Article pubs.acs.org/ac
Innovative Approach to Investigating the Microstructure of Calcified Tissues Using Specular Reflectance Fourier Transform-Infrared Microspectroscopy and Discriminant Analysis Catherine L. Nicholson,*,† Elwyn C. Firth,†,‡,⊥ Mark R. Waterland,§ Geoffrey Jones,§ Siva Ganesh,§,⊗ and Robert B. Stewart∥ †
Institute of Veterinary, Animal and Biomedical Sciences, ‡National Research Centre for Growth and Development, §Institute of Fundamental Sciences, and ∥Institute of Natural Resources, Massey University, Palmerston North, New Zealand S Supporting Information *
ABSTRACT: Although bone fracture has become a serious global health issue, current clinical assessments of fracture risk based on bone mineral density are unable to accurately predict whether an individual is likely to suffer a fracture. There is increasing recognition that the chemical structure and composition, or microstructure, of mineralized tissues has an important role to play in determining the fracture resistance of bone. The objective of this preliminary study was to evaluate the use of specular reflectance Fourier transform infrared (SR FT-IR) microspectroscopy in conjunction with discriminant analysis as an innovative technique for providing future insights into the origins of orthopedic abnormalities. The impetus for this approach was that SR FT-IR microspectroscopy would offer several advantages over conventional transmission methods. Bone samples were obtained from young racehorses at known fracture predilection sites and spectra were successfully obtained from calcified cartilage and subchondral bone for the first time. By applying discriminant analysis to the spectral data set in biologically relevant regions, microstructural differences between groups of individuals were found to be related to features associated with both the mineral and organic components of the bone. The preliminary findings also suggest that differences in bone microstructure may exist between healthy individuals of the same age, raising important questions around the normal limits of individual variation and whether individuals may be predisposed to later fracture as a result of detrimental microstructural changes during early growth and development.
B
Ca10(PO4)6(OH)2.21,23 However, the mineral found in calcified tissues is typically deficient in calcium compared with stoichiometric hydroxyapatite,22 may have little if any hydroxide, and should be more properly referred to as “biological apatite” to emphasize its variable composition.23 Substitutions involving sodium, potassium, magnesium, carbonate, and hydrogen phosphate are common and highly significant, since they can substantially change properties such as crystallinity, solubility, and hardness.23,24 The organic matrix of bone is composed principally of type I collagen and has a direct effect on bone strength.6,25 The type, age, and maturation of the collagen cross-links are known to be important factors in contributing to reduced mechanical properties during aging and in disease.18,26 Consequently, it is clear that both the mineral and organic components of the bone material should be considered together in any investigation of the relationship between microstructure and fracture resistance.
one is an outstanding example of smart materials design, responding and adapting to a continually changing environment throughout life.1−3 As a heterogeneous, anisotropic, nanocomposite material,4−6 multiple structural and material properties contribute to the fracture resistance of the whole bone.7,8 Such complexity makes predicting fracture susceptibility in individuals extremely difficult,9,10 exemplified by the escalating problem of osteoporosis.11−13 Assessment of bone architecture and mineral mass as indicators of strength has been the focus of much clinical research, but it is acknowledged that fracture risk cannot be accurately predicted based on measurements of bone mineral density alone.14−16 There is increasing recognition that the chemical structure and composition, or microstructure, of bone has a central role to play in influencing many important properties, including fracture resistance.17−22 By understanding how bone strength may be altered at the molecular level before fracture occurs, much earlier interventions and more effective treatment and preventative strategies could be designed. The nanocomposite bone material consists of a mineral phase, organic matrix, and water. The mineral is frequently referred to as “hydroxyapatite”, implying that this is a stoichiometric phase of known composition, specifically © 2012 American Chemical Society
Received: January 17, 2012 Accepted: March 7, 2012 Published: March 7, 2012 3369
dx.doi.org/10.1021/ac300123r | Anal. Chem. 2012, 84, 3369−3375
Analytical Chemistry
Article
principal components analysis, cluster analysis, and partial leastsquares38,39 offer a potentially much more objective, powerful, and informative alternative approach, but to date they have had very limited application to spectroscopic data collected from bone40,41 or cartilage.42,43 Discriminant analysis, based on identifying combinations of variables that can be used to separate members of two or more groups from each other, may be a useful tool for examining microstructural data obtained from bone. This study was designed to investigate whether SR FT-IR microspectroscopy in conjunction with discriminant analysis could be used as an innovative technique to gain new insights into the origins of orthopedic abnormalities. The study was conducted using samples from the third metatarsal bone (Mt3) of nine young Thoroughbred racehorses of different ages and disease status. The medial and lateral parasagittal grooves of the Mt3 are known to be common initiation sites for fracture in these animals.44−46 Although it has been generally believed that the heavy training demands placed on racehorses at a very young age may be largely to blame for later musculoskeletal injuries, there is now evidence to show that morphological abnormalities at the sites