|
|
Proc Natl Acad Sci U S A. 2006 December 19; 103(51): 19255–19260. Published online 2006 December 8. doi: 10.1073/pnas.0606406103. | PMCID: PMC1748213 |
Copyright © 2006 by The National Academy of Sciences of the USA Engineering, Biophysics Analysis of cartilage matrix fixed charge density and three-dimensional morphology via contrast-enhanced microcomputed tomography Ashley W. Palmer, Robert E. Guldberg, and Marc E. Levenston* George W. Woodruff School of Mechanical Engineering and Parker H. Petit Institute for Bioengineering and Bioscience, Georgia Institute of Technology, Atlanta, GA 30332 Received July 27, 2006. Small animal models of osteoarthritis are often used for evaluating the efficacy of pharmacologic treatments and cartilage repair strategies, but noninvasive techniques capable of monitoring matrix-level changes are limited by the joint size and the low radiopacity of soft tissues. Here we present a technique for the noninvasive imaging of cartilage at micrometer-level resolution based on detecting the equilibrium partitioning of an ionic contrast agent via microcomputed tomography. The approach exploits electrochemical interactions between the molecular charges present in the cartilage matrix and an ionic contrast agent, resulting in a nonuniform equilibrium partitioning of the ionic contrast agent reflecting the proteoglycan distribution. In an in vitro model of cartilage degeneration we observed changes in x-ray attenuation magnitude and distribution consistent with biochemical and histological analyses of sulfated glycosaminoglycans, and x-ray attenuation was found to be a strong predictor of sulfated glycosaminoglycan density. Equilibration with the contrast agent also permits direct in situ visualization and quantification of cartilage surface morphology. Equilibrium partitioning of an ionic contrast agent via microcomputed tomography thus provides a powerful approach to quantitatively assess 3D cartilage composition and morphology for studies of cartilage degradation and repair. Keywords: noninvasive imaging, proteoglycans, cartilage degeneration, osteoarthritis Analysis of small-animal models is limited by the availability of quantitative evaluation techniques for studying the extracellular matrix (ECM) changes associated with osteoarthritis (OA) and cartilage repair. Histology is traditionally used to monitor the spatial distribution of matrix macromolecules but is time-consuming and subject to distortion artifacts and tissue damage, and it produces only semiquantitative analysis of 2D sections that may provide inaccurate 3D representations. Biochemical assays are available to quantify the amount and type of matrix macromolecules in cartilage, but these assays fail to provide their spatial distributions, particularly in small animals where the limited thickness and volume of cartilage make it difficult or impossible to extract samples from multiple regions. Additionally, longitudinal monitoring of changes with time are impossible because of the destructive nature of these histological and biochemical techniques. Proteoglycans (PGs) are a particularly appropriate target for studying OA and for evaluating the efficacy of cartilage defect repair. PGs comprise 5–10% of articular cartilage by wet mass ( 1) and are key regulators of its equilibrium and dynamic mechanical properties. This regulation is the result of interactions between ionic interstitial fluid and negatively charged sulfated glycosaminoglycans (sGAGs) attached to the PG backbone ( 2). The amount and distribution of PGs changes substantially during development ( 3), during degeneration and repair ( 4, 5), and in response to blunt trauma ( 6). Of particular clinical interest, initial stages of OA are marked by aggrecanase-mediated cleavage of PGs ( 7), resulting in altered sulfation patterns ( 8) and progressive depletion of PGs from the ECM. Furthermore, changes in the ECM composition of articular cartilage precede and have been linked to changes in mechanical properties ( 9, 10). Therefore, the ability to noninvasively detect changes in PG content could provide an indirect indication of the integrity of the cartilage matrix and a means to monitor disease progression or reversal. Here we report the development of a contrast-based microcomputed tomography (μCT) technique for