|Home | About | Journals | Submit | Contact Us | Français|
Sodium multiple-quantum filtered (MQF) NMR spectroscopy may potentially be used to measure proteoglycan (PG) depletion in cartilage caused by osteoarthritis (OA). The purpose of this work was to quantify the effect of interleukin-1 (IL-1β)-induced macromolecule depletion on the residual quadrupolar interaction (RQI) of sodium in bovine cartilage plugs.
Fifteen 8-mm-diameter cartilage plug specimens were cored from the articular surface of fresh bovine patellae. All plugs were kept in culture media and nine of the plugs were subjected to interleukin-1 (IL-1β)-induced degeneration of cartilage for 4, 6, and 7 days. Sodium NMR spectra were obtained from each sample with a 1-cm-diameter solenoid coil in a 2T whole-body magnet interfaced to a custom-built spectrometer. We employed a previously described theoretical model to analyze triple-quantum filtered (TQF) and double-quantum filtered magic angle (DQFMA) spectra obtained from normal cartilage and cartilage treated with IL-1β. The model assumes a static Gaussian distribution of the RQI frequency, ωQ, in the sample. TQF and DQFMA spectra from each sample were fitted with the appropriate signal expressions to determine σ (the root mean square (RMS) ωQ), T2f, and T2s. An inversion-recovery sequence was used to determine T1 of each plug. A spectrophotometric assay was used to determine the amount of PG depleted from each plug. Histology was performed to visualize the PG loss in cartilage plugs. We defined σ as the measure of changes in macroscopic order in the tissue.
Simulated spectra from the theoretical model were in excellent agreement with the experimental data. We were able to determine the relaxation times as well as σ of each specimen from their corresponding fits. T2f ranged between 2.26–3.50 msec, decreasing with increased PG loss. Over the range of PG depletion investigated, T2s increased from 12.3 msec to 14.9 msec, and T1 increased from 16 msec to 21 msec, while σ decreased from 180 Hz to 120 Hz. The order of macromolecules in the cartilage tissue decreased substantially with PG loss. Histology sections clearly showed qualitative visualization of the PG loss in cartilage following treatment with IL-1β.
We demonstrated that IL-β-induced macromolecule depletion in cartilage not only changes the relaxation characteristics of sodium but also changes RQI of the tissue. Using MQF sodium spectroscopy we quantified the changes in σ and showed that loss of macromolecules reduces the degree of order in the tissue.
The early stage of osteoarthritis (OA) is associated with the preferential loss of proteoglycans (PGs) from the extracellular matrix (ECM) of cartilage (1,2). PGs maintain a fixed negative charge in the ECM and attract sodium ions and water molecules. PG loss is associated with a loss of sodium in cartilage (3). Recently it was shown (4–7) that sodium NMR can be used to measure the PG content of cartilage.
In cartilage, sodium ions are in a slow motion regime (ωOτC ~ 1, where ωO is the Larmor precession frequency and τC is the correlation time), which results in biexponential relaxation rates (8). Furthermore, cartilage, like most biological tissues, is heterogeneous in nature. However, some order exists at the local level. This ordering is primarily due to the anisotropic arrangement of collagen fibril bundles, PG, and other macromolecules in the ECM. The ordering in the ECM in cartilage produces a non-zero electric field gradient (EFG). The interaction of this EFG with quadrupolar nuclei, like 23Na, results in a residual quadrupolar interaction (RQI). The quadrupolar interaction is dependent on the orientation of the EFG to the external static magnetic field, B0. Recently, the presence of RQI in cartilage was demonstrated by sodium multiple-quantum filtered (MQF) NMR spectroscopy (9). The magnitude of quadrupolar interaction frequency (ωQ) represents the degree of order of the macromolecules in the tissue. Hence it is possible to probe this parameter, which in turn may provide information about the structural integrity of biological tissues. Sodium MQF-NMR spectroscopy and imaging of biological tissues have been performed both in vivo (10–12) and in vitro (8,9,13,14). MQF-NMR spectroscopy is uniquely suited for measuring ωQ, as well as biexponential relaxation times.
