Near Infrared Spectroscopic Assessment of Loosely and Tightly Bound Cortical Bone Water

Abstract

Water is an important component of bone and plays a key role in its mechanical and structural integrity. Water molecules in bone are located in different locations, including loosely or tightly bound to the matrix and/or mineral (biological apatite) phases. Identification of water location and interactions with matrix components impact bone function but have been challenging to assess. Here, we used near infrared (NIR) spectroscopy to identify loosely and tightly bound water present in cortical bone. In hydrated samples, NIR spectra have two primary water absorption bands at frequencies of ~5200 and 7000 cm⁻¹. Using lyophilization and hydrogen-deuterium exchange assays, we showed that these absorption bands are primarily associated with loosely bound bone water. Using further demineralization assays, thermal denaturation, and comparison to standards, we found that these absorption bands have underlying components associated with water molecules tightly bound to bone. In dehydrated samples, the peak at ~5200 cm⁻¹ was assigned to a combination of water tightly bound to collagen and to mineral, whereas the peak at 7000 cm⁻¹ was exclusively associated with tightly bound mineral water. We also found significant positive correlations between the NIR mineral absorption bands and the mineral content as determined by an established mid infrared spectroscopic parameter, phosphate/amide I. Moreover, the NIR water data showed correlation trends with tissue mineral density (TMD) in cortical bone tissues. These observations reveal the ability of NIR spectroscopy to non-destructively identify loosely and tightly bound water in bone, which could have further applications in biomineralization and biomedical studies.

INTRODUCTION

The quality of bone mechanical and structural properties are affected not only by changes in the primary components of mineral and collagen, but also by extent of hydration¹. Bone water plays a crucial role in stabilizing collagen’s triple helix through intra and inter molecular bonding², contributing to the mechanical properties of bone. Water present on the surface of the mineral also contributes to mechanical properties of bone by mediating the structuring and organization of the biological apatite crystals³. An approximate molecular formula for biological apatite can be suggested as Ca_7.5(PO₄)_2.8(HPO₄)_2.6(CO₃)_0.6(OH)_0.2•(H₂O)x, where X indicates water molecules bound within the lattice structure or to the surface of the crystals^4–6. Recently, Nyman et al. ^7–10 performed an extensive study on the effect of dehydration on the biomechanical properties of bone using the techniques of evaporation, and thermal and solvent dehydration on samples. They found that with age, the amount of bound water was reduced, and resulted in a related decrease in the overall bone toughness. Thus, it was concluded that bone hydration contributes to optimal biomechanical properties of mineralized collagen fibers.

Water is present in different regions within the bone matrix, generally defined as loosely or tightly bound to collagen or mineral^2,3,9. The term “loosely bound” refers to water present at the surface and interface of collagen fibrils and mineral crystals (physical adsorption), which can be removed through evaporation or lyophilization. “Tightly bound” water is trapped within the collagen triple helix or the mineral crystal lattice (chemical adsorption), requiring high temperature conditions for removal⁹. Calorimetry of frozen tissues is the conventional method to differentiate between loosely bound and tightly bound water⁹. However, this method is time consuming, destructive and dependent on factors such as the freezing temperature of tissues⁹. Non-destructive modalities such as magnetic resonance (MR) spectroscopy and imaging and vibrational spectroscopy (infrared and Raman) have been used to quantify the total amount of water in bone, as well as to differentiate loosely and tightly bound bone water^11–15. MR spectroscopy and imaging can be used to detect collagen-associated water; however, due to a short T₂ signal, it is very difficult to assess mineral-related water content¹⁶.

Infrared spectroscopy in the near infrared (NIR) frequency range (12000–4000 cm⁻¹) is well-known for its sensitivity to water¹⁷, and to organic moieties, with absorption bands in the NIR spectrum arising from combinations and overtones of O-H, N-H, C-O and C-H vibrations ¹⁸. NIR analysis has been previously used to not only differentiate between loosely and tightly bound water in musculoskeletal tissues such as cartilage, but also to identify key matrix and water absorption bands in cortical bone^19,20. NIR matrix absorption bands associated with collagen have been identified at 4260 cm⁻¹, 4608cm⁻¹ and 4890 cm ⁻¹, and arise from the combinations of CH₂ bending, C-H stretching and bending and N-H bending vibrations, respectively⁸. Two primary NIR cortical bone water absorption bands occur at the frequencies of ~5200 and ~7000 cm^{−1 11}, and arise from a combination of O-H bending and stretching, and from the O-H stretching first overtone, respectively. Recently, another NIR absorption band at 7000 cm⁻¹ has been described as potentially arising from a structural OH vibration related to hydroxyapatite ²¹. However, further studies are necessary to confirm this absorption band is also associated with bone mineral, and in general, to determine how the NIR water absorption bands can inform about the environments of water in bone.