high-resolution imaging of PGs in soft tissues. μCT is an x-ray-based imaging modality capable of 3D, quantitative morphological analysis, and current systems provide images at micrometer-level voxel resolutions. μCT imaging has traditionally been applied to hard tissues and is the standard for quantifying trabecular bone microstructural changes associated with aging and osteoporosis ( 11, 12). However, μCT has not been useful for direct imaging of soft tissues because of the low x-ray absorption of nonmineralized tissues. To compensate for the poor radiopacity of soft tissues we have developed a contrast-enhanced μCT technique in which samples are equilibrated in a solution containing an ionic computed tomography (CT) contrast agent before μCT scanning, yielding an equilibrium distribution of the ionic contrast agent that is inversely related to the density of the negatively charged sGAGs. In monitoring PG content we rely on detection of the equilibrium partitioning of an ionic contrast agent via μCT (EPIC-μCT). In the studies presented here we used Hexabrix 320, a clinically available CT contrast agent containing ioxaglate, a negatively charged hexaiodinated dimer. To demonstrate the potential of this approach to examine the ECM of cartilage we first present data validating the ability of EPIC-μCT to monitor PG content and distribution in an in vitro model of IL-1-induced cartilage degradation, a common model for studying cell-mediated cartilage catabolism ( 13, 14). The data indicate consistent agreement with traditional biochemical and histological analysis of PGs and demonstrate a strong correlation between x-ray attenuation and PG content. Second, we present data illustrating the potential of EPIC-μCT for 3D morphometric analyses of articular cartilage in situ on a rabbit femur. The EPIC-μCT images reproduce visible features of articular surface topography and ligament insertions, and segmentation of cartilage from the underlying bone allows quantitative analysis of cartilage morphology. In Vitro Cartilage Degeneration Model: EPIC-μCT Images. 3D images obtained from EPIC-μCT scanning were used to visualize the sGAG distribution in full-thickness control and IL-1-stimulated explants (A). Consistent with the assumed inverse partitioning of the negatively charged ioxaglate dimer within the ECM of cartilage, low x-ray attenuation corresponded to regions of high sGAG content and high x-ray attenuation indicated regions of low sGAG content. Control explants showed little visible variation in x-ray attenuation with culture time. In contrast, increases in x-ray attenuation were apparent for IL-1-stimulated explants beginning at day 4. In our full-thickness explant model, increases in x-ray attenuation appeared first in the superficial zone, in the deep zone, and on the outer surfaces of the IL-1-stimulated explants and progressed toward the interior with time. In Vitro Cartilage Degeneration Model: Safranin-O Histology. Histological sections revealed similar patterns of sGAG depletion. In longitudinal sections taken from the center of full-thickness explants, control explants exhibited a dense, uniform distribution of sGAGs (B). In contrast, IL-1-stimulated explants exhibited visible decreases in sGAG staining beginning at day 4 and continuing throughout the culture period. The decreased staining in the IL-1 group consistently progressed from the superficial zone, the deep zone, and the outer surfaces inward, consistent with the locations of increased attenuation in the EPIC-μCT images. In Vitro Cartilage Degeneration Model: Quantitative Changes in PG Content. Consistent with the EPIC-μCT and histology images, the overall average x-ray attenuation for the full-thickness control explants did not significantly vary with culture time (C). In contrast, the attenuation of IL-1-stimulated explants progressively increased, with levels significantly higher than in control explants beginning at day 6. The increased release of PGs to the media in response to IL-1 treatment was consistent with the increase in x-ray attenuation. Full-thickness control explants released a low level of sGAGs throughout the 10-day culture (D). In contrast, IL-1-stimulated cartilage showed a characteristic rise in media sGAG levels that peaked at day 8 and was significantly elevated over