of fracture initiation are present before training or racing has even begun.47 This strongly suggests that factors during very early bone growth and development may be responsible for later orthopedic disease and that certain individuals may be inherently predisposed to fracture. As an excellent animal model relevant to human orthopedic health,48,49 the horse may also provide useful insights into human musculoskeletal disorders. The impetus for this study was to develop a novel spectroscopic technique for microstructural analysis of calcified tissues which avoided the inherent difficulty of obtaining reproducibly thin sections while simultaneously enabling morphological investigation by BSEM. The preliminary nature of this work, using a limited number of animals, precluded a detailed investigation of microstructural differences between bone samples from clinically normal and abnormal animals or between control and fracture predilection sites. However, because the samples could be grouped according to age and disease status, it was expected that microstructural differences should be apparent between groups despite the limited number of animals. The aims of this study were therefore (1) to develop and evaluate the potential of SR FT-IR microspectroscopy as a proof-of-concept for determining whether further study using larger numbers of animals was justified and (2) to explore the use of discriminant analysis as a technique for analysis of the resulting spectral data set.
Fourier transform infrared (FT-IR) microspectroscopy is an excellent tool for investigating bone microstructure since both the mineral and matrix can be assessed simultaneously.27,28 While it lends itself to future development for in vivo applications in much the same way as Raman microspectroscopy,27,29 for example, by use of a fiber optic probe during arthroscopic surgery, much fundamental work remains to be carried out. Microstructural changes in bone, such as those that occur during aging or as a result of disease, have been investigated using vibrational spectroscopic techniques.7,30−32 FT-IR spectroscopic studies of bone are usually performed in transmission mode, which requires very thin sections (approximately 5 μm) to be taken from the sample.27 The preparation of calcified tissue sections less than 10 μm in thickness reproducibly is not a routine procedure and may be difficult to accurately achieve. Conventional studies are therefore limited to calculating the ratios of bone constituents such as the mineral/matrix ratio to eliminate errors associated with variations in sample thickness. This requirement for thin sectioning is also a major disadvantage if it is desirable that the sample remains intact, for example, if other characterization techniques are to be applied to the same sample. Furthermore, sample thickness strongly influences observed spectral distortions, although no reported methods currently exist which can account for these in a satisfactory way. Therefore, for a scattering material such as bone, the potential variability in sampling volume is another drawback of thin sections. Specular reflectance FT-IR (SR FT-IR) microspectroscopy is an alternative method to conventional transmission FT-IR microspectroscopy that enables thick, intact samples to be examined. The SR spectrum comprises absorption and dispersion components, the latter being related to the refractive index of the material.33,34 The appearance of an SR spectrum is similar to that of the first derivative of a transmission spectrum, in which an individual peak may have positive and negative peaks relative to the baseline, in contrast with the more familiar Gaussian−Lorentzian peaks typical of transmission spectra. SR spectra differ from transmission spectra in their spectral patterns, peak positions, and intensity ratios and cannot be used for quantitative analyses. However, by applying the mathematical transform known as the Kramers−Kronig transformation to a sample spectrum, SR spectra can be converted to absorbance-like spectra which closely resemble the corresponding transmission spectra.33 The transformed spectra can then be used for qualitative and quantitative analyses. There do not appear to have been any published studies of bone or cartilage using SR FT-IR microspectroscopy, but its application to the investigation of human teeth35,36 and urinary stones37 suggests that it could reasonably be expected to be useful in studying other mineralized tissues. SR FT-IR microspectroscopy could be particularly useful in conjunction with a nonspectroscopic imaging technique such as backscattered electron microscopy (BSEM), since morphological information could be directly correlated with microstructural data in precisely the same location of the sample. Analysis of spectroscopic data sets generated from studies of calcified tissues is conventionally performed using curve fitting procedures and calculating peak heights or areas directly from spectra for determination of mineral and collagen content, crystallinity, collagen maturity, and other parameters of interest.27 The use of curve fitting and other spectral processing procedures may be rather subjective and introduce errors into the resulting data. Multivariate statistical methods such as
■
MATERIALS AND METHODS Bones. The Mt3 was obtained from nine thoroughbred horses of different ages. Four bones were from newborn foals while four were from a cohort of 5 month old foals which had been managed together under identical conditions. All eight bones had been stored in a freezer at −20 °C prior to this study and were from clinically normal animals with no known musculoskeletal abnormalities. One bone was from a 3 year old horse with visible gross abnormalities in both the medial and lateral parasagittal grooves (Figure S-1 in the Supporting Information), i.e., the regions associated with fracture initiation. This sample was supplied prepared, having already been embedded in polymer resin and polished (sample provided by Prof. Alan Boyde, Queen Mary University of London).