The present study uses several experimental approaches to identify and quantify NIR absorption bands associated with bone water loosely or tightly bound to collagen and mineral. We anticipate that the establishment of NIR methods to elucidate differences in bone water and to quantify bone mineral content will have applications for non-destructive analysis of bone and mineral quality in a wide range of sample types.

METHODS

Overall approach

Several different approaches were employed to spectroscopically evaluate loosely bound water, and water that is tightly bound to either collagen or mineral in cortical bone (described below, and in Figure 1). The techniques included lyophilization, hydrogen/deuterium (H/D) exchange, demineralization, denaturation, comparison to NIR spectra of bone standards with different degrees of matrix mineralization, and comparison to NIR spectra of calcium phosphate standards with or without bound water. Changes in the NIR spectral contour of the two established cortical bone water absorption bands at 5200 and 7000 cm⁻¹ were analyzed in samples with known compositional differences, and conclusions were drawn to better understand how distinct types of bone water can be identified and quantified based on NIR spectra.

Figure 1:

Flow chart of different approaches and techniques used to identify and analyze NIR water peaks associated with collagen and/or mineral components of bone (TMD = tissue mineral density).

Sample processing for NIR identification of water absorption bands associated with collagen and mineral in bone

Removal of loosely bound bone water: All tissue samples were initially lyophilized (Martin Christ-Alpha 1–2) for 5 minutes to remove excess water. Five minutes of lyophilization was chosen to remove macroscopic surface water, and also likely removed macroscopic pore water, which has been shown to correspond to a loss of 1.4–5.7 % of original bone mass²². The majority of the remaining loosely bound water was removed by lyophilization for 48 hours. To confirm that all loosely bound water was removed, lyophilized samples were subjected to H/D exchange by complete submersion in liquid D₂O (Sigma Aldrich) from 1 minute to 24 hours. After immersion in D₂O, the samples were dabbed with absorbent wipes immediately before analysis. H/D exchange involves replacement of available hydrogens with deuterium atoms in O-H, N-H or S-H bonds. This process would result in all freely available hydrogen ions in the native bone water exchanging with deuterium ions, resulting in the formation of X-D bonds, where X = O, N or S. Since deuterium is a heavier isotope of hydrogen, X-D bonds will vibrate at a lower NIR frequency than X-H bonds²³. Thus, by investigation of NIR spectra of bone before and after soaking in deuterium, we can elucidate the type and location of bound water in the NIR region. If there is no change in spectra after soaking in deuterium, there were no freely available hydrogens (water molecules). It is important to note that D₂O has an absorption band close to 5200 cm⁻¹, which coincides with a water absorption band of interest. However, the intensity of the D₂O absorption band in this region is much lower that of water²⁴, thus we hypothesize that an H/D exchange of water by D₂O would still result in a significant decrease in the absorption band seen at 5200 cm⁻¹ in bone samples.

Tightly bound collagen water: To specifically identify absorption bands associated with water tightly bound to collagen, we used previously characterized demineralized porcine cortical bone samples²⁵. Type I collagen and stoichiometric hydroxyapatite (Sigma Aldrich) powders were pressed into 100 mg pure pellets and used as standards for reference. The samples were lyophilized for 48 hours to remove loosely bound water. The demineralized bone samples were heated at 75 degrees Celsius for two hours in an oven²⁶ (Precision 658, Thermo Scientific), a procedure previously described for protein denaturation. Denaturing unravels collagen’s triple helix, exposing the tightly associated water molecules within the structure and loosening the protein-water bonds²⁷. Therefore, we hypothesized that water tightly bound to collagen will be lost after denaturization, which would be observed as a decrease in the corresponding NIR water absorption band.

Tightly bound mineral water: To identify absorption bands arising from water tightly bound to mineral, we investigated the NIR spectra of standard calcium phosphate compounds with and without bound water, brushite (CaHPO₄·2H₂O) and monetite (CaHPO₄), respectively. The sole difference in their composition is the presence of tightly bound water. Brushite and monetite powders (Sigma Aldrich) were pressed into 100 mg pure pellets and lyophilized for 48 hours to remove loosely bound water. Following lyophilization, the samples underwent H/D exchange to ensure that any remaining water peaks observed in the spectra corresponded to tightly bound water molecules. We hypothesized that a NIR absorption band associated with mineral could be assigned to water tightly bound to the mineral if it corresponded to a absorption band in the O-H region present in the spectrum of brushite and absent in the spectrum of monetite.

