We have used Fourier transform infrared spectroscopy (FTIR) to characterise the chemical and structural composition of the tendons of the rotator cuff and to identify structural differences among anatomically distinct tears. Such information may help to identify biomarkers of tears and to provide insight into the rates of healing of different sizes of tear. The infrared spectra of 81 partial, small, medium, large and massive tears were measured using FTIR and compared with 11 uninjured control tendons. All the spectra were classified using standard techniques of multivariate analysis.
FTIR readily differentiates between normal and torn tendons, and different sizes of tear. We identified the key discriminating molecules and spectra altered in torn tendons to be carbohydrates/phospholipids (1030 cm−1 to 1200 cm−1), collagen (1300 cm−1 to 1700 cm−1 and 3000 cm−1 to 3350 cm−1) and lipids (2800 cm−1 to 3000 cm−1).
Our study has shown that FTIR spectroscopy can identify tears of the rotator cuff of varying size based upon distinguishable chemical and structural features. The onset of a tear is mainly associated with altered structural arrangements of collagen, with changes in lipids and carbohydrates. The approach described is rapid and has the potential to be used peri-operatively to determine the quality of the tendon and the extent of the disease, thus guiding surgical repair.
Tears of the rotator cuff cause pain and functional restriction in up to 30% of adults.1 The financial cost is vast, due to the length of time away from work2 and the annual healthcare costs, which together are estimated to be $3 billion in the United States alone.3 Currently, there is no definitive treatment for the tears.4 Conservative management including physiotherapy, and the administration of non-steroidal anti-inflammatory agents or glucocorticoids addresses the pain rather than modifying the disease, and surgical repairs are associated with high rates of re-rupture and poor outcome.5 There is therefore a need for a better understanding of the pathophysiology of tears of the rotator cuff. Larger tears have higher failure rates.6,7 Disease biomarkers need to be identified which may help to indicate the early stages of the disease, monitor its progression and predict the response to treatment.
The aetiology of a tear includes mechanical and biochemical factors. The extracellular matrix of a tendon helps to transmit tensile and shear strength and to dictate its mechanical properties. Changes in the extracellular matrix of torn tendons of the rotator cuff have been demonstrated previously. These include an increase in collagen I and III, alteration of matrix metalloproteinases (MMPs) and an accumulation of glycosaminoglycans and lipids.8,9 Torn tendons have been shown to have an altered chemical composition with varying levels of proteoglycans,10 cytokines,11 oxygen free radicals,12 atrophy13 and chondrocyte markers.14 There is increasing evidence to support the concept of a progressive multistage disease process in which small tears have different characteristics from larger ones. An increase in the size of a tear has been shown to correlate with changes in the composition of the cellular and extracellular matrix, a reduction in metabolism and presumed viability15 and an increase in apoptosis.16 The implications of these cellular and extracellular matrix changes are unclear.
We used Fourier transform infrared spectroscopy (FTIR) to characterise the chemical and structural composition of tears of the rotator cuff of varying size. All the molecules can be excited to higher vibrational states using light at specific wavelengths which correspond to the frequency of excited vibration modes.17 This information can be used to map the absorption positions and help identify the chemical properties of the tissue. FTIR spectroscopy can provide highly sensitive, unique infrared chemical fingerprints of specimens, which can yield new insights into compositional changes in tears of the rotator cuff. A key advantage of FTIR is that it is a quick, non-manipulative and non-destructive test which can identify a wide range of chemical targets and can possibly be used within the operating room. The biomedical applications of FTIR are numerous, including mapping of the collagen and proteoglycan content of cartilage and the study of various tissues such as brain, breast, bone, cartilage, heart, skin and human mesenchymal stem cells.18–20 In order to overcome confounding variables of the thickness and lack of transparency of the specimens, we used attenuated total reflectance (ATR) FTIR for our study.
