Age is an important risk factor for osteoarthritis (OA). During aging, nonenzymatic glycation results in the accumulation of advanced glycation end products (AGEs) in cartilage collagen. We studied the effect of AGE crosslinking on the stiffness of the collagen network in human articular cartilage.
To increase AGE levels, human adult articular cartilage was incubated with threose. The stiffness of the collagen network was measured as the instantaneous deformation (ID) of the cartilage and as the change in tensile stress in the collagen network as a function of hydration (osmotic stress technique). AGE levels in the collagen network were determined as: Nε-(carboxy[m]ethyl)lysine, pentosidine, amino acid modification (loss of arginine and [hydroxy-]lysine), AGE fluorescence (360/460 nm), and digestibility by bacterial collagenase.
Incubation of cartilage with threose resulted in a dose-dependent increase in AGEs and a concomitant decrease in ID (r = −0.81, P < 0.001; up to a 40% decrease at 200 mM threose), i.e., increased stiffness, which was confirmed by results from the osmotic stress technique. The decreased ID strongly correlated with AGE levels (e.g., AGE fluorescence r = −0.81, P < 0.0001). Coincubation with arginine or lysine (glycation inhibitors) attenuated the threose-induced decrease in ID (P < 0.05).
Increasing cartilage AGE crosslinking by in vitro incubation with threose resulted in increased stiffness of the collagen network. Increased stiffness by AGE crosslinking may contribute to the age-related failure of the collagen network in human articular cartilage to resist damage. Thus, the age-related accumulation of AGE crosslinks presents a putative molecular mechanism whereby age is a predisposing factor for the development of OA.
Osteoarthritis (OA) is a common chronic disabling disorder for which age is the single greatest risk factor (1, 2). Although age-related changes in articular cartilage are likely to play a role, the mechanism by which age increases the susceptibility to joint degeneration is largely unknown. Swelling of cartilage, which is proportional to the amount of damaged collagen (3), is the initial event in cartilage degeneration (4), indicating that the development of OA starts with failure of the cartilage collagen network. With increasing age, the stiffness of the collagen network in articular cartilage increases (5, 6), which may result in an age-related increase in susceptibility to mechanically induced damage. Indeed, measurements of the tensile properties of articular cartilage show marked changes with age (7, 8) and indicate that the resistance of the collagen network to fatigue damage decreases with increasing age (9, 10). We hypothesize that the age-related increase in stiffness of the cartilage collagen network results in a decreased resistance of this network to mechanical failure and, consequently, an increase in the risk of developing OA.
The question that has yet to be answered is: What causes the stiffness of the articular cartilage collagen network to increase with age? Because of the exceptionally long half-life (>100 years) of collagen molecules in articular cartilage (11), they are susceptible to the accumulation of end products of nonenzymatic glycation. Nonenzymatic glycation of proteins is initiated by the reaction of sugars with lysine and arginine residues in proteins, and eventually leads to the formation and age-related accumulation of advanced glycation end products (AGEs), such as Nε-(carboxymethyl)lysine (CML) (12) and Nε-(carboxyethyl)lysine (CEL) (13), and crosslinks, such as pentosidine (14, 15), methylglyoxal-lysine dimer (MOLD) (16), and threosidine (17).
In articular cartilage, relatively high levels of AGEs accumulate with increasing age (18–20). In a preliminary experiment, generation of AGEs resulted in a stiffening of the articular cartilage collagen network (21), consistent with reports on other collagenous tissues (22, 23). Furthermore, accumulation of AGEs makes the collagen network more brittle, as was described for articular cartilage (24), lens capsules (22), and bone (Catanese J, et al: unpublished observations). We hypothesize that AGE crosslinking results in increased stiffness of the cartilage collagen network and that this subsequently leads to increased susceptibility of the collagen network to mechanical failure (brittleness). Thus, accumulation of AGEs could be a molecular mechanism that causes age to be a major predisposing factor for the development of OA.
Threose is a highly reactive carbohydrate that is formed by nonenzymatic degradation of ascorbic acid both in vitro and in vivo (25, 26). Among the ascorbic acid degradation products, L-threose is the most abundant (∼20–25%) and the most reactive degradation product (25–27). L-threose rapidly modifies Lys residues in proteins (28), leading to the formation of characteristic AGEs, including CML (29), formyl threosyl pyrrole (30), threosidine (17), and pentosidine (30).