Sample processing for NIR quantification of mineral content

To assess whether a mineral-associated absorption band from bone mineral is present in the NIR region, we compared the spectra of previously characterized mineralized and demineralized porcine cortical bone samples²⁵. We hypothesized that removing mineral from bone would lead to a reduction in a mineral-associated absorption band, regardless of whether it arises directly from the mineral phosphate, or from water tightly bound to mineral. The intensity of the potential mineral NIR absorption band was then evaluated for correlation with the MIR-determined mineral content of serially demineralized bone samples. Tibiae from young bovine bone (Research 87) were pulverized (Spex Freezer/Mill 6770) to powder, and one gram of bone powder was mixed with 100 ml of 10 mM EDTA-Tris solution (Sigma Aldrich) for demineralization and stirred together for one week with intermittent EDTA changes. The demineralization process was monitored by MIR assessment of aliquots of bone over a one-week period, and NIR data obtained from the lyophilized demineralized aliquots pressed into pellets (Carver Press 4350), as described below. We hypothesized that the successful identification of an NIR bone mineral absorption band would be marked by a strong positive correlation (R>0.8, p < 0.05) between the specific NIR absorption band and the MIR-determined mineral content of the bone samples.

Correlation of NIR spectral data with cortical bone mineral density

We hypothesized that specific NIR absorption bands associated with mineral, matrix and/or water components correlated to variations in human cortical bone tissue mineral density (TMD) determined by micro-computed tomography (micro-CT).

Human Bone Tissues: Cadaveric human tibiae segments harvested from 10% along the distal end obtained from 6 female donors (33 to 88 years) with no evidence of skeletal disease (NDRI, Philadelphia, PA) were cut cross-sectionally to uniform 500 μm thick sections using a diamond wafering saw (Buehler Isomet 1000). The sections were ultrasonicated (60 khz FS60D Fisher Scientific) for 120 minutes in 1% tergazyme detergent solution to remove bone marrow and stored in a solution of phosphate buffer solution (PBS) and protease inhibitor (PI) at −20°C until use. The sections were then thawed and lyophilized for 5 minutes to remove remnant PBS/PI solution on the surface of the bone samples. Near infrared spectral imaging data (NIRSI) in an environmentally-controlled chamber (described below) were collected prior to micro-CT imaging.

Bone tissue processing and micro-CT image acquisition and data analysis: Bone sections were lyophilized for 48 hours prior to micro-CT data collection. This was done to avoid inconsistencies associated with water loss during micro-CT data collection. It was previously confirmed in a subset of 3 samples that micro-CT data parameters from wet and lyophilized samples did not differ (data not shown). Data from the lyophilized cortical bone cross-sections were obtained with a high-resolution cone-beam micro-CT scanner (SkyScan 1172; Bruker microCT, Kontich, Belgium) equipped with a Hamamatsu C9300 11fremMp camera using the double-side and oversize sample options. The parameters used for micro-CT data collection were as follows: Source voltage, 62 kV; source current, 131 mA; pixel resolution, 9.89µ; rotation step, 0.2°; 5 frame averaging and random movement compensation was 10. The acquisition time for each sample was 10 hours. The raw micro-CT images were reconstructed using cone-beam reconstruction software (SkyScan NRecon) with a ring artifact correction of 10, and a beam hardening correction of 60%. The reconstructed images were rotated in the transverse direction using DataViewer software. Cortical bone only was selected as the volume of interest (VOI) using a uniform threshold of 80 Hounsfield units (HU). Micro CT data were also collected from known standards of calcium hydroxyapatite (CaHA) (0.25 and 0.75 g/ cm⁻³ 4 mm in diameter) which were used as calibration values from which cortical bone TMD values were calculated ²⁸.