It was previously shown (15), using isolated model systems of cartilage ECM components, that the contribution to RQI in cartilage is due to ordering of collagen bundles and that PG does not contribute to the RQI. Although these were meticulous studies, since they were performed on isolated collagen and PG molecules, it is difficult to determine the effects of PG on the RQI in the presence of collagen in intact cartilage. This can only be accomplished by selective depletion of PG from cartilage using an appropriate enzyme or cytokine. Interleukin-1 (IL-1β) is a cytokine that enhances formation of metalloproteinases by chondrocytes. Proteinases are responsible for PG depletion in cartilage both in vivo (16,17) and in vitro (18,19). An increased level of IL-1β has been observed in the early stages of OA (20). It was shown previously (21) that sodium relaxation rates change with IL-1β-induced cartilage degeneration. However, this simplistic model did not incorporate ωQ into the calculations.
In this work, we investigated the effect of degradation of the ECM of articular cartilage with IL-1β on ωQ, the short (T2f) and long (T2s) components of the transverse relaxation time, as well as the longitudinal relaxation time (T1) of sodium. The magnitude of σ (the RMS ωQ) was used as a measure of the ordering of macromolecules in the tissue. We demonstrated that the macromolecule depletion induced by IL-1β resulted in changes in the local order of the ECM, which was reflected not only in the relaxation times but also in the σ.
The spin dynamics of quadrupolar nuclei, in biological tissues, can be determined by forming the appropriate interaction Hamiltonian and computing the response to a given pulse sequence. The Hamiltonian for the combined Zeeman interaction and a much smaller quadrupolar interaction can be expressed as:
Where ÎZ is the nuclear spin angular momentum operator, eQ is the nuclear quadrupole moment, and eq is the principal value of the residual EFG along its major axis. Angles θ and are the polar and azimuthal angles, respectively, between the orientation of the EFG and the static magnetic field, B0, and η is the asymmetry parameter of the EFG.
Several models have been used to analyze MQF-NMR spectra of spin-3/2 in biological systems (14). Eliav and Navon (15) showed that a model that assumes a Gaussian-weighted distribution of ωQ values provided the best fit for the MQF-NMR signal from fresh bovine nasal cartilage. This model provided the best fit for Jeener-Broekaert spectra obtained from bovine articular cartilage as well (22). In the present investigation, we used a similar model that assumes a static Gaussian distribution of ωQ to describe the sodium MQF-NMR signal from bovine cartilage (23,24). In this model, the tissue is comprised of several volumes called “domains.” The average volume of the domain is greater than the average volume traversed by a diffusing sodium ion within its coherence lifetime. Within each domain, the macromolecules are nonrandomly arranged and interact anisotropically with fluid and hydrated ions. As a result there exists a residual EFG, of magnitude eq and direction that is determined by the orientation of the domain and the asymmetry parameter. Each domain has a quadrupolar interaction frequency, ωQ, given by:
If the sample is a liquid crystal, where all the domains have the same residual EFG and are oriented parallel to each other, we expect an NMR spectrum with three narrow lines. A random orientation of the EFG in the different domains will result in a Pake powder pattern NMR spectrum. Biological tissues have a wide range of domain orientations, and a range of EFG and η values. Therefore, a Gaussian distribution of ωQ is assumed for such poorly ordered systems. Within this model, the equations for the sodium MQF-NMR signals have to be appropriately weighted by a Gaussian distribution of ωQ as given by the function:
The Gaussian distribution is centered on zero, i.e., the mean of the quadrupolar interaction frequencies, Q = 0 and σ is defined as the RMS of ωQ values over all domains.