Our hypothesis was that torn tendons would have an altered chemical composition and FTIR spectra, and that partial and smaller tears could be chemically differentiated from larger tears by their infrared fingerprint.
Materials and Methods
Collection of specimens.
We studied 92 specimens. These were collected from 91 patients undergoing surgical repair of a defect in the rotator cuff, which had been identified pre-operatively by ultrasound. The mean age of the patients was 65.7 years (45.0 to 89.0). The specimens were obtained from consecutive patients who had been recruited into the United Kingdom Rotator Cuff Trial in which they were randomised into the operative repair arm for either open or arthroscopic repair over a period of 18 months. The inclusion and exclusion criteria are outlined in Table I⇓. All tears occurred after the age of 45 years and a study by Nobuhara, Hata and Komai21 found that 95% of their patients with tears of the rotator cuff were aged 45 years or older.
The size of the tear was measured intra-operatively at its widest diameter and biopsy specimens were taken from its edge. The tears were categorised according to the classification suggested by Post, Silver and Singh22 as partial (n = 7), small (n = 16), medium (n = 23), large (n = 17) and massive (n = 18). There was no significant difference in the duration of symptoms in the different groups. A control group consisted of 11 patients, matched for age and gender, undergoing other types of shoulder surgery with no evidence of disease of the rotator cuff. Their mean age was 58 years (46 to 79).
The specimens were immediately placed in formalin for preservation or freshly frozen by placing them into liquid nitrogen and storing them at −80°C. Tendons which had not been treated in formalin were included in the study as a control group in order to assess the effect of formalin and thus verify the robustness of our vibrational spectroscopy classification. Two separate punch biopsies 3 mm in diameter were taken from each patient. Approval for the study was obtained from our institutional review board and all patients gave informed consent.
Fourier transform infrared spectroscopy.
We used a Nicolet 6700 spectrometer (Thermo Fisher, Madison, Wisconsin) with an ATR single-bounce diamond accessory and a liquid-nitrogen-cooled detector. Each specimen was orientated along the long axis of the collagen fibres. A constant pressure was applied to the sample to obtain a consistent absolute absorbance. The diamond internal reflectance element has a refractive index of 2.417, allowing an angle of incidence of 45°. The depth of penetration of an evanescent wave for the diamond is about 1.0 μm to 1.5 μm. Spectral information was collected using OMNIC 7.6 software (Thermo Fisher Scientific, Worcester, Massachusetts) in the spectral region ranging from 400 cm−1 to 7000 cm−1, at a resolution of 4 cm−1. Each spectrum represented a mean of 64 scans taken per specimen, with two spectral readings taken from each separate punch. The results for each patient were averaged. The formalin buffer was used as background spectrum. It was collected before each sample reading and subtracted from each tendon spectrum.
Pure reference compounds of common extracellular matrix molecules, including collagen I, II and III, chondroitin sulphate, decorin and elastin, were obtained from Sigma Aldrich (St. Louis, Missouri). FTIR spectra were collected for six samples of each pure specimen as outlined above and averaged to assist with vibrational band assignment. Chondroitin sulphate is the most common proteoglycan found in tendons of the rotator cuff and decorin is a common glycosaminoglycan.10
All the data were run on a Matlab (Mathworks, Natick, Massachusetts) and Pychem multivariate analysis software package,23 and the data and mathematical modelling were validated during the analysis. The data window was reduced to 700 cm−1 to 4000 cm−1 since the near infrared region (> 4000 cm−1) is very weak and has complex biologically relevant band assignments which are harder to interpret. In order to remove the contribution of atmospheric CO2, the corresponding spectral region of 2250 cm−1 to 2420 cm−1 was replaced by a linear trend. An ATR correction was applied to address the wavelength dependency of penetration. In order to ensure accurate quantitative analysis, the data were smoothed, normalised, variance-scaled and derivatised. The spectra were reduced and classified using standard multivariate analysis to determine the spectral features which differentiated between different states of disease. Similar approaches have previously been described in human studies using FTIR.24,25 Since the wave numbers relevant to biological