ∼20 cm⁻¹ (Fig. 3). The amide III and CH ₂ wag bands did not show significant variations in band position (less than ∼2 cm⁻¹). The amide I band also showed variations in intensity with age (Fig. 3). The changes in band shape were made more quantitative by fitting the amide I region to overlapping Gaussian peaks as described in Sec. 2.3. (We do not know a priori if changes in the underlying structure of the band produce the observed shifts in its apparent center frequency or whether it is the opposite, that is, that the shifts in the band center cause the changes in the underlying structure.) The results of this procedure are shown in Fig. 4 in which four individual peaks—one each corresponding to the CH ₂ wag and the amide II bands, and two for the amide I band—are used to fit the spectra between 1400 and 1800 cm⁻¹.

Figure 4

Details of the peak fitting to the CH ₂ wag and amide I region of the UV Raman spectrum for human cortical bone from a young (top) and old (bottom) donor. A good description of the spectrum is found by nonlinear least squares fitting to four peaks centered at ∼1450 cm⁻¹ (CH ₂ wag), 1510 cm⁻¹ (amide II), and 1610 and 1655 cm⁻¹ (amide I with underlying structure). The donor age is indicated in each case.

Peak fitting results from all spectra were evaluated using linear regression analysis against age and fracture toughness data. The peak position of all four fit peaks was only weakly correlated with age. Figure 5 shows peak areas for the two peaks underlying the amide I band (∼1610 and 1655 cm⁻¹) against age. The areas of both peaks showed good correlation with age (r=−0.6138 and 0.6570 for peaks at 1610 and 1655 cm⁻¹, respectively), as did the ratio of the areas (r=−0.6026). The area of these peaks show some significant variation with age (P=0.0787 and 0.0545 for peaks at 1610 and 1655 cm⁻¹, respectively), as do the ratio of the areas, 1610/1655 cm⁻¹ (P=0.0858). The changes in the ratios of the peak areas are consistent with shift in the center of the amide I band; it appears that an increase in the intensity of the higher energy (1655 cm⁻¹) feature in samples from older donors is responsible for the evident shift of the center of the asymmetric amide I band.

Figure 5

Variation in the area of the two peaks (∼1610 and 1655 cm⁻¹) underlying the amide I band with donor age. The correlation coefficient, r, and the P value for significance are included in each plot and a linear regression of the data is also shown. The “area” was defined as the product of the height and the full width at half-maximum (FWHM) of the peak.

4.

Discussion

The amide bands, particularly amides I and III, are believed to be good indicators of protein conformation because of the role of the amide moiety in cross linking and bonding.⁴² Examination of the present data shows that both the height and location of the resonance-enhanced amide I band vary with age (Fig. 3), supporting the proposed null hypothesis. The band center position moves to a higher energy with increasing age due to the larger contribution from the high energy component, as was evident when the underlying peaks were analyzed [Fig. 5(c)].

We had previously demonstrated the effect of aging on the fracture toughness “resistance curve” (R curve) behavior of human cortical bone using the same specimens used here and reported decreases in both the crack-initiation and crack-growth toughnesses that were correlated with donor age.¹⁸ [Although the use of a single-value measure of the fracture toughness is appropriate for certain materials, in many ductile and brittle materials, including cortical bone,²² the fracture resistance actually increases with crack extension, requiring a resistance curve (R curve) fracture-mechanics approach.^{43 44} In particular, R curves are necessary to describe the fracture resistance of materials toughened by crack-tip shielding,^{45 46 47} i.e., mechanisms such as crack bridging or constrained microcracking, which develop in the crack wake as the crack extends. In such instances, crack extension commences at a crack-initiation toughness, and with further crack extension requires a higher driving force until typically a “plateau” or steady-state toughness is reached. The corresponding slope of the R curve can be considered as a measure of the crack-growth toughness.] Both measures of the toughness are needed for a complete understanding of the fracture resistance of materials such as cortical bone that show evidence of extrinsic toughening.^{18 22} [Crack propagation can be considered as a mutual competition between two classes of mechanisms: intrinsic mechanisms, which are microstructural damage mechanisms that operate ahead of the crack tip and promote crack growth, and extrinsic mechanisms, which act to “shield” the crack from the applied driving force and operate principally in the wake of the crack tip to inhibit crack growth.^{45 46 47}] Linear regression analyses of the area ratio of the 1610 and 1655 cm⁻¹ peaks (the two peaks underlying the amide I band as observed in deep-UV Raman spectroscopy) against crack-initiation and crack-growth toughness are shown in Fig. 6. A more significant correlation is seen in the case of the crack-initiation toughness (r=0.69 and P=0.04) compared to the crack-growth toughness (r=0.44 and P=0.23) . These results suggest that the “quality” of the collagen has a stronger influence on crack-initiation toughness as opposed to crack-propagation toughness.