In this study, we are particularly interested in the double-quantum filtered magic angle (DQFMA), triple-quantum filtered (TQF) signals. The DQFMA signal is generated by sodium in purely anisotropic environments, whereas the TQF signal originates primarily from sodium in isotropic slow motion environments (9,25,26). The real part of the TQF signal from a spin-3/2 system with a static Gaussian distribution of ωQ is given by:
The corresponding DQFMA signal is:
Where t is the detection time (begins immediately after the first pulse), τ is the preparation time, and the variables s1 and s2 are the transverse relaxation rate constants, 1/T2f and 1/T2s, respectively. Equations  and  were used to fit the sodium MQF-NMR signal obtained experimentally. The fits were used to determine σ and the relaxation time constants, T2s and T2f.
Samples of bovine articular cartilage were obtained from a local slaughterhouse (Bierig Bros., Vineland, NJ) within 3 hours after slaughter. Using a sterile cork borer, 15 8-mm-diameter, cylindrical cartilage plugs were cored from the same region in 15 different patellae. IL-1β treatment was based on a previously described technique (27). All chemical compounds were obtained from GIBCO BRL (Gaithersburg, MD) unless otherwise indicated. Samples were washed in phosphate-buffered saline (PBS) (Sigma Chemicals, St. Louis, MO) containing 100 units/mL penicillin/streptomycin, 0.25 μg/mL amphotericine B, and 1 mM hepes. The samples were separated into three groups of five plugs and each group was placed into sterile six-well culture plates for 24 hours. These plates contained Dulbecco’s Modified Eagle’s Medium (DMEM) with 10 μg/mL of insulin, 10 μg/mL of transferrin, 100 units/mL of penicillin, 100 μg/mL of streptomycin, 0.25 μg/mL amphotericine B, 2 mM L-glutamine, and 50 μg/mL ascorbic acid (Sigma Chemicals, St. Louis, MO). After equilibration, the medium was removed and the cartilage plugs were placed in fresh medium. Three plugs from each group were subjected to a solution of DMEM containing 20 ng/mL of IL-1β (Sigma, St. Louis, MO), while the other two samples were placed in normal DMEM solution to serve as controls. The medium was refreshed every 3 days. Samples from the first group were removed from culture after 4 days for NMR measurements. Likewise, the second and third groups were removed from culture after 6 and 7 days, respectively, and the NMR experiments were performed. In this way, samples with different degrees of PG loss were obtained. Media from each sample were saved for further analysis to determine PG loss from the plugs in culture. The samples were placed in a sealed tube for the NMR experiments to avoid dehydration. The position of the cartilage plug inside the magnet was such that the normal to the articular surface was perpendicular to the static magnetic field, B0.
All NMR experiments were performed on a 2T whole-body Oxford magnet attached to a custom-built pulse-programming computer interface. A seven-turn, 1-cm-diameter solenoid coil tuned to 23Na at 2T (22.7 MHz) was used to obtain all sodium NMR spectra. Single quantum spectra were obtained with a standard pulse-acquisition sequence. A typical 90° pulse length was 25 μs. Sodium T1 was measured from spectra obtained with a standard inversion recovery (IR) sequence with 20 inversion times ranging from 2 to 40 msec. Multiple-quantum filtering was performed using pulse sequences described by Reddy et al (28). Sodium TQF spectra were used to determine T2s, and DQFMA spectra were employed to calculate T2f and σ.
TQF spectra were obtained with the pulse sequence shown in Figure 1A. The pulse sequence consists of four radiofrequency (RF) pulses. The first pulse is a 90° pulse, which produces single-quantum coherence. It is followed by a 180° refocusing pulse after a time τ/2 (half the preparation time). During the preparation time, the magnetization evolves in the presence of quadrupolar relaxation as well as the RQI (15). Two more 90° pulses separated by a time δ (evolution time) are applied another τ/2 period after the refocusing pulse. The first of these two pulses creates multiple-quantum coherences (triple-quantum in this case), which are then converted by the second pulse to odd-ranked tensors ( ), which evolve into the observable single-quantum signal ( ). The RF pulses were phase-cycled with six RF phases ( = 30°, 90°, 150°, 210°, 270°, and 330°) while the acquisition phase was alternated between 0° and 180°. The six phase-cycled signals were then combined to obtain one TQF spectrum. In the measurement of relaxation times the τ time was varied 18 times from 0.4 to 35 msec and the δ time was kept as small as possible (400 μs). Each spectrum was averaged 100 times to enhance the signal-to-noise ratio (SNR). The pulse sequence repetition time (TR) was 100 msec, so the total acquisition time was 18 minutes for each sample.
