Investigation of Molecular Motion of Proteoglycans in Cartilage by 13C Magnetic Resonance

‘92 nmr spectral parameters were measured for intact bovine nasal cartilage tissue, the purified proteoglycan ag- gregate, and chondroitin 4-sulfate. A comparison of integrated intensities obtained for four different samples of fresh tissue with an ethylene glycol standard indicated that at least 80% of the total glycosaminoglycan carbons in the tissue contributed to the spectrum. This result was con-firmed by intensity measurements obtained at 56” on fresh tissue and at 37” after extensive papain digestion of fresh tissue. Spin lattice relaxation times and nuclear Overhauser enhancements were analyzed in terms of the following models of molecular motion: of correlation times; anisotropic motion. The analysis indicates that the segmental motions of glycosaminoglycan chains are characterized by a broad distribution of correlation times centered at about 50 ns. Slow motion contributions to glycosaminoglycan line widths were reduced by dipolar decoupling decoupled spectra,

. The branched, highly negatively charged proteoglycan molecules occupy large solvent domains per unit mass (2,3). The proteoglycans are present in the matrix primarily as large aggregates in which 20 to 50 molecules are bound through noncovalent, specific interactions between a portion of the core protein and hyaluronic acid (4)(5)(6)(7)(8); small link proteins stabilize the aggregates (8)(9)(10). for its relative contents of the unsaturated disaccharide digestion products, 2-acetamido-2-deoxy-3-0-(/3-n-gluco-4-enepyranosyluronic acid)-4-O-sulfa-n-galactose and 2.acetamido-2deoxy-3-0-(P-o-gluco-4-enepyranosyluronic acid)-6-O-sulfo-n-galactose (see Fig. 1) using the paper chromatographic method developed by Saito et al. (15). Only small amounts of the nonsulfated disaccharide were detected. The summary of these data is presented in Table I. samples. In the modified configuration a sample of D,O, located a few centimeters from the 13C coil, provided a deuterium signal used for the lock. The ':'C coil had a length of 22 mm and the filling factor was maximized by making the inside diameter of the "'C coil only slightly larger than the outside diameter of the lo-mm tube. For this reason, samples were not spun. This was not a serious limitation in the present case, since nonspinning line widths of less than 3 Hz were easily achieved.
The ':'C transmitter delivered 30 Normally, a field of 0.8 G (yH,/h = 3.5 kHz) was used for scalar decoupling and a 15 G field (yH2/2r = 65 kHz) was used for dipolar decoupling. The ':'C coil had dimensions of 6 mm x 18 mm and produced an H, field of about 60 G which rotated the magnetization through 90" in 3.9 us. Since the 6-mm probe required only about 20% as much sample as the lo-mm probe, the former probe was used to obtain solution spectra when sample quantities were limited, as was the case with the solutions of bovine nasal aggregate. The sensitivity of the 6-mm probe was 40 to 45% that of the lo-mm probe.
Sample temperatures in the Dewared probes were regulated with a Varian temperature controller and were measured with a copperconstantan thermocouple.
The temperature of intact cartilage was measured (in the nmr probe) on a sample having a concentric hole into which the thermocouple was inserted. Data Acquisition -Free induction decay signals were accumulated in a NIC-80 data system (12 K total memory) equipped with a NIC-293 controller and a Diablo disc drive. The disc interactive Nicolet program NTCFT, version 1002, was used to perform the usual mathematical operations on the FID. Data were obtained in single channel and in the quadrature modes. The latter provided ~'2 times better sensitivity than single channel operation. The images in the quadrature mode were seldom more than a few per cent; hence, the software correction to orthogonalize the two FID signals was seldon employed. Software-controlled pulse widths and delays were generated by timers in the NIC-293.

Absolute
Intensity Measurements -Throughout the course of these experiments, the gain of the spectrometer was monitored by periodic measurement of the signal intensity of a neat ethylene glycol sample having a height equal to that of the ':'C coil. The integrated signal intensity/scan varied less than 10% over the time period of the experiments.
Before each measurement, the input impedance of the probe plus sample was tuned to 50 0. The integrated intensity of the power spectrum of an H,O sample was also periodically measured and compared with the glycol intensity to monitor sensitivity of the spectrometer.