NIR spectroscopy data collection and analysis

NIR spectral data from all samples were obtained in diffuse reflectance mode using an ASD LabSpec 4 Standard-Res spectrometer coupled to a NIR fiber optic probe comprised of low-OH quartz fibers with a 3 mm diameter tip (Malvern Panalytical). This spectrometer is a portable dispersive instrument with an optical resolution of ~10 nm in the NIR range. Fifty co-added scans were collected and ratioed to a background spectrum collected from a Spectralon reference standard. At least three spectra were collected from each sample and averaged. The X-axis values in the spectra obtained from the ASD spectrometer have regular linear intervals to wavelength, with 1 nm spacing between datapoints. To show the spectra with the X-axis in wavenumber scale, we used data interpolation to obtain spectra with 8 cm⁻¹ equidistant spacing between datapoints. NIR raw spectra tend to have broad and overlapping absorption bands, and therefore, a second derivative filter (SavitzkyGolay, 2^nd order polynomial and 27 points of smoothing) was applied to raw spectra to resolve underlying peaks, while maintaining signal-to-noise quality. Second derivative peak height data have previously been shown to strongly correlate to integrated areas of absorption bands¹¹, which reflect relative amounts of the components of interest. However, it is important to consider that this correlation may only be correct when the bandwidths remain constant, which may be unlikely to happen among absorption bands from different water species. Nonetheless, an advantage of using derivatives is their ability to filter-out strong broad absorption bands, thereby revealing the presence of weak sharp bands. Here, the second derivative spectra were multiplied by negative one to display positive peaks. An environmental sample chamber with an inflow of either dry air or deuterium vapor, dependent on the specific experiment, was used during NIR data collection ²⁹. Dry air was circulated inside the chamber during NIR data collection of lyophilized samples to prevent absorption of water from the environment. Deuterium vapor flow was supplied into the chamber during H/D exchange experiments to prevent exchange of deuterium ions back to hydrogen ions in the samples.

MIR data collection and analysis

MIR data from lyophilized mineralized and demineralized bone samples were acquired in attenuated total reflection (ATR) mode using a Thermo Scientific Nicolet iS5 spectrometer equipped with an iD7 ATR accessory with a diamond crystal. Spectra were collected with 4 cm⁻¹ resolution and 32 co-added scans from samples and from an air background. At least three spectra were collected from each sample and averaged. The MIR raw spectra were baseline corrected and normalized to amide I, the absorption band that arises from the protein component of bone, primarily collagen. MIR spectral analysis of relative mineral content was based on the ratio of the second derivative peak heights of the primary phosphate peak at 1030 cm⁻¹, arising from the asymmetric stretching of P-O bonds, to the amide I vibration of proteins at 1660 cm⁻¹, which arises from carbonyl stretching, termed the phosphate/amide I ratio. This ratio has previously been shown to correlate to bone ash weight content, a measure of tissue mineralization³⁰.

Statistical analysis

All spectral experiments were repeated twice to confirm reproducibility of data. A Pearson correlation was used to evaluate the strength of linear relationships in the data, with the R value reported, and a p-value < 0.05 considered to be statistically significant. Correlations were investigated between NIR peak heights and ratios and MIR-determined phosphate/amide I. Additionally, tissue mineral density was correlated to NIR-derived water content, and to age of sample donors.

RESULTS

NIR identification of water associated with collagen and mineral in bone

Analysis of the raw (Fig. 2a) and second derivative (Fig. 2b) NIR spectra of hydrated, lyophilized and deuterated bone allowed differentiation of the primary water and matrix absorption bands. Lyophilization of the samples led to a marked decrease in the absorption band intensities at 5200 and 7000 cm⁻¹, confirming that they are primarily associated with loosely bound water in hydrated bone, as described in previous studies³¹. This decrease in intensity was not observed in the matrix absorption bands, such as of collagen at 4608 cm^{−1 32}. There was, however, some intensity remaining in both primary water regions after lyophilization. H/D exchange was performed for 24 hours to confirm whether all loosely bound water was removed by lyophilization. Interestingly, the 7000 cm⁻¹ absorption band had similar intensities in lyophilized and deuterated samples, but in the 5200 cm⁻¹ region, there was a small decrease in peak intensity, which indicated that lyophilization had not removed all the loosely bound water. Nevertheless, in both regions, at 5200 and 7000 cm⁻¹, absorption bands remained after lyophilization and H/D exchange, indicating that these peaks can be identified as water tightly bound to bone. The studies below then assessed which bone component the remaining absorption bands were associated with.

Figure 2:

Raw (a) and inverted second derivative (b) bone spectra. Two main water absorption bands are observed at ~5200 and 7000 cm⁻¹. Lyophilization and H/D exchange show markedly lower peak intensities at the water absorption bands, confirming that those peaks in hydrated samples are primarily associated with loosely bound water. In dehydrated (lyophilized) samples, the remaining peaks in the water regions are identified as water tightly bound to bone. The collagen peak at ~4608 cm⁻¹ was not reduced by lyophilization or H/D exchange.