The present study was designed to determine the effects of AGE crosslinking on the stiffness of the collagen network in articular cartilage. For that purpose, we measured the instantaneous deformation (ID) of the tissue in unconfined compression (31). As shown previously, the ID is mainly controlled by the mechanical properties of the collagen network in articular cartilage (31). These ID results were confirmed by the osmotic stress technique described by Basser et al (6), a method that is precisely adapted to measuring the stress in the collagen network itself. To prevent other age-related changes in articular cartilage from confounding the interpretation of these experiments, cartilage was glycated in vitro (using threose as a model compound). In the present study, not only the stiffness, but also the extent, of amino acid modification and AGE crosslinking of the collagen network was determined. Stiffness measurements were correlated with AGE levels. Furthermore, to confirm that the effects of threose on mechanical properties of the collagen network were indeed the result of increased AGE crosslinking, we tested the effect of incubation of cartilage with threose in combination with Lys or Arg, both of which are inhibitors of the glycation process (32, 33).
MATERIALS AND METHODS
Tissue samples. Macroscopically normal human articular cartilage was obtained postmortem, within 18 hours after death, from subjects who had no clinical history of joint disorders. Samples of full-thickness cartilage, excluding the underlying bone, were taken from the femoral condyles of 30-, 31-, and 37-year-old donors and from the superior region of the femoral head of 24-, 30-, and 43-year-old donors. All tissue samples were stored at −20°C until analyzed.
Carbohydrate incubations. Healthy, full-depth cylindrical cartilage samples (5.5 mm diameter) were punched from the joints and immersed in 0.15M NaCl for 24 hours at 4°C. Prior to the incubations, the fixed charge density (FCD; see below) was measured to distribute samples evenly among the incubation groups according to their FCD. Incubations with threose were carried out in phosphate buffered saline (pH 7.4) containing 25 mM EDTA (Spectrum Medical Industries, Los Angeles, CA) as proteinase inhibitor in 2 independent experiments. Cartilage samples from 37- and 43-year-old donors were incubated with 0–200 mM threose (Sigma, St. Louis, MO) and with 200 mM threose in combination with 20 mM L-lysine (Sigma) or L-arginine (Sigma). All incubations were performed for 6 days at 37°C; control samples were incubated in the same solution without threose. Two or 3 cartilage samples were used per condition.
After the incubations, samples were washed for 24 hours in 0.15M NaCl at 4°C with a change of the NaCl solution after 16 hours. Samples were subjected to unconfined compressive loading to determine the ID (see below) both before and after the incubations. Using an approach similar to that for threose, cartilage samples from 30- and 31-year-old donors were incubated with methylglyoxal (0–20 mM) and ribose (0–200 mM; Sigma). Methylglyoxal was freshly prepared by acid hydrolysis of 1,1-dimethoxyacetone (Sigma), and the preparation was purified by distillation (34).
FCD. FCD is defined as the concentration of negatively charged groups in the tissue (mEq/gm) and represents the glycosaminoglycan content (35). The FCD was determined by means of the tracer cation method described elsewhere (36), using 0.015M NaCl labeled with 22Na as the equilibrating solution.
Instantaneous deformation. As a measure of the stiffness of the collagen network, the ID of the cartilage was determined; this has previously been shown to be mainly influenced by the properties of the collagen network (31). ID was measured by subjecting the full-depth samples to unconfined compressive step-loading in a custom-built apparatus (31). During the experiment, the sample was immersed in 0.15M NaCl at 4°C in a transparent glass cell. The top surface of the sample was compressed against a transparent rigid plunger and viewed with a microscope equipped with a photo camera. Prior to loading, a preload of ∼0.03 mPa was applied to ensure contact between the surface of the sample and the plunger. The sample was then loaded so that the initial pressure on its top surface was 3.0 mPa. The ID, expressed as the percentage increase in area of the articular surface, was measured at 1 second after loading. The ID of the samples was measured before and after the incubations, and data are presented as the ratio of the ID after incubation to the ID before incubation.
Since the proteoglycan content of cartilage modulates the tensile stresses in the collagen network, it plays an indirect role in determining the ID (31). Addition of EDTA to the carbohydrate solutions was used to prevent matrix metalloproteinase– and aggrecanase-mediated proteoglycan loss from the tissue during the incubation, which was confirmed by FCD measurements before and after incubation.