control explants beginning at day 4. The progressive loss of sGAGs to the media was consistent with both reduced safranin-O staining and increased attenuation in EPIC-μCT analysis. Correlation of X-Ray Attenuation and sGAG Concentration. To directly examine the relationship between explant composition and x-ray attenuation, both control and IL-1-stimulated cartilage explants were digested and analyzed for sGAG content after EPIC-μCT scanning. A linear regression analysis revealed a significant relationship (r2 = 0.91, P < 0.0005) between x-ray attenuation and sGAG/H2O (E). The negative Pearson coefficient, r = −0.95, agreed with the assumed inverse distribution of a negatively charged contrast agent and sGAGs within the ECM. In Situ EPIC-μCT Imaging of a Rabbit Joint. The reconstructed image of an immature rabbit femur after exposure to a 40%/60% solution of Hexabrix/0.15 M PBS reveals articular surface topography and soft tissue features (A and C) absent from a control image after desorption of the contrast agent (B). In particular, the EPIC-μCT images demonstrate the ability to distinguish ligament insertion points (black arrows) and a scalpel mark on the articular surface (white arrow). After segmentation of cartilage from bone using thresholds identified from the histogram, a thickness map of the articular cartilage surface layer (D) indicates thickness values that are consistent with the anticipated regional variations in rabbit articular cartilage thickness. Noninvasive techniques capable of detecting PG content and distribution and tissue morphology could dramatically improve upon existing evaluation options for monitoring the efficacy of OA treatments and cartilage repair strategies. This work details the development of EPIC-μCT, a technique for the 3D, quantitative imaging of PGs in soft tissues. By using IL-1 stimulation of articular cartilage as an in vitro degradation model, the ability of EPIC-μCT to monitor the progressive loss of sGAGs from cartilage explants was demonstrated. In addition, the strong correlation found between x-ray attenuation and sGAG density indicates that this technique can be used not only to monitor relative changes in sGAG content and distribution but also to quantitatively evaluate sGAG content, a feature that may prove useful for in vitro analysis of matrix changes in studies of tissue degeneration and development of tissue engineered constructs. EPIC-μCT also shows the ability to produce 3D images of the articular surface of cartilage in situ, illustrating potential applications in monitoring surface contours and engineered construct integration in defect repair studies as well as regional variations in cartilage thickness. The EPIC-μCT technique is complementary to several other methods that have recently been introduced. Contrast-enhanced MRI techniques including dGEMRIC have shown success in monitoring PG concentration in cartilage in vitro and clinically ( 15, 16), but the resolution of current clinical MRI systems (≈300 μm) and high-powered research MRI systems (≈25 μm in-plane) may limit their application in small-animal models. The use of gadolinium-based contrast agents with μCT has also shown the potential to monitor PG content in cartilage explants ( 17). Optical coherence tomography ( 18) and Fourier transform infrared spectroscopy ( 19) may also be useful in the noninvasive monitoring of cartilage ECM changes, potentially providing information on collagen and PG contents. Although the latter methods provide high-resolution planar images, quantitative analyses of 3D distributions and morphology are not currently possible. Although it is not currently viable as a clinical imaging modality, the ability to simultaneously provide quantitative data regarding tissue composition and morphology makes EPIC-μCT a powerful research tool. With the noninvasive, high-resolution monitoring capability of μCT, the greatest benefit of this technique may lie in the assessment of ECM changes in rodent models of cartilage degeneration and regeneration. A promising extension of this work is the application of EPIC-μCT for end-stage analysis of intact cartilage surfaces after dissection of small-animal joints. Furthermore, the capability of commercial μCT scanners for in vivo analysis raises the exciting possibility of longitudinal