materials were not all independent, principal component analysis (PCA) allowed reduction of data co-linearity and dimensionality to observe whether the data clustered naturally and to identify possible outliers. Partial least squares (PLS)26 highlighted the most salient spectral features which differentiated between normal and different types of tear and helped to predict the categories of tear which the spectra would fall into. Discriminant function analyses (DFA) of the principal components was used to train a supervised classification model.25 These were plotted using a hierarchical cluster analysis (HCA) to produce a classification tree. The multivariate analyses were cross-validated by randomly splitting the samples into train, test and validation sets and ensuring that the data modelling was accurate. In order to identify the chemical components which optimally distinguished different tendon tears, we used a genetic algorithm.23 The obtained wave-numbers were then cross-referenced with tables listing previously published wave-numbers for biologically relevant materials and with the spectra collected from the pure model compounds. Finally, the multivariate results were statistically tested using a general linear model (GLM) which encompassed linear regression, analysis of variance and analysis of covariance. This approach provided a robust test for multivariate data.27
Samples of the tendon used for FTIR analysis were embedded in paraffin and sections 10 μm in size were cut perpendicular to the long axis of the collagen fibres using a Leica-LM microtome (Leica Microsystems, Wetzlar, Germany). Histological assessment of two random sections from each sample was performed after staining with haematoxylin and eosin and Alcian Blue. The worst affected areas were selected to assess the basic collagen structure as were the areas which were rich in glycosaminoglycans.28 The sections stained with haemotoxylin and eosin were classified using Riley’s score,29 which describes the organisation, orientation and quality of the collagen fibres. Grade 1 describes normal tendons. Grade 2 shows mild degeneration with patchy waviness of collagen fibres and shorter nuclei with an Indian file alignment. Increased collagen hyalinisation and loss of collagen orientation are assigned as grade 3, moderate degeneration. Severe degeneration is designated as grade 4 and describes diffuse hyalinisation with complete loss of orientation of the collagen. Glycosaminoglycans were assessed using the Movin score29,30 based on Alcian Blue staining and classified as 0 (normal), 1 (slightly abnormal), 2 (abnormal) and 3 (markedly abnormal).29 Two separate reviewers (SC, MB) examined and graded the specimens independently and were blinded to the type of tear.
The mean spectra of each tendon, in the mid-infrared region of interest, are shown in Figure 1⇓. Even without any data modelling, it is possible to see the separation of normal tissue and different sizes of tear using FTIR, which suggests that they contain chemically distinguishable groups. Figure 1⇓ indicates that the spectra show no peak shift and that the main changes are in the relative intensity of the spectral peaks. This suggests that there is a change in the relative amounts of chemical groups in the different types of tear.
The changes in the different types of tear were examined by comparing difference spectra in which the spectrum of the normal tendon was substracted from the different groups of tear spectra (Fig. 2⇓). Typically, if no changes were observed between groups then the difference spectrum would show a flat line at zero ▵Abs, as seen between the wavelength of 3700 cm−1 and 4000 cm−1 in Figure 2⇓. We observe, however, a series of positive (1 and 11) and negative peaks (2 to 10), with the positive peaks accounting for the presence of new compounds, and negative peaks for disappearing compounds. Table II⇓ summarises the peaks found from the difference analysis in Figure 2⇓ with corresponding band assignments to indicate the chemical changes which are occurring. Negative peaks dominated below 1800 cm−1 (lines to 2 to 6, amide bands) suggesting a loss or change of chemical and structural composition compared with the normal tendon. However, one positive peak dominated the spectral region centred on 1000 cm−1 (peak 1) which was strong only within the partial tendon (spectrum A in Figure 2⇓). Above 2700 cm−1 more random behaviour is seen with changes in lipids (peaks 7 and 8), collagen (peak 9, amide B band) and hydroxyl groups which reflect the formalin signal (peaks 10 and 11). The PCA of each of the different groups of tear showed that the first two principal components could explain at least 87.4% of the variance.