Bone samples were analyzed after demineralization to further understand whether the small peaks remaining at 5200 and 7000 cm⁻¹ in lyophilized samples were related to water tightly bound to either collagen or mineral. Analysis of the second derivative of the spectra (Fig. 3a) shows that the peak at 7000 cm⁻¹ is the only one that substantially decreases after demineralization, thus strongly indicating that this peak arises from a moiety with an OH stretch that is primarily associated with bone mineral. Additionally, there was a noticeable shift of the 5200 cm⁻¹ peak to a lower frequency with demineralization. Comparison to NIR spectra of pure collagen and hydroxyapatite standards helped to identify underlying components of the 5200 cm⁻¹ peak. It is clear that while the peak seen in bone has a maxima closer to that of hydroxyapatite at 5225 cm⁻¹, after demineralization, this peak position shifts and becomes similar to that observed in collagen at 5170 cm⁻¹ (Fig. 3b). Therefore, we can conclude that the broad 5200 cm⁻¹ absorption band in lyophilized bone has underlying highly overlapped components that arise from mineral-bound OH (5225cm⁻¹ component), as well as from collagen-bound OH (5170 cm⁻¹ component).

Figure 3:

(a) Inverted second derivative spectra of lyophilized and demineralized lyophilized bone. Demineralization of bone resulted in increased heights of peaks associated with matrix absorption bands, and changes in spectral contour of the two primary water absorption bands. In particular, the peak at 7000 cm⁻¹ is reduced, showing that it is clearly associated with the presence of mineral. The peak at 5200 cm⁻¹ shows a noticeable shift. (b) Inverted second derivative spectra of bone, collagen and mineral standards. In dehydrated bone, two underlying components can be identified in the 5200 cm⁻¹ peak region, which are primarily associated with collagen (5170 cm⁻¹) and mineral (5225 cm⁻¹).

To assess whether the 5170 cm⁻¹ absorption band arises from water tightly bound to collagen, or from another OH-containing molecule, demineralized bone was denatured, aiming to disrupt the protein structure and loosen any water that was tightly bound to it. Analysis of the second derivative of the demineralized bone spectra after denaturation (Fig. 4) shows that only the 5170 cm⁻¹ peak is substantially reduced after denaturing. This indicates that in lyophilized samples, this peak can be primarily identified as water tightly bound to collagen. The other absorption bands in the NIR spectra of collagen are still present with denaturation.

Figure 4:

After denaturing demineralized bone, there is a pronounced decrease in peak intensity at the ~5170 cm⁻¹ water absorption band. This indicated that this peak in dehydrated samples is partially associated with water molecules tightly bound to collagen, which were lost after protein unfolding caused by denaturation.

As described above, the 7000 and 5225 cm⁻¹ absorption bands in lyophilized bone are OH absorption bands related to the mineral phase. However, we sought to confirm the assignment of these peaks to a specific mineral-related molecular bond vibration, either bound water, or a hydroxyl group within the apatite structure, by examination of the NIR spectra of mineral standards with and without bound water, brushite and monetite, respectively. The samples were lyophilized and subjected to H/D exchange to ensure observation only of tightly bound OH peaks. Second derivatives of the spectra (Fig. 5) shows that while brushite has clear peaks at 7000 and 5225 cm⁻¹, these peaks are absent in monetite. The difference in spectra between monetite and brushite can only be attributed to tightly bound water present in the composition of brushite. Therefore, we can identify the 7000 and 5225 cm⁻¹ peaks in lyophilized bone as water tightly bound to the mineral phase.

Figure 5:

Comparison of the spectrum of brushite (CaHPO₄·2H₂O) and monetite (CaHPO₄) shows that the peak at ~7000 and 5200 cm⁻¹ arises from mineral-bound water, present in brushite but not in monetite. H/D exchange confirms that the water is tightly bound to the mineral, as no frequency shifts are evident.

A summary of the loosely and tightly bound water peaks identified in hydrated and lyophilized bone is found in Table 1.

Table 1:

Assignments of NIR spectral absorption bands in hydrated and lyophilized bone samples.

Frequency (cm−1)	Hydrated bone	Dehydrated bone
5200	Loosely bound water	Combination of tightly bound water associated with collagen and mineral. Underlying components at: - 5170 cm−1: water associated with collagen - 5225 cm−1: water associated with mineral
7000	Loosely bound water	Tightly bound water associated with mineral

NIR quantification of mineral content

NIR spectra are not sensitive to the primary vibrations that arise from phosphate in mineral²¹, and thus to date cannot be used to quantify mineral. However, it may be possible to use the above-described NIR peaks of water tightly bound to the mineral as a NIR biomarker to reflect the presence of bone mineral. Thus, we hypothesized that the NIR mineral-associated water peaks can be used to quantify bone mineral content in lyophilized samples, after loosely bound water is removed. For this analysis, serially demineralized samples were evaluated, and MIR was used to obtain standard mineral content values, by quantifying the established phosphate/amide I peak ratio. Bone demineralization can be illustrated by the MIR raw spectra (Fig. 6a), which shows the relative decrease of the phosphate peak with progressive demineralization. It is also possible to note that the OH absorption band is not observed in the MIR spectra, illustrating the OH deficiency of bone mineral^33,34. In the second derivative of the NIR spectra obtained from lyophilized aliquots of progressively demineralized samples (Fig. 6b), both the 5200 and 7000 cm⁻¹ absorption bands also decreased in accordance with progression of demineralization over time.