monitoring of cartilage matrix changes during disease or treatment progression in individual animals. A critical step in advancing this technique is the development of accurate methodology for segmenting cartilage from bone in 3D data sets, because the x-ray attenuation of bone and Hexabrix-equilibrated soft tissue can overlap. The overlap in x-ray attenuation depends on the mineral content of the subchondral bone as well as the fixed charge density (FCD) and water content of the cartilage, all of which can vary with age, health, and species. As demonstrated in the present study, one approach is to dilute the contrast agent sufficiently that the attenuation ranges of the soft tissue and bone become distinct, improving the delineation of the cartilage/bone interface and thus the accuracy of cartilage thickness measurements. However, such dilution simultaneously reduces the sensitivity of the contrast agent concentration to FCD, limiting the PG monitoring capabilities of the technique. If intermediate concentrations cannot be identified that allow both morphological and compositional analysis, separate scanning after equilibration in dilute and concentrated contrast agent solutions may be required to fully realize the potential of EPIC-μCT. Additionally, in vivo imaging raises additional technical issues involving the delivery and retention of a sufficient volume and concentration of the contrast agent. Intraarticular injection offers a simple and direct approach, although clearance of the contrast agent from the joint space could impair adequate equilibration. Distinguishing between apposed cartilage layers, underlying bone, and adjacent soft tissues in a small-animal joint will be technically challenging and may or may not be resolvable by varying the imaging window and the concentration of Hexabrix delivered. Despite these potential limitations, the present study establishes the basis for a quantitative, high-resolution, 3D imaging technique capable of nondestructively monitoring cartilage matrix organization in vitro that may become a powerful tool for analyzing small-animal models of cartilage degeneration and regeneration. Principle of the Technique. Similar to the dGEMRIC technique ( 15), EPIC-μCT relies on the partitioning of a charged contrast agent within a charged ECM. In cartilage, type II collagen carries a nearly neutral net charge whereas the sulfated GAGs primarily associated with the large, aggregating PG aggrecan confer a highly negative net charge to the ECM. The interstitial fluid carries a corresponding excess of positively charged solutes, resulting in bulk macroscopic electroneutrality. Given these properties, conditions of electrochemical equilibrium predict that a negatively charged solute capable of permeating the tissue would distribute inversely to the negatively charged sGAGs and therefore spatially target regions of lower sGAG density. For a region of tissue with a FCD ρ m (in Coulombs/liter) submerged in a bath containing a single monovalent salt at a concentration Co, the Gibbs–Donnan theory for an ideal solution predicts the interstitial anion concentration within the tissue C − as
where F is Faraday's constant. For a tissue such as cartilage, a negatively charged solute capable of permeating the tissue would distribute inversely to the negatively charged sGAGs and therefore spatially target regions of lower sGAG density. If the anion concentration could be noninvasively detected it would thus provide an indirect measure of the matrix FCD (and therefore the sGAG concentration). It should be noted that although Eq. 1 describes the distribution of an ideal anionic solute in a dilute solution, factors including steric interactions between solutes, solvation effects, and restricted availability of intrafibrillar water will influence the activity (or effective concentration) of a solute within the tissue, resulting in deviations from the ideal behavior. Furthermore, most commercial contrast agents are multicomponent mixtures, complicating theoretical predictions of solute behavior. Additionally, a macroscopic imaging technique such as EPIC-μCT will actually detect the apparent macroscopic concentration  −:
where ![[var phi]](/corehtml/pmc/pmcents/x03C6.gif) f is the fluid volume fraction (≈0.8 for articular cartilage) ( 1). Consequently, the quantitative relationship between the concentration of the anionic contrast agent and the tissue FCD will not exactly follow Eq. 1 and must be determined empirically for a given contrast formulation or dilution and tissue type. Serial dilutions of the contrast agent (as in A) could then be used as phantoms to calibrate attenuation as a function of contrast agent concentration, allowing standardization of the quantification to account for different scan parameters or differences between μCT systems. Although this approach would work in theory for any ionic contrast agent, the present study explored contrast-enhanced μCT imaging of cartilage using Hexabrix 320 (Mallinckrodt, Hazelwood, MO), an ionic contrast agent labeled for clinical use in contrast-enhanced CT scanning of the head and body and for direct injection in the uterus, fallopian tubes, and joint spaces. Hexabrix 320 is a sterile, nonpyrogenic, aqueous solution consisting of 39.3% (wt/vol) ioxaglate meglumine and 19.6% (wt/vol) ioxaglate sodium. Dissociated ioxaglate provides six iodine atoms per monovalent anion, resulting in a greater x-ray attenuation at a given osmolality than is achieved with triiodinated monomers (A) and thus providing a greater sensitivity to differences in FCD. Considering the important functional contributions of the FCD of cartilage that arise from the interactions of bound sGAGs with mobile ions in the interstitial fluid, the osmolarity of the contrast agent was also a concern, because significant mismatches would yield swelling or shrinkage artifacts. Although slightly hypertonic to cartilage (≈300–400 mOsm/liter), Hexabrix at 460 mOsm/liter represents one of the closest matches among ionic iodinated contrast agents. In addition, the commercial formulation of Hexabrix is pH balanced to 6–7.6 pH, minimizing any concerns of the effect of short-term exposure on cell viability and metabolic activity. In Vitro Evaluation Model. To evaluate the potential for EPIC-μCT to monitor changes in the ECM of cartilage, degradation was produced in vitro by incubation of bovine articular cartilage explants with IL-1α. Briefly, IL-1α and IL-1β are regulatory and proinflammatory cytokines implicated in the progression of OA. Within hours of treatment with IL-1, an up-regulation in cell-mediated aggrecanase expression is noted in articular cartilage ( 13, 20, 21). The increased aggrecanase expression leads to progressive degradation of the PG network. After PG depletion the collagen network undergoes matrix metalloproteinase-mediated degradation ( 20– 22). Given the similarities between matrix degradation in the IL-1 model and in degenerative conditions, IL-1 stimulation of articular cartilage is a well suited model for this evaluation study. Tissue Explant Preparation and Culture. Sixty-six 4-mm-diameter, full-thickness explants were harvested aseptically with a disposable biopsy punch (Miltex, York, PA) from both stifle joints of a 2- to 4-week-old calf (Research 87, Boylston, MA). By using a custom cutting block, explants were trimmed of all visible subchondral bone by cutting parallel to the articular surface. The tallest and shortest heights of each sample were measured with digital calipers and averaged to obtain the explant thickness. The full-thickness explants had an average thickness of 5.00 ± 0.07 mm with a range of 3.85–6.35 mm. Sixty explants were assigned to either control or IL-1-treated groups, and six explants were designated as day-0 controls, with particular care taken to ensure a uniform distribution of thicknesses and harvest legs across groups. Explants were cultured in 24-well plates (Becton Dickinson, Franklin Lakes, NJ) in 1 ml of either control medium [high glucose DMEM (Invitrogen, Carlsbad, CA) supplemented with 50 μg/ml ascorbate (Sigma, St. Louis, MO), 0.1 mM nonessential amino acids (Invitrogen), and 50 μg/ml gentamicin (Invitrogen)] or control medium supplemented with 20 ng/ml recombinant human IL-1α (PeproTech, Rocky Hill, NJ). This IL-1 concentration has previously been shown to yield nearly complete depletion of sGAGs from bovine articular cartilage explants within 2 weeks ( 21). The control and IL-1-treated explants were cultured for up to 10 days, with media changed and collected every 48 h. Six explants per group were removed from culture at