3370
dx.doi.org/10.1021/ac300123r | Anal. Chem. 2012, 84, 3369−3375
Analytical Chemistry
Article
Preparation of Bone Sections. A detailed description of the methodology for obtaining bone slices from the distal Mt3 has been previously reported50 and consisted of using a diamond saw to remove a 2 mm thick palmar (posterior) bone slice inclined at approximately 30° to the long axis of the bone. Three 5 mm × 5 mm sections were then cut from each palmar slice to encompass the regions of the lateral and medial parasagittal grooves (areas of fracture predilection) and the medial condylar surface (not associated with fracture). The bone sections were then dehydrated in absolute ethanol. Preparation of Polymer-Embedded Polished Bone Samples. The dehydrated bone sections were placed in small glass vials (Wheaton Industries Inc., Philadelphia) for embedding in polymethylmethacrylate (PMMA). The methylmethacrylate monomer (Alfa Aesar, Lancs, U.K.) was activated by addition of azobisisobutyronitrile initiator (synthesized by Dr. Andrew Kay, IRL, Lower Hutt, NZ), and samples were left in an oven at 40 °C until polymerization was complete. The samples were then removed by simply smashing the glass vials. The highly polished mirror-like surface necessary for sample analysis was achieved using increasingly fine grades of carborundum paper followed by manual polishing with 3 μm Al2O3 powder (Buehler, Illinois) and finally 0.3 μm Al2O3 (Linde A alumina polishing compound, Praxair Electronics, Indianapolis). Backscattered Electron Microscopy. BSEM was performed prior to SR FT-IR microspectroscopy to ensure that no subclinical morphological abnormalities were evident in any of the samples from the foals. This was not performed on the 3 year old horse which was already known to have gross abnormalities in both parasagittal grooves (Figure S-1 in the Supporting Information). The polymer-embedded polished bone sections were mounted on 19 mm aluminum stubs and carbon-coated using a carbon flash coating system. Imaging was carried out using a scanning electron microscope with a BSE detector (FEI Quanta 200, Eindhoven, The Netherlands). SR FT-IR Microspectroscopy. After BSEM imaging, the thin carbon coating on each sample was easily removed by careful application of very fine carborundum paper. The surface was then repolished with the Al2O3 powders before each sample was analyzed by FT-IR microspectroscopy in reflection mode (Nicolet 6700 FT-IR spectrometer coupled to Nicolet Continuum FT-IR Microscope, Thermo Electron Corporation, Madison). Each spectrum was collected within a sample area of 25 μm × 25 μm. A total of 45 spectra were collected in each of the three 5 mm × 5 mm sections and comprised (i) 15 in articular calcified cartilage (ACC) adjacent to the hyaline articular cartilage (HAC), referred to as calcified cartilage row 1 (CC1), (ii) 15 in ACC adjacent to the subchondral bone (SCB), referred to as calcified cartilage row 2 (CC2), and (iii) 15 in SCB adjacent to ACC, referred to as SCB (Figure S-2A,B in the Supporting Information). In the case of the newborn foals, there were difficulties in finding sufficiently large sampling areas and distinguishing the ACC from SCB restricted collection of spectra to CC1 only. In total, 45 spectra were collected for each of the four newborn foals while 135 spectra were obtained for each the four 5 month old foals and the abnormal horse. Spectra were collected over the wavelength range 650−2000 cm−1 with a spectral resolution of 8 cm−1, and each spectrum represented an average of 512 repeated spectra. Background correction was carried out in accordance with conventional reflectance methodology using a gold mirror. Using OMNIC