Cartilage samples were prepared for absolute intensity measurements by packing the tissue to a height equal to that of the liquid in the standard glycol sample and then tuning the input impedance of the probe plus sample to 50 II. However, the line shape was sufficiently well defined to show that at least two components, broad (A -120 Hz) and and narrow (A 5 40 Hz), were present, with the broad line component predominant.
The simulation in Fig. 9o  -':'C spectra are normally obtained with a field of sufficient power (yH&r 2 2kHz) applied at the proton resonance frequency to remove the proton-carbon scalar coupling. Since dipolar interactions are about 100 times larger than scaler interactions, the latter are not affected by scaler decoupling. In contrast with scalar interactions, dipolar interactions are averaged out by fast isotropic molecular reorientations which result in spectra having line widths of a few hertz or less. Macromolecules in viscous solutions or in the solid state are often characterized by motions which are slow or spatially restricted and incomplete averaging of the dipolar interaction occurs, leading to very broad line widths. Application of a strong field, yHJ27-r -60 kHz, at the proton resonance frequency virtually eliminates line width contributions arising from static dipolar interactions as well as from motions for which 7 2 (yH,) ', Hence, application of dipolar decoupling to systems having slow or restricted motion will produce spectra with sharper lines or greater intensity, or both, than spectra obtained using scalar decoupling (17)(18)(19)(20)  that motion of the glycosaminoglycan chains is more restricted in the tissue, since 13C line widths decrease with increasing chain mobility. However, the fact that the line widths observed in the tissue are 5100 Hz implies that the vast majority of correlation times characterizing reorientation of the C-H vectors in the glycosaminoglycan are less than 1 ps (13). This statement applies only to that fraction (f,) of the glycosaminoglycan chains which contribute intensity to the spectrum. In order to obtain a quantitative measure off,, one must obtain a spectrum under conditions in which the integrated intensity of the nmr signal is proportional to the number of carbons which contribute to the signal. The spectrum observed for bovine nasal cartilage in Fig. 2c was obtained under such conditions, i.e. the time between 90" pulses was 8 times larger than the T, values of the protonated carbons, and the NOE was suppressed using gated decoupling. Furthermore, the spectrum was obtained in less than 1.2 h on tissue freshly isolated as described under "Experimental Procedures," conditions which minimize endogenous proteolysis. Hence, we think that the intensity of the protonated carbons in the 80 to 140 ppm range, Fig. 2c, is proportional to the number of mobile carbons in the backbone of the glycosaminoglycan chains in the native tissue.
Integrated intensities were measured for four separate samples which were subsequently dried and weighed (Table II). The integrated intensities were converted into weight of 'V by calibrating the instrument with a sample of neat ethylene glycol to determine the integrated intensity/scan/mg of 13C as described under "Experimental Procedures." The weight of mobile glycosaminoglycan was then readily calculated from the chemical structure and the weight of 13C in the backbone of the glycosaminoglycan chains. The integrated intensity accounts for all the glycosaminoglycap, measured by weight, in the tissue (Table II). Although this result suggests that all the glycosaminoglycan chains in the tissue are mobile (f, = l), we emphasize that the uncertainty in f, is large, -~20%, primarily because of uncertainties in the shape of the base-line (see under "Experimental Procedures"). Because of the large uncertainty in f,, two other methods were used to estimate the fraction of mobile glycosaminoglycan chains in the tissue. First, intensities from a spectrum of a fresh tissue sample were compared with intensities obtained after the tissue had been treated with papain; and, second, the intensities of a fresh tissue were measured at several temperatures. The line widths were significantly smaller in the papain-treated spectrum, Fig. 3b, resulting in greater peak heights. The integrated intensities in the 80 to 120 ppm range, however, were the same as found in Fig. 3a, within experimental error, 20%. The papain-treated sample exhibited new signals in the 120 to 190 ppm range arising from protonated carbons in flexible peptides released by the papain digestion. When the digest was continued for 12 h at 50", substantial increases in peptide resonance intensities, Fig. 3c, were observed. These resonances appeared because most of the noncollagenous protein and some collagen in the tissue were digested under these conditions. The 