Osmotic stress technique. To corroborate the effect of AGE crosslinking on the stiffness of the cartilage collagen network, the tensile stress in the collagen network as a function of hydration was determined for control and threose-incubated cartilage samples (6). Full-depth cartilage plugs obtained from the femoral head of 24- and 30-year-old donors were sliced on a freezing microtome (Leitz, Wetzlar, Germany); 400-μm sections were removed from the cartilage surface as well as from the deep zone. The test specimens thus represented mainly “middle zone” cartilage. We chose to use sections from the middle zone because, in normal cartilage, the FCD is known to be relatively uniform in this region (37). Samples were incubated in 0.15M NaCl containing 300 mM L-threose (Sigma), 25 mM EDTA (Spectrum), 5 units/ml penicillin (Sigma), and 5 μg/ml streptomycin (Sigma). The incubation was carried out for 5 days at 37°C.
The osmotic stress technique was carried out on sets of control samples and threose-incubated samples taken from adjacent sites on the femoral head (n = 4). The method involved incubation of samples enclosed in dialysis tubing (Spectrum) in calibrated, osmotically active polyethylene glycol (PEG) (20 kd; Fluka, Buchs, Switzerland) solutions. The tensile stress exerted by the collagen network, Pc, can be calculated from the “balance of forces” at equilibrium hydration (6). At equilibrium hydration, Pc, together with the externally applied osmotic stress, ϕPEG, both of which tend to squeeze water out of the tissue, are balanced by the osmotic pressure of the cartilage proteoglycans, ϕPG. Pc is plotted as a function of collagen network hydration (normalized tissue volume: Vnormalized = [Vtotal − Vcollagen]/Vcollagen), which was also obtained using the method described by Basser et al (6). The larger the change in collagen tensile stress per change in hydration (i.e., the steeper the slope of the Pc/Vnormalized curve) the greater the tensile stiffness of the collagen network.
Determination of collagen AGE levels. The methods used for analysis of cartilage collagen AGE levels have been extensively described in our previous work (20). Articular cartilage collagen was purified by depleting the tissue of all noncollagenous proteins by sequential enzymatic treatment with chondroitinase ABC (Sigma), trypsin (Boehringer Mannheim, Mannheim, Germany), and Streptomyces hyaluronidase (Sigma) (20).
For analysis of CML, CEL, and pentosidine, purified cartilage collagen was reduced with sodium borohydride (Sigma) (20). After acid hydrolysis, an aliquot of these samples was removed for pentosidine and amino acid analysis. The CML, CEL, and Lys content of the collagen hydrolyzates was measured as their N-trifluoroacetyl methyl esters by isotope- dilution selected ion monitoring gas chromatography–mass spectrometry using deuterated internal standards (12, 13). The pentosidine and amino acid content of the hydrolysates were determined by high-performance liquid chromatography (38, 39). The pentosidine content of the collagen samples was expressed as mmoles per mole of collagen, while the CML, CEL, Arg, hydroxylysine, and Lys contents of the collagen samples were expressed as moles per mole of collagen, assuming 300 hydroxyproline residues per triple-helical collagen molecule (39). From the comparison of the sum of the Arg, Hyl, and Lys residues per collagen molecule in threose-treated samples with control samples, the mean percentage of modification of these glycation-sensitive amino acids was calculated, assuming 0% modification in the control samples.
AGE fluorescence (λex = 360 nm, λem = 460 nm) in purified cartilage collagen was measured after digestion of the collagen with papain (from papaya latex; Sigma) (20). Papain buffer was used as a blank. AGE fluorescence was expressed as relative fluorescence units and was normalized to the hydroxyproline content of the digest, measured after acid hydrolysis (papain contributed <1% of the hydroxyproline in the digests) (40, 41).
The susceptibility of purified cartilage collagen to collagenase digestion was measured following digestion with Clostridium histolyticum collagenase (CLS 2; Worthington, Freehold, NJ) (20). After incubation, the supernatant and remaining tissue were separated, and the amount of collagen in both fractions was estimated by measuring the amount of hydroxyproline (40, 41).