Figure 6:

Serial demineralization of cortical bone after lyophilization in (a) raw MIR spectra and (b) NIR inverted second derivative spectra. MIR raw spectra show that increasing demineralization leads to a decrease the phosphate absorption band. NIR spectra show that both water absorption bands, at 5200 and 7000 cm⁻¹, decrease with increasing demineralization, indicating that both are associated with bone mineral. Inset in (a) shows that the OH absorption band is not observed in the spectra, illustrating the OH deficiency of bone mineral.

In serially demineralized lyophilized bone samples, the 5200 cm⁻¹ (Fig. 7a) and 7000 cm⁻¹ (Fig. 7b) peak heights showed significant positive correlation to mineral content. It is interesting to note that the values obtained in the correlations are stronger with the 7000 cm⁻¹ peak. This is most likely due to the observation above, that this peak is associated primarily with mineral, while the 5200 cm⁻¹ peak has contributions from both mineral and collagen components. In contrast, the collagen absorption band at 4608 cm⁻¹ had a negative correlation with mineral content. These data clearly demonstrate that that the NIR absorption bands thought to be associated with mineral are indeed more present in mineralized samples, while the collagen-associated NIR peaks are more pronounced in demineralized samples. Accordingly, the ratios of the mineral peaks to the collagen peak also showed significant positive correlation to MIR-determined mineral content, as shown in both the 5200/4608 ratio (Fig. 7d) and the 7000/4608 ratio (Fig. 7e). These results provide strong evidence that quantification of the peaks of water tightly bound to mineral at 5200 and 7000 cm⁻¹ can be used to investigate mineral content in lyophilized samples.

Figure 7:

NIR-determined second derivative (inverted) peak heights obtained from serially demineralized lyophilized bone correlated to MIR phosphate/amide I ratio, which reflects mineral content. Individual peak heights at a) 5200 cm⁻¹ and b) 7000 cm⁻¹ had a significant correlation to mineral content. c) The collagen peak at 4608 cm⁻¹ had a negative correlation to mineral content, as expected. NIR water/matrix ratios (d,e) also show strong correlations to mineral content.

Correlation of NIR spectral data with human cortical bone properties

An interesting application of the NIR spectral analysis of peaks identified and quantified above may be for clinically relevant studies of mineralization. In a series of female tibiae samples obtained from donors with a range of ages (Fig. 8), micro-CT determined TMD significantly correlated with donor age (Fig. 9a). Interestingly, in hydrated bones, quantification of the 5200/4608 peak ratio had a negative trend with increasing TMD (Fig. 9b), whereas, in lyophilized bones, the 7000/4608 peak ratio has a positive trend with TMD, although significance was not reached (Fig. 9c). This suggests that with increasing tissue mineralization, total water content (combined loosely and tightly bound, assessed in the hydrated bones) decreases, while the specific water component tightly bound to the mineral increases, reflecting an increase in mineral content. It is possible that increased sample size will further establish these relationships. Nevertheless, these preliminary results shed insights into the potential application of NIR approaches to non-destructively evaluate bone quality.

Figure 8:

Micro-CT data were obtained from cross-sections of tibial bone from female donors, ranging from 33 years old (a) to 88 years old (b). Qualitatively, a decrease in cortical bone thickness with age and increased porosity is evident.

Figure 9:

a) A significant correlation was found between TMD and donor age. b) In hydrated samples, NIR quantification of water content tends to show a negative correlation with TMD. c) In lyophilized samples, quantification of tightly bound mineral water, reflecting mineral content, tends to show a positive correlation with TMD.

DISCUSSION

The primary contribution of this study is the demonstration that NIR spectroscopy can be used to identify different bone water components associated with collagen and mineral. We show that, in hydrated bone samples, the peaks at ~5200 and 7000 cm⁻¹ can be used to evaluate loosely bound water components; whereas in dehydrated (lyophilized) samples, the peak at ~5200 cm⁻¹ reflects a combination of water tightly bound to collagen and to mineral, and the peak at 7000 cm⁻¹ can inform more specifically about mineral-bound water. A particular novelty of this work is the identification of a NIR absorption band assigned to bone mineral, biological hydroxyapatite. Using the NIR peak of mineral-associated tightly bound water as a biomarker of bone mineral opens a promising window into the application of non-destructive approaches to evaluate bone properties in a large variety of samples, albeit in a dry state.