80 to 120 ppm region of the spectrum, which contains resonances due to the glycosaminoglycan chains, however, was not changed by the higher temperature digestion. Fig. 4 contains spectra at 4" and 56" of the fresh tissue sample whose spectrum at 37" is shown in Fig. 2b. Comparison of Fig. 2b and 4b shows the expected sharpening of spectral lines at the higher temperature as well as peptide resonances in the 120 to 190 ppm range. The latter result is a consequence of temperature-accelerated endogenous proteolysis or thermal denaturation, or both. Significantly, the integrated intensity of the glycosaminoglycan resonances in the 80 to 120 ppm range is the same at 56" as at 37". At 4" the line widths are so large that the uncertainty in the intensity measurement is about a factor of 2; hence, quantitative comparison with intensity measured at 37" is not possible. Taken together, the above results provide strong evidence that at least 80% of the glycosaminoglycan chains contribute intensity to the nmr spectrum in native bovine nasal cartilage and hence are mobile.  Table   III contains a list of the T1, A, and NOE values determined for the C-H carbons of the glycosaminoglycan chains in the native tissue and in soluble chondroitin I-sulfate chains isolated from rat chondrosarcoma proteoglycans as described under "Experimental Procedures." The T, values were obtained from inversion-recovery spectra (Fig. 51, the line widths by computer simulation of the observed spectra, and the NOE values by comparing intensities of spectra obtained first with continuous and then with gated proton decoupling (e.g.  2b with 2~). All carbons in the 80 to 140 ppm region in each spectrum had the same T, values and the same NOE values within the quoted experimental uncertainties. These parameters can be related to correlation times characterizing glycosaminoglycan chain motion provided that (a) the relaxation mechanism is known and (b) a model for the chain motion is assumed. At 15 MHz both theory and experiment show that the Y!-'H dipole-dipole interaction is the overwhelming mechanism for relaxation of protonated carbons (29). As illustrated in Fig. 6, the relationship between correlation times and the nmr parameters depends upon the motional model assumed. The solid curues in Fig. 6 were calculated assuming that reorientation of the internuclear C-H bond axis is isotropic and is characterized by a single correlation time. If reorientation proceeds by rotational diffusion, then 7 = 1/(6Rl, where R is the rotational diffusion coefficient. A more complicated situation occurs when two or more diffusion constants are required to describe the motion. The reorientation of an axially symmetric rigid body, such as a cylinder or an ellipsoid of revolution, is described by two diffusion constants, R, and RO, which characterize, respectively, the rates of reorientation parallel and perpendicular to the symmetry axis. Woessner (30) derived expressions relating the nmr parameters to the diffusion constants and 8, the angle between the C-H axis vector and the rotational symmetry axis. The dashed curues in Fig. 6 illustrate the results of this calculation for 0 = 60" and RJR, = 10. When 13 = 0", the C-H vector is parallel to the symmetry axis and the solid curves, Fig. 6, apply, since reorientation is completely characterized by the single diffusion constant Ri? = 11~~. A third motional model, which has often been applied to the analysis of dielectric and nmr relaxation data of polymers, assumes that isotropic reorientation is characterized by a distribution of correlation times rather than a single correlation time (27,28,31,32). A distribution of correlation times describes more realistically the reorientation of a polymer backbone, since a polymer can have many conformations, each characterized by a different correlation time. To describe the distribution of chain motions, we will use a normalized log x2 distribution introduced by Schaefer (27)  The nmr parameters plotted as a function of i in Fig. 6 (dashed-dot curves) were calculated by means of a distribution, having p = 14 and b = 1000, which had been used to analyze nmr data obtained for synthetic polymers (27) and elastin (29). These parameters yield a very broad asymmetric distribution of correlation times as seen in Fig. 7 Fig. 6 apply for anisotropic reorientation described by two diffusion constants, R, and R,, where R, = 10 R, and the angle between the C-H bond axis and the symmetry axis equals 60". As is seen from column 3, Table IV   Perrin's theory of reorientation of rigid ellipsoids of revolution predicts that a rod-like particle having R, = 10 RP has a long axis, a,, equal to 6.502 where a2 is the diameter of the ellipsoid (30, 34). Thus the nmr data is compatible with the presence of local rod-like glycosaminoglycan conformations having axial dimensions whose ratio is 6.5.