Statistical analysis. Linear regression analyses, 1-way analysis of variance with post hoc Tukey's highest significant difference tests for multiple-group comparisons, and Student's t-tests for 2-group comparisons were performed with SPSS version 10.0 for Windows (SPSS, Chicago, IL). P values less than 0.05 were considered significant. When no individual data are given, the data represent the mean ± SD.
Effect of threose on AGE levels in articular cartilage collagen. AGE levels were determined in collagen that was purified from cartilage samples that had been incubated with different concentrations of threose. The levels of CML and CEL increased significantly (2.7- and 7.2–fold, respectively) (Figure 1A) following incubation with threose, while only a slight increase in the collagen pentosidine concentration was observed at 200 mM threose (1.2-fold; data not shown). Besides the concentrations of these chemically well-characterized AGEs, more general measures of glycation and AGE crosslinking were determined in the threose-incubated cartilage samples.
Threose incubation resulted in extensive modification of Arg, Hyl, and Lys residues in cartilage collagen (mean ± SD 17.7 ± 7.2, 57.2 ± 7.4, and 60.0 ± 10.7%, respectively, at 200 mM of threose, n = 2 independent experiments; the overall modification of Arg, Hyl, and Lys is shown in Figure 1B). The well-characterized AGEs (CML, CEL, and pentosidine) account for only a small fraction of all the AGEs formed during incubation with threose in vitro: the total amount of CML, CEL, and pentosidine represented only 1.13 ± 0.21% of the Lys modifications. AGE fluorescence, a general measure of crosslinking AGEs, was also increased in threose-incubated samples (Figure 1C). Consistently, threose reduced the digestibility of the collagen by bacterial collagenase, which represents a functional measure of cartilage collagen crosslinking (Figure 1D). Except for pentosidine, all AGE measures (CML and CEL levels, overall amino acid modification, AGE fluorescence, and collagenase digestibility) correlated well with one another (r > 0.80 in all cases).
Effect of threose on the mechanical properties of the articular cartilage collagen network. ID is expressed as the ratio of the ID after incubation to the ID before incubation. If the incubation does not affect the ID, this ratio will equal 1. As expected, the ratio of the control samples was 1.00 ± 0.05 (Figure 2, incubation with 0 mM threose). A ratio <1 reflects a decrease in deformation due to threose incubation. Clearly, incubation of cartilage with up to 200 mM threose resulted in a dose-dependent decrease in ID (Figure 2). At 200 mM threose, the ID was decreased by 40% (P < 0.0005 compared with control), reflecting a major change in the mechanical properties of the collagen network.
Curves of the tensile stress in the collagen network (Pc) as a function of hydration (Vnormalized; example shown in Figure 3) were obtained by the osmotic stress technique described by Basser et al (6). Incubation with threose (300 mM, 5 days, 37°C) resulted in an increase in the slope of the Pc/Vnormalized curve from a mean ± SD of 3.6 ± 1.3 for control samples to 7.0 ± 1.0 for threose-incubated samples (P < 0.01, n = 4; slope determined at the steepest part of the curve). The steeper slope indicates an increase in tensile stiffness of the collagen network in human articular cartilage. Increased AGE crosslinking of collagen in the threose-incubated cartilage samples was confirmed by analyses of AGE fluorescence, pentosidine, and amino acid modification. Consistent with the ID experiments, all AGE measures were increased in collagen purified from the threose-incubated samples compared with the control samples; pentosidine and AGE fluorescence increased 2.4- and 9–fold, respectively, while the overall amino acid modification in the threose-incubated samples was 28.5 ± 1.5% (data not shown).
Correlation of ID of cartilage with AGE levels. In the present study, we investigated the effect of in vitro glycation by threose on the mechanical properties of the cartilage collagen network, in combination with measurements of AGE levels. Because AGE levels were measured specifically in collagen (purified from cartilage by enzymatic removal of all noncollagenous proteins) and because mechanical tests previously validated as measuring the stiffness of the collagen network (6, 31) were used, ID measurements could be related directly to AGE levels.