software (version 7.3, Thermo Electron Corporation, Madison), each SR spectrum was converted to an absorbance-like spectrum by applying the Kramers−Kronig transformation. Spectra were normalized to 1 absorbance unit, and a linear baseline correction was performed. Reference spectra were obtained for biological apatite obtained from sintered bovine cortical bone denuded of collagen by pyrolysis (provided by Dr. Michael Mucalo, University of Waikato, NZ), type I collagen from rat tail (Sigma, product number C7661), and pure PMMA. The reference materials were prepared in an identical fashion to the samples by embedding in PMMA and polishing. Discriminant Analysis. Discriminant analysis was carried out to identify the metrics, i.e., variables (wavenumbers), that could best discriminate members of known groups from each other. Group classification was defined on the basis of animal age and clinical status (newborn/normal, 5 month old/normal, 3 year old/abnormal). The analysis was initially performed on the entire data set which consisted of 855 observations and 350 metrics. Subsequent analyses then focused on subsets of the data. The subsetting was carried out by restricting the wavelength range of interest to focus on biologically relevant spectral regions, i.e., those relating to the bone mineral or collagenous matrix (Table 1). In this case, the number of Table 1. Spectral Band Assignments for Biologically Significant Regions in FT-IR Spectraa biological significance
wavelength region (cm−1) [transmission FT-IR]27
wavelength region (cm−1) [SR FTIR]34,35
band assignment27,34,35
carbonate I (mineral) phosphate (mineral)
850−890
850−890
ν2 CO32−
900−1200
980−1200
amide III (collagen matrix) amide II (collagen matrix) amide I (collagen matrix)
1200−1300
1200−1350
ν 1 + ν 3 PO43− ν 2− 3 HPO4 C−N stretch N− H bend
1500−1600
1500−1600
C−N stretch N− H bend
1585−1720
1600−1700
CO stretch
a
Data are presented from transmission studies of bone and cartilage and specular reflectance studies of human dentin.
metrics included in each subset was reduced by manually choosing the wavelength regions over which the analysis was performed. This was justified on the basis that the regions with no absorbance clearly make no contribution to analysis of the data, and regions of no known biological significance cannot be interpreted biochemically.39,51 A further subsetting was used to examine individual sections (lateral parasagittal groove, medial parasagittal groove, medial condylar surface). Discriminant analysis was carried out using the program R [R Foundation for Statistical Computing, Vienna, Austria, http://www.r-project. org].
■
RESULTS
BSEM Analyses. No morphological abnormalities were observed in the ACC or SCB of any of the samples from the foals. Representative images for a newborn and 5 month old foal are shown in Figure S-2A,B in the Supporting Information, respectively. The very “open” appearance of the distal Mt3 soon after birth was quite different to its more typical morphology 3371
dx.doi.org/10.1021/ac300123r | Anal. Chem. 2012, 84, 3369−3375
Analytical Chemistry
Article
only a few months later. In the newborn foals, there was no clear boundary between the ACC and SCB. SR FT-IR Microspectroscopy. The typical appearance of the derivative-like SR FT-IR spectra before and the absorbancelike spectra after the Kramers−Kronig transformation was applied is illustrated in Figure S-3 in the Supporting Information. The SR FT-IR spectra for the reference materials (biological apatite and collagen) and embedding matrix (PMMA) are shown in Figure 1. These were used for band
prominent, the mineral peak much reduced in intensity, and the shapes of these peaks somewhat different for this individual. For illustrative purposes, Figure S-5 in the Supporting Information shows a comparison between a bone sample collected in reflectance mode and a different sample of approximately 10 μm thickness prepared by cryosectioning and collected in transmission mode. Discriminant Analysis. A discriminant analysis plot of the entire data set is shown in Figure 3. Samples were separated
Figure 1. Reference SR FT-IR spectra of biological apatite from sintered bovine cortical bone denuded of collagen (BAP), type I collagen from rat tail, and pure polymethylmethacrylate embedding matrix (PMMA). The phosphate peak arises from the inorganic mineral phase of calcified cartilage and subchondral bone while the amide regions arise from peptide bonding and cross-linking in the collagenous matrix.