days 0, 2, 4, 6, 8, and 10 for EPIC-μCT scanning and subsequent destructive analysis. μCT Scanning and Analysis. After removal from culture, each explant was incubated with 1 ml of full-strength Hexabrix supplemented with protease inhibitors at 37°C under gentle agitation for 24 h, exceeding the time required for full equilibration (B). Explants were then patted to remove excess fluid and loaded six per batch into custom μCT chambers. All scanning was performed in air by using a VivaCT 40 (Scanco Medical, Bassersdorf, Switzerland) at 45 kVa, 133 mA, 200-ms integration time, and a voxel size of 21 μm. Although scan time varied with the thickness of the explant, two explants were scanned in parallel in an average of 15 min. After scanning, explants were immersed in 2 ml of 0.15 M PBS supplemented with protease inhibitors for 24 h at 4°C to allow for desorption of the Hexabrix. One explant per condition per time point was reserved for subsequent histological analysis. The remaining explants were placed in fresh PBS and protease inhibitors and stored at −20°C until further biochemical analysis. By using Scanco Medical software, a histogram of the x-ray attenuation readings was generated and used to determine an appropriate threshold for image processing. The average x-ray attenuation was than calculated for the thresholded volume of each full-thickness explant. To visualize the spatial variation in sGAG content, 3D color images representing the range of x-ray attenuation values within a sample were generated by using Scanco Medical image processing software. Note that valid comparisons between samples and accurate quantification of FCD require that consistent scan and threshold parameters be used for all samples and standards. Biochemical Analysis. After EPIC-μCT scanning, explants reserved for biochemical analysis were lyophilized for 18 h and solubilized with 1 mg/ml proteinase K (EMD Biosciences, San Diego, CA) in 100 mM ammonium acetate (pH 7.0). The digested explants and conditioned media were then assayed for sGAGs via the 1,9-dimethylmethylene blue colorimetric assay with shark chondroitin sulfate (Sigma) as a standard ( 23). Histology. For comparison with the 3D spatial images generated by EPIC-μCT, a subset of full-thickness explants was stained for sGAGs by safranin-O. Explants reserved for histology were placed in 10% neutral buffered formalin for 72 h at 4°C before being transferred to 70% ethanol at 4°C until paraffin embedding and processing. Explants were cut down the longitudinal axis, creating two cylindrical halves. After paraffin embedding, 5-μm-thick sections through the entire explant thickness were removed by using a rotary microtome. Sections were stained for sGAGs by using a 0.l% safranin-O solution with a 0.2% aqueous solution of fast green used as a counterstain. In Situ Imaging of Rabbit Femur. A femur was obtained postmortem from an ≈3-month-old rabbit from an unrelated study and dissected free of surrounding tissue (A). In preliminary experiments, the ability to visualize soft tissue during in situ imaging of an immature rabbit femur was highly dependent on the concentration of the Hexabrix equilibration solution. Use of the full-strength Hexabrix solution (100%) provided excellent visualization of surface topology, but the enhanced radiopacity of the soft tissue proved too similar to that of bone to allow reproducible identification of the tissue interface. As the Hexabrix solution was diluted with 0.15 M PBS, the radiopacity of the soft tissue linearly decreased and the cartilage could be more readily distinguished from the underlying bone (C Inset). Note that dilution of the Hexabrix solution also decreases the sensitivity of the attenuation to differences in tissue FCD, and additional analyses are required to determine whether an intermediate contrast agent concentration will allow the cartilage to be distinguished from the bone while simultaneously facilitating accurate quantification of the FCD. Based on results of dilution studies, the femur was immersed for 1 h in 40%/60% (vol/vol) Hexabrix/0.15 M PBS at 37°C under gentle agitation. The femur was scanned at a 21-μm voxel resolution in a VivaCT 40 and evaluated after selecting an appropriate threshold (C). A histogram of the x-ray attenuation values obtained revealed two partially