Figure 6

Variation in the crack initiation and growth toughness with peak area ratio for the 1610 and 1655 cm⁻¹ peaks underlying the amide I band. The correlation coefficient, r, and the P value for significance are included in each plot and a linear regression of the data is also shown.

We have previously demonstrated that the crack-growth toughness in cortical bone is primary the result of the extrinsic toughening mechanism of crack bridging.²² The primary source of such bridging in human bone appears to be the formation of so-called uncracked ligaments in the crack wake,^{18 22} which are essentially “bridges” of material spanning the crack behind the crack tip. These are created either by the nonuniform advance of the crack front and/or by the imperfect linking of microcracks ahead of the crack tip with the main crack. Our previous observations showed that cracks growing in bone were deflected around the osteons, and propagated preferentially along the presumably weaker cement lines. Such behavior would imply that the tissue microstructure would be expected to play a more important role in older bone and is consistent with our previous observations that the more extensive remodeling in older bone⁴⁸ leads to lower amounts of extrinsic toughening by crack bridging, and hence, lower crack-growth toughness.¹⁸ Consequently, it is not surprising that more significant correlations were found in the case of the crack-initiation toughness, rather than the crack-growth toughness, as we can presume that the former is more strongly affected by the collagen quality. Thus, these results show that when considering the effect of aging in mineralized tissues, it is important to account for not just the changes at nanostructural dimensions, but also at the microstructural level.

At the molecular (subnanostructural) level, it has been suggested that the deleterious effects of aging can be attributed to changes in the cross-linking chemistry in bone.^{49 50} Changes in the nature of the cross linking are known to affect the biomechanical properties of bone.⁵¹ Although most studies looking at age-related changes in the cross linking have utilized high-performance liquid chromatography (e.g., Refs. 12, and 52 53 54), Paschalis etal.^{31 32} have deconvoluted FTIR spectroscopic data into overlapping component peaks to probe the secondary structures in bone collagen. These authors reported that two peaks at ∼1660 and 1690 cm⁻¹ underlying the amide I band are particularly indicative of the cross linking, with the area ratio between them (1660/1690) corresponding to the nonreducible/reducible cross-link ratio. This ratio was observed to increase with age in demineralized bovine bone. Our data did not reveal an underlying peak at ∼1690 cm⁻¹; this component of the amide I band is not resonance enhanced. However, we did observe an aging-induced increase in the contribution of a peak at ∼1655 cm⁻¹, consistent with an increase in the nonreducible (Pyr) cross-link content in bone (Fig. 4). It has also been suggested that, with time, the reducible DHLNL matures into the nonreducible Pyr cross link, leading to an increase in the content of the latter.⁵⁵ These results imply that the protein conformation in the organic matrix (which is ∼90–95% collagen) does indeed change with age. More detailed studies with appropriate purified peptide standards are currently being undertaken to quantify such changes.

In summary, we have used deep-UV Raman spectroscopy to show that aging affects the organic component of the human cortical bone, in particular the amide I band. Moreover, we have demonstrated that deep-UV excitation leads to very distinct, well-defined Raman bands from the organic matrix, with almost complete absence of fluorescence, and without the need for demineralization; the technique is therefore an excellent candidate for nondestructive spectroscopic evaluation of mineralized tissues. Spectral changes are also directly correlated, for the first time, to the mechanical properties, more specifically, to the crack-initiation and crack-growth fracture toughness. However, it must be noted that a limitation of this study is the relatively small sample size, specifically 18 specimens taken from 9 separate donors. More detailed investigations which look at the specific effects of variables such as gender, anatomical location, microstructural variations, and bone mineral density, are required to obtain a more complete understanding and are currently being undertaken. Although we took precautions to minimize microstructural variations by extracting the specimens from roughly the same positions from all the humeri used, and the relatively large laser spot size (together with the use of a rotating stage) helps sample a substantial portion of each specimen thereby eliminating bias from local microstructural variations, we expect that some of the variables listed above contributed to the observed scatter. Clearly, further investigations are necessary to assess how aging-related changes affect bone at the subnanostructural levels, particularly to define the cross-linking chemistry and the consequent effect on “bone quality.” Such an approach is important from the perspective of gaining a full understanding of the effects of aging on the fracture resistance of mineralized tissues.