Normal and torn tendons can be readily differentiated (Fig. 3⇓). In order to determine the degree of relatedness of the different diseased tendons, we used DFA-HCA analysis. Figure 3⇓ shows the results of the classification analysis in the form of a hierarchical tree. The Euclidean distance represents the degree of relatedness of the samples. The classification within the tree indicates that normal tendons are a separate chemical entity from torn tendons. Chemically, the torn groups can be classed into a) partial and small tears and b) medium, large and massive tears. Partial tears are chemically distinguishable from normal and small torn tendons, confirming that they are, biologically, a separate entity, even although they do share more similar chemical features with small tears. Large and massive tears are more related to each other than to medium tears, and the high similarity suggests a large chemical overlap. Although informative, the tree does not provide any information about the basis of the differences seen between the types of disease.
The DFA analysis suggested that the following key spectral regions and their corresponding biological compounds accounted for most of the differences between normal and torn tendons: carbohydrates, phospholipids (1030 cm−1 to 1200 cm−1); collagen structures (1300 cm−1 to 1700 cm−1, 3000 cm−1 to 3350 cm−1); and lipids (2800 cm−1 to 3000 cm−1).
Further testing of differences between the groups was undertaken using a GLM analysis, with age as a covariate and gender as a random factor. We found that neither age nor gender varied significantly among the groups. Our study encompassed a reasonably wide distribution of age of patients, but most importantly the age profile was similar and matched among the different groups and hence changes between the groups were unlikely to be age-related.
In order to identify the specific chemical basis of discriminating variables and the chemical evolution in the different groups of tear, we used a genetic algorithm to identify the significant wave-numbers which accounted for most variability between groups (Table III⇓), and cross-referenced them with known wave-numbers for biologically relevant materials31 (Table II⇑). Collagen I, II and III decreased in the torn tendons, collagen II most noticeably and collagen III to the least extent. There was a progressive decrease in collagens I and II, with a smaller decrease in collagen III moving from normal, to partial, to small and then medium tears. The greatest decrease was evident in collagen II, with higher levels in massive tears compared with small tears. Large and massive tears had higher levels of collagen I than small and medium tears. Elastin increased in small to massive tears, with no change in partial tears when compared with normal tissue. The onset of partial tears involves only small structural changes in collagen, particularly in collagen II. This may be due either to degradation of collagen I into other types of collagen or to cross-linking. Small tears mostly involve changes in lipids, and to some extent sugars such as proteoglycans. For medium tears, the chemical groups which appear to be most affected are collagen structure and sugars, suggesting a change in the proteoglycans. Most chemical groups are altered in large tears, whereas massive tears show the greatest change in collagen, with some proteoglycan. The exact proteoglycans involved are unclear since no difference in chondroitin sulphate or decorin was detected between groups.
Using formalin to preserve specimens did not prevent the ascertainment of differentiating chemical features, but differentiation was more apparent amongst the fresh groups. We found, counter-intuitively, that the formalin levelling effect was more apparent in the larger tears, suggesting cross-linking was dependent upon the size of the tendon.
Generally, increased tear size resulted in higher Riley histological grades,29 indicating increased disruption of the structure and organisation of collagen fibres (Table IV⇓). Increasing tear size was also associated with higher Movin scores30 and showed increased abundance and decreased structural integrity of proteoglycans.