The decrease in ID due to threose correlated significantly with the sum of the well-characterized AGEs CML, CEL, and pentosidine in the collagen (r = −0.90, P < 0.01) (Figure 4A). The ID also decreased with increasing overall amino acid modification and AGE fluorescence in the collagen (r = −0.73, P < 0.0005 [Figure 4B]; and r = −0.81, P < 0.0001 [Figure 4C], respectively). Furthermore, the threose-mediated decrease in ID was related to a decrease in digestibility of the collagen by bacterial collagenase (indicative of increased collagen crosslinking; r = 0.78, P < 0.0005) (Figure 4D).
Inhibition of the effect of threose on ID of cartilage. To confirm that the threose-induced decrease in ID was indeed due to the increased AGE crosslinking, Arg and Lys were added to the threose incubations as inhibitors of the glycation and crosslinking process (32, 33). These amino acids were expected to compete with the cartilage proteins for the available threose, thereby preventing threose from forming AGE crosslinks in the cartilage proteins. The highest threose concentration (200 mM) used in the present study resulted in a 40% decrease in ID (Figures 2 and 5). This decrease in ID by 200 mM threose could be inhibited by 50% in the presence of 20 mM Arg (P < 0.05 compared with 200 mM threose alone) (Figure 5). Inhibition levels as high as 68% were observed in the presence of 20 mM Lys (P < 0.05 compared with 200 mM threose alone) (Figure 5). The resulting ID was unchanged from control levels (P = 0.09) (Figure 5). The fact that Lys is a more potent inhibitor than Arg is consistent with the higher reactivity of threose to Lys than to Arg (this study and ref. 28).
Effect of other carbohydrates on ID of articular cartilage. In addition to the effect of threose, the effects of ribose and methylglyoxal on the ID of cartilage were investigated to see whether the results obtained with threose could be generalized to other carbohydrates. The decrease in ID was significantly correlated with AGE fluorescence after ribose and methylglyoxal incubation (1-tailed regression analysis r = −0.64, P < 0.02, and r = −0.62, P < 0.05) (Figures 6A and B, respectively). Thus, experiments with both ribose and methylglyoxal confirm the relation between AGE levels and ID that is observed after threose incubation.
In this study, we showed that incubation of human articular cartilage with threose resulted in extensive AGE formation in the cartilage collagen network. Consistent with previous reports, CML and CEL were formed in substantially greater amounts than pentosidine (29, 30). Threose preferentially reacted with the amino group of Hyl and Lys residues and, to a lesser extent, with the guanidino group of Arg residues. The observed 3-fold difference in reactivity between Hyl/Lys and Arg is consistent with results reported by Lee et al (28). Also consistent with previous findings using other carbohydrates is our finding that only 1.1% of the Lys loss in cartilage collagen in our incubations could be explained by the formation of the well-characterized AGEs CML, CEL, and pentosidine (13, 16).
In addition, our recent in vivo findings indicated that only 4.7% of the age-related loss of (hydroxy-)lysine residues in articular cartilage collagen can be explained by CML, CEL, and pentosidine (20). Thus, these AGEs do not complete the mass balance and should only be considered as markers of overall AGE crosslinking. Other AGEs, including formyl threosyl pyrrole, a major threose-derived AGE (30), and threosidine, a Lys–Lys crosslink (17), as well as MOLD and GOLD (16), were not measured in this study, but probably also contribute to threose-mediated crosslinking of collagen. The crosslinking nature of the AGEs formed during the threose incubation was shown not only by the threose-induced AGE fluorescence (30), but also by the decrease in digestibility of cartilage collagen by bacterial collagenase after threose incubation (20).
Incubation of articular cartilage with threose clearly resulted in a decrease in the ID of cartilage, and this decreased deformation correlated well with AGE levels measured in purified collagen. The correlation of ID with AGE levels is not unique to threose; similar correlations were found after incubation of cartilage with ribose, which results in the formation of AGE crosslinks, such as pentosidine (14, 15) and the nonfluorescent amino acid component NFC-1 (42), and methylglyoxal, which forms AGE crosslinks such as MOLD (16). Since the ID is mainly controlled by the properties of the articular cartilage collagen network (31), our findings suggest that AGE crosslinking results in increased stiffness of the collagen network. The increase in stiffness of the collagen network due to AGE crosslinking was confirmed by our results with the osmotic stress technique, a method that accurately measures the stress in the collagen network itself (6).