overlapping peaks corresponding to contrast-enhanced soft tissues and calcified bone (C). The soft tissue peak was isolated from the data set by using lower and upper thresholds, and a corresponding soft tissue volume was generated. By using a direct distance transformation algorithm (Scanco Medical), a thickness map of the soft tissue volume was generated (D). After desorption of the contrast agent for 1 h in 0.15 M PBS, the femur was rescanned at 21 μm and evaluated at the same threshold settings as the control image (C). Statistics. Unless otherwise indicated, all results are reported as mean ± SEM. Minitab Release 12.23 (Minitab, State College, PA) was used for all statistical analyses. Average x-ray attenuation and sGAGs released to the media with time were evaluated by using a two-factor (treatment and culture time) general linear model and by using Tukey's test (P < 0.05) for pairwise comparisons. The relationship between explant sGAG concentration and average x-ray attenuation was examined via linear regression analysis. We thank Angela Lin, Srinidhi Nagaraja, and Galen Robertson for technical assistance with the μCT imaging and processing and Jhade Woodall for assistance with the in situ imaging work. This work was supported by the Georgia Tech/Emory Center for the Engineering of Living Tissues, an Engineering Research Center Program of the National Science Foundation under Award EEC-9731643; by the Arthritis Foundation through an Arthritis Investigator Award; by the National Institutes of Health through Grant R21AR053716; and by TI:GER National Science Foundation's Integrative Graduate Education in Research Training Grant 0221600 fellowship (to A.W.P.). | | sGAG | sulfated glycosaminoglycan | | | CT | computed tomography | | | μCT | microcomputed tomography | | | PG | proteoglycan | | | FCD | fixed charge density | | | ECM | extracellular matrix | | | EPIC-μCT | equilibrium partitioning of an ionic contrast agent via μCT. |
1. Mow VC, Ratcliffe A. In: Basic Orthopaedic Biomechanics. Mow VC, Hayes WC, editors. Philadelphia: Lippincott-Raven; 1997. pp. 113–177. 2. Maroudas A, Bannon C. Biorheology. 1981;18:619–632. [PubMed] 3. Williamson AK, Chen AC, Masuda K, Thonar EJ, Sah RL. J Orthop Res. 2003;21:872–880. [PubMed] 4. Williams A, Oppenheimer RA, Gray ML, Burstein D. Arthritis Res Ther. 2003;5:R97–R105. [PubMed] 5. Rayan V, Hardingham T. Matrix Biol. 1994;14:263–271. [PubMed] 6. Patwari P, Cook MN, DiMicco MA, Blake SM, James IE, Kumar S, Cole AA, Lark MW, Grodzinsky AJ. Arthritis Rheum. 2003;48:1292–1301. [PubMed] 7. Sandy JD, Verscharen C. Biochem J. 2001;358:615–626. [PubMed] 8. Plaas AH, West LA, Wong-Palms S, Nelson FR. J Biol Chem. 1998;273:12642–12649. [PubMed] 9. Setton LA, Mow VC, Muller FJ, Pita JC, Howell DS. Agents Actions Suppl. 1993;39:27–48. [PubMed] 10. Williamson AK, Chen AC, Sah RL. J Orthop Res. 2001;19:1113–1121. [PubMed] 11. Rupprecht M, Pogoda P, Mumme M, Rueger JM, Puschel K, Amling M. J Orthop Res. 2006;24:664–674. [PubMed] 12. Jiang Y, Zhao J, Liao EY, Dai RC, Wu XP, Genant HK. J Bone Miner Metab. 2005;23(Suppl):122–131. [PubMed] 13. Arner EC, Hughes CE, Decicco CP, Caterson B, Tortorella MD. Osteoarthritis Cartilage. 1998;6:214–228. [PubMed] 14. Billinghurst RC, Wu W, Ionescu M, Reiner A, Dahlberg L, Chen J, van Wart H, Poole AR. Arthritis Rheum. 2000;43:664–672. [PubMed] 15. Bashir A, Gray ML, Burstein D. Magn Reson Med. 1996;36:665–673. [PubMed] 16. Kim YJ, Jaramillo D, Millis MB, Gray ML, Burstein D. J Bone Joint Surg Am. 2003;85-A:1987–1992. [PubMed] 17. Cockman MD, Blanton CA, Chmielewski PA, Dong L, Dufresne TE, Hookfin EB, Karb MJ, Liu S, Wehmeyer KR. Osteoarthritis Cartilage. 2006;14:210–214. [PubMed] 18. Patel NA, Zoeller J, Stamper DL, Fujimoto JG, Brezinski ME. IEEE Trans Med Imaging. 2005;24:155–159. [PubMed] 19. Camacho NP, West P, Torzilli PA, Mendelsohn R. Biopolymers. 2001;62:1–8. [PubMed] 20. Bonassar LJ, Sandy JD, Lark MW, Plaas AH, Frank EH, Grodzinsky AJ. Arch Biochem Biophys. 1997;344:404–412. [PubMed] 21. Legare A, Garon M, Guardo R, Savard P, Poole AR, Buschmann MD. J Orthop Res. 2002;20:819–826. [PubMed] 22. Kozaci LD, Buttle DJ, Hollander AP. Arthritis Rheum. 1997;40:164–174. [PubMed] 23. Farndale RW, Sayers CA, Barrett AJ. Connect Tissue Res. 1982;9:247–248. [PubMed]
|
This article has been 