Non-destructive techniques such as MR spectroscopy and imaging have helped discriminate bone water into two primary categories, free and collagen-bound. In magnetic resonance imaging (MRI) studies, free water is defined as water that is in the macroscopic and microscopic pores of bone tissues, and the remaining water, approximately 60–70% in cortical bone, is identified by MR spectroscopy studies as collagen-bound water^11,16. However, this definition does not discriminate between water loosely and tightly bound to collagen, which could be an important factor contributing to bone quality³⁵. Further, an important limitation of these techniques is that they provide little information about mineral-bound water, since the MR signal from mineral-bound water is too short to be identified and quantified. Therefore, this has motivated studies into other nondestructive and noninvasive techniques, such as vibrational spectroscopy (Raman and infrared) to evaluate bone hydration. Spectroscopic techniques have the ability to identify unique absorption bands that can also be used to quantify the amount of collagen, mineral and water in bone, and to further elucidate and classify water components beyond just free and collagen-bound water.

It is important to briefly clarify the use of the terms “loosely” and “tightly” bound water. Here and in other papers^7–10,32, water molecules bound to bone matrix can be classified by the strength of their interaction with bone mineral and proteins. In this case, water molecules can be loosely bound to surface of collagen and mineral (physical adsorption) or tightly bound (trapped) within their structure (chemical adsorption). While loosely bound water can be easily removed by evaporation or lyophilization, tightly bound water requires conditions such as high temperature to be removed. However, it is possible that this terminology can be misleading when vibrational data are discussed, because bond strengths (and masses) determine the position of infrared bands. For this reason, it is important to explicit that the terms “loosely” and “tightly” bound used here refer to the strength by which water molecules are bound to other components of the bone matrix.

Unal et al. published studies based on the use of Raman spectroscopy to probe water components bound to collagen and/or mineral in cortical bone^10,35. Through serial evaporation and D₂O treatments on bovine cortical bone, they identified unique Raman absorption band s that are associated with water loosely and tightly bound to both mineral and collagen, respectively. In a previous study³⁶, unbound water was removed from cortical bovine samples by oven-drying drying samples at 40°C for 48 hours, and bound water was removed by submerging samples in ethanol for 36 hours to completely dehydrate the tissues. There, they demonstrated that Raman-identified water absorption band s correlated to both mechanical and structural bone properties³⁶. They also concluded that Raman-determined unbound water correlated to increasing porosity, and Raman-derived water absorption band s of unbound and bound components correlated significantly to gravimetric derived unbound and bound water content respectively. Further, mineral bound water was primarily associated with bone strength while collagen bound water was associated with bone toughness. Together, those data demonstrate that a vibrational spectroscopic approach is clearly useful for assessment of bone water linked to bone quality. Here, we present a step forward in the study of bone water by adding near infrared spectroscopic analysis to the possible approaches for more specific assessment of water, matrix and mineral in samples of interest.

Studies have previously suggested the NIR absorption band at 7000 cm⁻¹ arose from tightly bound water²¹ associated with mineral¹¹; however, since the absorptivity of this peak is very low, it was not conclusively identified as related to mineral content. The methods used in the present study allowed us to confirm, and quantify, the mineral-associated absorption bands in the NIR spectral region. However, at this point, practical use of this absorption band to reflect mineral content is limited to lyophilized samples. Nonetheless, as shown by the strong correlation between this NIR peak and the MIR phosphate/amide I ratio in serially demineralized samples, this may still be useful as an assessment of bone mineral in dry samples, which could include harvested samples such as bone from animal studies.

Furthermore, calcium phosphates with and without bound water (brushite and monetite, respectively) were used as standards to aid in defining the NIR absorption band at 7000 cm⁻¹ as water tightly bound to the mineral crystal. It is important to mention that bone mineral is highly deficient in OH³³, and a resolved peak attributable to a structural OH absorption band is not present in standard MIR spectra of biological apatite³⁴; therefore, it is reasonable that the corresponding overtone and/or combination peak would not be observed in the NIR region.