It should be noted that AGE levels as generated by the incubation of young articular cartilage with threose, ribose, and methylglyoxal are higher than those found in aged human articular cartilage, although the levels are of the same order of magnitude (20). Since different sugars preferentially form certain AGEs and since AGE formation in vivo is due to the combined action of several sugars, a comparison of AGE levels in vitro and in vivo should be based on general AGE measures. As an example, our sugar incubations resulted in modification of glycation-sensitive amino acids (Arg, Hyl, and Lys): maximum of 44% with threose, 9% with ribose, and 35% with methylglyoxal. Modification of ∼10% of these amino acids is detected in aged human articular cartilage (20).
The addition of Lys or Arg to cartilage collagen being incubated with threose profoundly reduced the effect of threose on the stiffness of the collagen network, as has been shown for rat tail tendons incubated with glucose (33). Lys appeared to be a more potent inhibitor than Arg, which is consistent with the higher reactivity of threose with Lys than with Arg (ref. 28 and the present study). Thus, specifically interfering with the glycation process by adding Lys or Arg to the threose incubations resulted in inhibition of stiffening of the collagen network. This indicates that the effect of threose on the stiffness of the collagen network is not only highly correlated with AGE levels, but also is indeed due to AGE crosslinking.
In the present study, we used the ascorbic acid degradation product threose as a model glycating agent. As a consequence, our results may also contribute to our understanding of the role of ascorbic acid in the development of OA (43). Ascorbic acid may have a protective effect on the development of OA (44, 45) by its antioxidant properties and its ability to stimulate chondrocyte collagen and aggrecan synthesis (46, 47). In contrast, our present results suggest that high levels of ascorbic acid may increase the susceptibility of cartilage to degeneration through AGE crosslinking of the collagen network by threose. Thus, ascorbic acid can potentially have an effect on joint health by a variety of means. A trial is currently being conducted in a spontaneous in vivo OA model in guinea pigs, exploring the effects of high and moderate levels of ascorbic acid on the development of OA compared with a minimal, but nonscorbutic, level of the vitamin.
Our data from the osmotic stress technique (6) and preliminary data from ID measurements indicate that the stiffness of the cartilage collagen network increases with age, whereas studies using tensile tests have shown a decrease in cartilage stiffness with age (7, 8). The discrepancy between these studies likely results from the different methodologies used. Using the osmotic stress technique that was specifically designed to assess the stiffness of the collagen network, we currently show that the age-related increase in collagen stiffness can be readily explained by the age-related increase in AGE levels (20). Recent studies also demonstrate that in vitro–enhanced AGE levels increase cartilage stiffness in tensile testing (24). Therefore, it is surprising that the 50-fold increase in cartilage AGE levels with age does not result in an increase in stiffness when measured in tensile tests in an age range of human articular cartilage (7, 8). Furthermore, the age-related decrease in the water content of human articular cartilage (48), in combination with an age-related increase in FCD (5, 6), strengthens our conclusion that the stiffness of the collagen fiber network in cartilage must increase with age.
Based on our in vitro results, we conclude that AGE crosslinking of articular cartilage collagen results in increased stiffness of the collagen network. Although AGE levels in vivo are somewhat lower than in our in vitro incubations, the age-related accumulation of AGE crosslinks in collagen in human adult articular cartilage seems a plausible mechanism to explain the observed age-related increase in collagen network stiffness in human articular cartilage (5, 6). As a result of this stiffening, accumulation of crosslinking AGEs may make the cartilage collagen network more brittle (24). Increased stiffness and brittleness may, in turn, contribute to the age-related failure of cartilage to resist mechanical damage, and thus be a factor that predisposes aged cartilage to damage and, eventually, the development of OA.
Furthermore, the inhibitory effect of Lys and Arg on the induction of stiffness in the collagen network through AGE crosslinking presents interesting therapeutic possibilities. According to our hypothesis and present data, agents that prevent AGE accumulation (49) as well as agents that are proposed to “break” already formed AGE crosslinks (50, 51) may, through prevention of an increase in collagen network stiffness, slow down the development of OA.
We thank Professor Vincent Monnier at Case Western Reserve University, Cleveland, OH, for providing pentosidine, and Tom VanDenBroek and Esther Oldehinkel, TNO Prevention and Health, Leiden, The Netherlands, for their technical assistance. Dr. Virginia Byers Kraus at Duke University Medical Center, Durham, NC, is gratefully acknowledged for critically reading the manuscript.