Figure 3. Discriminant analysis plot of the full data set illustrating clear separation of samples into three distinct groups. The abnormal 3 year old horse is distinguished from the other animals by the first linear discriminant function (LD1), while the second linear discriminant function (LD2) separates the newborn foals from the older animals.
assignment of sample spectra in conjunction with published data acquired in both transmission and reflectance modes (Table 1). Sample spectra from all animals, sections, and rows were broadly similar in appearance (representative spectra are shown for the abnormal horse in Figure S-4A−C in the Supporting Information) with the exception of one particular case. Figure 2 shows that one of the 5 month old foals (horse no. 2) was clearly different from the others of the same age in CC1 spectra of the lateral parasagittal groove. In contrast to every other spectrum recorded, the collagen peaks were very
into three distinct groups, namely, newborn foals, 5 month old foals and the abnormal horse. The first (LD1) and second (LD2) linear discriminant functions explained 49% and 37% of the difference between the groups, respectively. LD1 separated the abnormal horse from the other animals, while LD2 separated the newborns from the other animals. Results of discriminant analysis are shown in Figure 4a,b for the phosphate and amide I subsets, respectively. No separation of the groups was observed in the carbonate I, amide II, or amide III regions. Further subsetting of the data to examine the three sections in each spectral region produced essentially the same results except that for the lateral parasagittal groove, the amide I region showed one of the 5 month old foals (horse no. 2) separated from the other clinically normal foals (Figure 5).
■
DISCUSSION While the 3 year old horse was known to have abnormalities in both parasagittal grooves, BSEM imaging of the distal Mt3 bone of the newborn and 5 month old foals did not reveal any morphological features indicative of abnormality in these animals. This was important to confirm, as defects in the ACC and SCB in the region of the parasagittal grooves have been reported in 2 year old thoroughbred horses with no clinical signs of musculoskeletal abnormality.46,52 Fatigue fractures of the Mt3 are known to initiate in the ACC of the lateral or medial parasagittal groove before propagating into the subchondral bone and then into the bulk epiphyseal trabecular bone.47 Using BSEM, changes in the mineralized tissue at the site of origin of these fractures have revealed abnormalities in
Figure 2. SR FT-IR spectra of the 5 month old foals in the articular calcified cartilage adjacent to the hyaline articular cartilage in the lateral parasagittal groove. Horse no. 2 is strikingly different from the others in having much less mineral (black arrow) relative to collagen (red arrows). The arrowed peaks for this horse are also somewhat different in shape to the other animals. 3372
dx.doi.org/10.1021/ac300123r | Anal. Chem. 2012, 84, 3369−3375
Analytical Chemistry
Article
Figure 4. Discriminant analysis plots of subsets of the full data set representing (a) phosphate and (b) amide I spectral regions for the first two linear discriminant (LD) functions. The phosphate region was best able to discriminate between the newborns, 5 month old foals, and the abnormal 3 year old horse while the amide I region separated the abnormal animal from the others. Horse IDs: 1−4 = 5 month old foals, 5 = abnormal 3 year old horse, 6−9 = newborn foals.
microstructure of these tissues. Interpretation of the spectra was straightforward, since reference specular reflectance spectra were obtained for biological apatite and collagen, published reflectance data were available for apatite and collagen from teeth, and because the reflectance spectra were almost identical to conventional transmission spectra available for cartilage and bone once the Kramers−Kronig transformation had been applied. These factors allowed unequivocal band assignments to be made in the biologically significant regions of the spectrum, as summarized in Table 1. In common with transmission FT-IR spectroscopy, the phosphate and carbonate bands in the SR FT-IR spectra relate to the amount and composition of the bone mineral present in the sample, while the amide bands relate to the amount and properties of the collagenous matrix.27 Ideally, validation of the SR FT-IR microspectroscopy technique should be demonstrated by comparison to transmission FT-IR microspectroscopy, which is considered the current “gold standard” for spectroscopic analysis of calcified materials. However, there are three reasons which preclude a direct comparison of transmission and reflectance data. First, data are collected from the sample volume in transmission mode but from the sample surface in reflectance mode. A quantitative comparison cannot therefore be made between the two techniques and this is a particularly important point for the analysis of a material as heterogeneous and anisotropic as bone. Second, reflectance spectra have a derivative-like shape and while the Kramers−Kronig transformation produces spectra that are visually similar to true absorbance spectra, the peaks are slightly different in shape and position. Consequently, reflectance spectra from samples can only be directly compared with other samples and references collected and corrected in exactly the same way. Third, we were unable to collect transmission data on these samples because, in our experience, it proved impossible to cut sections