It is also interesting to discuss the preliminary results obtained using human cadaveric bone samples. Aging triggers natural degradation of the organic matrix and an increase cortical bone TMD³⁷, which have been associated with a decrease in overall bound bone water³⁸. In addition, one study showed that in rodents, an increase in TMD was associated with an increase in collagen cross-linking and a reduction in collagen bound water¹⁵. The results presented in the current study support these findings, showing a strong positive correlation between age and TMD in female cortical bone samples. In a previous paper¹¹, we showed that both water absorption bands observed at ~7000 and 5200 cm⁻¹ correlated directly with MRI bound water data, and that the ~7000 cm⁻¹ water band was inversely correlated with donor age. Furthermore, here we observed that with increased TMD, NIR quantification of the overall bone water ratioed to collagen matrix (5200/4608) in hydrated samples decreases, whereas in lyophilized samples, the mineral/collagen ratio (7000/4608) increases. Although this data did not reach statistical significance for the small samples size available here (n=6), this can be highlighted as a potential application of NIR-determined water components as an indicator for bone quality assessment.

In addition to this potential medical application, NIR spectroscopic approaches to identify different water components is of great multidisciplinary interest. For example, it can be useful for applications in materials sciences^39,40 and chemical mineralogy, including paleontology^41,42. In particular, an interesting study used NIR spectroscopy to non-destructively evaluate diagenesis in fossils bones⁴¹. The authors described well-defined water absorption bands at 5318 and 7240 cm⁻¹. To evaluate the nature of these water species, they subjected the fossilized bones to ethanol vapors and acquired NIR spectra upon increasing times of exposure. They found that both water absorption bands decreased, indicating that these bands were related to labile water molecules that could be replaced by ethanol. They discussed that these water molecules could be located on the grain surface of the biological apatite, and that they could be important in the understanding of the organic-inorganic interactions in bone. Here, we corroborate that peaks at ~5200 and 7000 cm⁻¹ are related to water molecules associated with bone mineral. Moreover, we found that the peak at ~5200 cm⁻¹ is also related to water tightly bound to collagen. The differential analysis of these absorption bands could be an interesting approach in future studies exploring the diagenesis of water in bone mineral and collagen preserved over geologic time.

It is important to also briefly discuss the comparison of NIR with other vibrational spectroscopy methods. For example, a strength of NIR compared to MIR analysis is the ability to analyze thick samples. MIR studies generally require thin section sample analysis, grinding samples into powders, or surface-only analysis using ATR. NIR radiation has a greater depth of penetration into samples (several millimeters to centimeters) compared to MIR radiation (typically ~ 10 microns maximum), and thus sample preparation for NIR studies can be minimized. Further, the relatively low cost of NIR instrumentation may enable this approach for bone assessment to be widely accessible.

Comparison of Raman and NIR spectroscopy for evaluation of bone water leads to some notable strengths and weaknesses for each modality. Both techniques are relatively fast in data collection and have the ability to collect data from thicker samples. An advantage of NIR spectroscopy is that it is a low-cost technique with readily available commercial instruments, whereas the lack of widespread availability and high-cost of some Raman instrumentation for water assessment may challenge the wide-spread application of this approach. Raman spectra also tends to generate vast amounts of fluorescence, which can interfere with signals produced from the samples. However, in contrast to NIR techniques, Raman spectroscopy has been successful in the identification of mineral absorption bands in hydrated bone tissues³⁶, which has not been possible in NIR so far, including in the present work. It is to be noted that NIR spectra are composed of broad and overlapping peaks that are dominated by water absorption bands, and therefore, it challenging to elucidate mineral-associated water frequencies in hydrated tissues. It is also important to mention that acquiring data transcutaneously from underlying bone tissues is still challenging using both of these techniques, although investigations into such an approach are ongoing using both modalities^43–46.

We believe the primary limitation in this study, and in the use of NIR clinically, is the challenge of identifying water associated with mineral or collagen specifically in hydrated bone samples. NIR sensitivity to water causes the spectra to be dominated by water absorption bands throughout much of the range interest. Therefore, samples need to be lyophilized to elucidate specific water-related mineral absorption bands. Since there was a promising correlation between TMD and water content, a larger sample set could be investigated to better elucidate how NIR spectroscopy may contribute to understanding of age-related, and possibly disease and/or therapy-related, compositional changes in bone tissue.

CONCLUSIONS

The current study demonstrated that NIR spectroscopic analysis of cortical bone can be used to identify water bound to different components: loosely bound to the tissue, tightly bound to either collagen or mineral, and tightly bound primarily to mineral. Compared to MRI studies of bone water^13,38, using a NIR approach allows for a more specific identification and classification of bone water, but it is not yet appropriate for clinical studies. Nonetheless, the results detailed in this study provide a very strong foundation for future work focused on application of NIR spectroscopy in ex vivo studies to aid in elucidation of compositional changes that occur in cortical bone during aging and disease states.