A substantially larger axial ratio is compatible with the measured parameters provided that backbone carbon 0 values assume a small range. However, the precise axial ratio cannot be obtained since the range of 8 is unknown.
Analysis of Glycosaminoglycan Linewidths -Although the T, and NOE values calculated using the log x2 distribution (p = 14, b = 1000, ? = 65 ns) agree with the measured parameters, the calculated line width, 500 Hz, is 5 to 25 times larger than the glycosaminoglycan line widths obtained from the cartilage spectrum, Table III. This result indicates that the long tail of the log x2 function overemphasizes the large correlation times. Accordingly, a 100 Hz line width is calculated (+ = 65 ns) when correlation times ~500 ns are deleted from the distribution by truncating the distribution at (b) in Fig. 7. Thus, deleting these larger correlation times which comprise only 13% of the distribution yields a calculated line width which equals that found for the broad component of the glycosaminoglycan spectrum. Correlation times >250 ns comprise 20% of the distribution and their deletion (truncation at (a) in Fig. 7) yields a 35 Hz calculated line width, in approximate accord with the 20 Hz line width estimated for the narrow component of the glycosaminoglycan spectrum. In each case truncating the log x2 function had virtually no effect (~2%) on calculated T, and NOE values, since these parameters are determined by correlation times in the neighborhood of 10 ns.
Although the truncated log xp distribution accounts for the line widths, this function is deficient in two respects. First, slow motions (T >500 ns) are completely neglected whereas dipolar decoupling experiments (discussed below) show that such motions are present. Second, all motions are assumed to be isotropic. In spite of these deficiencies we have used the truncated log x2 function since it illustrates two important features of the chain motion: (a) a broad distribution of 7 values in the range of the Larmor frequency is needed to account for T, and NOE values, (b) the large correlation times, although comprising a minor fraction of the distribution, determine the line widths.
This latter point and the result that at least two line widths must be used to simulate the glycosaminoglycan spectrum, Fig. 9u, is strong evidence that the distribution of slow motions is not the same for all backbone carbons in a glycosaminoglycan chain. Dynamic heterogeneity is also suggested by the chemical structure of the proteoglycan molecule, since one end of each glycosaminoglycan chain is free and presumably more mobile than the other end of the chain which is attached to the core protein. In particular, one would expect increasing resistance to the cooperative slower motions involving longer chain segments as one moves inward from the free end of the chain.
Dipolar decoupling greatly reduces the line width heterogeneity, Fig. 8 a and b. Corrected line widths of 40 Hz (80% of the signal) and 20 Hz (20% of the signal) are obtained from the computer simulation, Fig. 9b, of the dipolar decoupled spectrum. This result strongly supports the idea, discussed above, that the slow motions of glycosaminoglycan chains in intact cartilage are heterogeneous. The sharpened glycosaminogly-can lines, Fig. ti, show that the dipolar broadening of the scalar decoupled spectrum, Fig. 6u, due to slow motion is reduced when the dipolar interaction is suppressed by high power decoupling.
In addition to sharpening the lines ofthe glycosaminoglycan carbons, dipolar decoupling dramatically increases the intensity in the 120 to 190 ppm range. The large background signal seen in this region, Fig. 3b, is due to collagen carbons, which have lines that are too broad to detect (>3 kHz) using low power decoupling. The fact that a large collagen signal is obtained without cross-polarization with a 2-s interval between 90" pulses implies that at least the side chains of the collagen in the tissue undergo rapid internal motion. Hence, the type II collagen in the bovine nasal cartilage (35) exhibits mobility which is like that,observed (20) for type I collagen in tendon.
Mobility of Proteoglycan Protein-The nmr data provide strong evidence that the glycosaminoglycan chains in the intact tissue are segmentally mobile. However, we were unable to obtain unambiguous information about the molecular dynamics of the proteoglycan protein since we could not distinguish resonances of the proteoglycan protein from those due to the other, more abundant proteins (  Fig. 1Oc with Fig. 10d shows that papain digestion did not affect the oligosaccharide resonances, but new resonances (shaded in Fig. 10d)