The use of FTIR allowed us to show that normal tissue and tears of the rotator cuff of varying size have different and distinguishable chemical properties. This is the first study, to our knowledge, which has used FTIR to assess the chemical composition of such tears. The biological changes associated with tears are a critical element in both their healing and propagation. We offer some insight into the chemical alterations which may underlie the structural and mechanical changes that predispose to failure of the tendon and the inability to withstand loads even after surgical repair. Our results suggest that the onset and progression of these tears primarily involves the disruption and alteration of the conformation in the structural arrangements of collagen with associated changes in proteoglycans and elastin. We cannot be certain whether the biological changes precede the tear, or are a reactive change to them, but it is plausible that some of the extracellular changes in the matrix will have occurred in response to mechanical or biological insults, and thus weaken the integrity of the structure of the collagen-matrix lattice, making tendons more prone to rupture. The structural changes suggested by the FTIR correlated with the histological analysis, showing a progressive alteration in the structure and organisation of collagen and proteoglycan.
The collagen spectra primarily correlates with vibrations in its peptide band which arise from amide A, I, II and III.32 Our study has identified structural changes in these peptides in torn tendons. Since collagen fibres are predominantly responsible for the tensile strength of tendons, and the structure of the collagen correlates with the mechanical parameters, the structural changes are likely to translate into decreased tensile strength.33 The intermolecular cross-linkages contribute to various mechanical parameters such as tensile strength and viscoelasticity. Thus alterations in collagen structure in torn tendons may disrupt these cross-linkages and thus impair the mechanics. This may account for the weakness and the inability to perform household tasks after development of tears of the rotator cuff,34 and the improvement in function35 and strength36 seen after successful repair of the tendon.
A reduction of collagen types I, II and III was detected. Previous studies have also described decreased collagen levels37 and this may predispose to rupture since a correlation has been shown between the density of collagen and the strength of the tendon.38 Many studies examining collagen subtypes have studied a much earlier time period since the tears, unlike our study in which the mean duration of symptoms was for one year. Using a rabbit model, Hirose et al39 showed increased type-I collagen in the edges of the tear for three weeks only. Increased mRNA levels of collagen I precursors were found in full-thickness tears for up to four months after injury.40 Our findings are supported by the observations of Riley et al,37 who also detected a reduction in collagen III in torn human rotator cuffs. Increased collagen III has been detected in only the first week of healing.39 It has been proposed that in scar tissue type-III collagen is replaced by type-I collagen during maturation,41 and any reduction of collagen subtypes in the torn tendons may reflect decreased healing and scar formation. Increased levels of type-III collagen have been identified at the sites of insertion of the rotator cuff in dogs which coincide with areas of greater stress and compression.42 Type-II collagen is less commonly found, but is associated with the fibrocartilage at the osteotendinous junction and in areas of the tendon undergoing compression.9 The absence of load and compression in the chronic tears in our study may explain the reduction in types-II and -III collagen and the correlation with size, since larger tears may transmit lower forces. Many other collagen subtypes have been identified, but are thought to be less important for the structure and function of the tendon.8
Our FTIR study showed changes in proteoglycans, which were confirmed by the histological findings. It is possible that changes in the extracellular matrix may be the initial insult or event which reduces the ability of tendons to withstand tension and transmit loads at the collagen-matrix interface, thus predisposing to rupture. It has previously been proposed that degeneration of the tendon occurs when its elastic capability is overcome by recurrent minor strains.43 The matrix within which the collagen fibres are embedded allows one collage subfibre to interact with another, increases the stiffness against bending and transmits shear forces at the fibre-matrix interface.44 Alterations in proteoglycans are likely to have important consequences since they are thought to play a role in resisting compressive forces. Altered levels of MMPs and tissue inhibitors of MMP9 have been shown in torn tendons. Inhibition of MMPs was shown to improve histological healing, but did not influence the mechanical properties.45 While the tendon of supraspinatus has high levels of chondroitin sulphate and decorin,10 no change in either was detected in the groups. Thus changes to other chemical components may be more critical in the development and differentiation between different sizes of tear. Our findings of alterations in the lipid composition and increased elastin are supported by previous findings showing fatty infiltration in degeneration of the rotator cuff, particularly in larger tears.46
The structural changes in the collagen and extracellular matrix suggested by our study may reflect an imbalance between synthesis and degradation. The tendon of supraspinatus usually has a higher rate of collagen turnover than other tendons such as that of biceps.47 Therefore any disruption in the turnover is likely to affect the control of collagen and proteoglycan.48 Propagation of a tear may be encouraged by the presence of fewer cells at the edges of tears with altered metabolism,15 which are less able to synthesise and to degrade the extracellular matrix in order to repair the structural damage. Importantly, a number of previous studies have indicated that larger tears are more likely to re-rupture,6 and it is possible that biochemical and cellular changes may underlie this. Smaller tears have been shown to have higher rates of healing than larger tears postoperatively.49,50 The edges of smaller tears have been shown to have higher metabolic rates and presumed viability,51 as well as increased cellular activity, in comparison with larger tears15 which could explain why smaller tears have a better ability to repair and to heal.
Further investigation will be required to examine whether biochemical alterations may result in altered rates of healing and/or higher rates of failure. Our study has provided extensive qualitative data, but was limited in its ability to extract quantitative data. Another limitation is that FTIR lacks specificity for many types of protein particularly proteoglycans and collagen, because of the strong spectral overlap of some of their common features, which can make differentiation between the two groups more difficult. Thus ratios of peak intensities alone cannot be used to assess the composition of the extracellular matrix. Infrared irradiation has a limited penetration of only a few tens of micrometres, limiting it to assessment of superficial surface tissue and possibly missing biochemical variations at greater depth. The spectra for each of the different sizes of tear were not completely distinct and some overlap in the different groups was observed. It is possible that there is a continuum of underlying chemical changes and hence each of the sizes is not completely or reliably distinguishable chemically. Some error may have been introduced during the classification of the size of the tear since they are three-dimensional entries which are sometimes judged in a two-dimensional plane when repaired arthroscopically. Because of the relatively small number of surgical repairs of partial tears, there was a smaller sample size of seven partial tears compared with the larger number in other categories. Preservation in formalin was shown to have a levelling effect on spectral readings, but nonetheless did not prevent the differentiation of different groups. The mean age of the patients with normal tendons was slightly lower (59 years) than that of those with torn tendons which could have been a confounding variable. Age did not vary significantly among the different tear groups.
Our cross-sectional study gives insight into the disease process at a single time-point. Further longitudinal studies would provide some information on the progression of disease and the underlying chemical changes. These may help to identify disease earlier and to monitor the response to treatment and progression of the disease, possibly from biopsies taken in an outpatient clinic. To date the earliest detection of tears of the rotator cuff has relied upon imaging which shows gross anatomical tears. There is currently no method for identifying early extracellular biochemical changes in situ. Since larger tears are associated with higher rates of failure, there is a need for earlier diagnosis and intervention. Analysis of the composition of tendon tissue using this technique was approached with no presuppositions about the likely chemical content, and theoretically all chemical components of the tissue were being accessed. FTIR is advantageous in that it is non-destructive, reagent-free and relatively quick. It has the potential to be used per-operatively with an FTIR probe52 to determine the quality of the tendon and to demarcate the extent of disease, thus guiding surgical repair. The technique could help to confirm further these biochemical changes. Our study has focused on changes at the edge of the tendon of the rotator cuff and it may be useful to see how far back within the tendon these changes extend. In the future a larger study combining both unfixed biopsy samples and, ideally, measurements in vivo, would help further to delineate chemical changes involved in these tears.
Understanding the possible differences between different sizes of tear is an important issue, and our study has shown significant chemical differences in tissues which require repair. These changes may play a role in the biological response to any form of treatment or intervention. We believe that our study has been of value in establishing FTIR as a tool which can be used to differentiate the chemical features of tears of the rotator cuff.
No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.
This research was funded by the NIHR Musculoskeletal Biomedical Research Unit and the Jean Shanks Foundation.
- Received July 1, 2010.
- Accepted November 16, 2010.
- © 2011 British Editorial Society of Bone and Joint Surgery