DQFMA spectra were obtained with the pulse sequence shown in Figure 1B. It is similar to the TQF pulse sequence except that the last two pulse flip angles are set to 54.7°. In this sequence, only the double-quantum coherence due to even-rank tensor ( ) will be detected. This coherence purely arises from the presence of RQI. The RF pulses were phase-cycled with four RF phases ( = 0°, 90°, 180°, and 270°). Spectra were obtained by varying the preparation time from 0.2 to 10 msec in 15 steps, while the evolution time was fixed at 400 μs. As with the TQF pulse sequence, the acquisition phase was alternated between 0° and 180°. Each DQFMA spectrum was averaged 300 times. The TR time was 100 msec for the DQFMA sequence, resulting in a total acquisition time of 30 minutes per sample.
After the NMR experiments, the amount of PG depleted from each plug was determined by subjecting the degradation media of each plug to a dimethylmethylene blue (DMMB) spectrophotometric assay (29). To determine the total PG content, the entire cartilage plug was digested with 500 μg/mL of papain for 24 hours after the NMR experiments. A solution of DMMB was prepared by mixing 8 mg of DMMB in 496.5 mL of water containing 1 g of sodium formate, 2.5 mL ethanol, and 1 mL formic acid (pH = 3.6 and A535 = 0.47). Absorbances at 535-nm wavelength were obtained with a Spectronic 20D+ spectrophotometer (Spectronic Instruments, Rochester, NY). Calibration was performed using 0–30 μl of 1 mg/mL of chondroitin sulfate C (Sigma Chemicals, St. Louis, MO) with additional PBS to make up 50 μl in a 4.5-mL methacrylate cuvette. Two mL of the DMMB solution were added to each standard and the % transmittance was read immediately. The standard curve was plotted using the absorbance of each standard (calculated from the % transmittance by Beer’s Law). Reagent blanks were used to recalibrate the spectrophotometer before each set of samples was read. Samples of 10 μl of the degradation media (and 2 μl of the papain digested media) with additional PBS to make up 50 μl were created. Two mL of the DMMB solution were added to each of the unknowns, and the transmittance was read. The amount of PG was determined from the absorbance and the standard curve. The total PG from each sample was determined by adding the PG in the degradation medium and the PG in the corresponding cartilage plug (obtained by digestion with papain).
To determine the effect of IL-1β degradation on collagen content in cartilage, in a separate experiment another nine cartilage plugs were cultured in the exact same way as the original 15 plugs. Then the cultured media and the papain digests were subjected to both PG assay (as described above) and collagen assay. Common-wealth Biotechnologies Inc. (Richmond, VA) performed the collagen assays in the following way. Two hundred μl of each sample were transferred to a pyrolyzed glass tube. The samples were taken to dryness and hydrolyzed in 1 mL of 6N HCl for 20 hours at 110°C. Following hydrolysis, the samples were again taken to dryness and dissolved in 200 μl of sample loading buffer. The amino acid content of each sample was then analyzed via chromatograms.
Histology was performed on three controls and three samples treated with IL-1β. Each cartilage plug was fixed with 10% formalin, decalcified for 2 days, and embedded in paraffin wax. A microtome (Microm, Germany) was used to slice 5-μm-thick sections across cartilage. Histology sections were mounted on slides and stained with Safranin-O (Sigma, St. Louis, MO) with a Fast Green counterstain to distinguish normal and PG-depleted cartilage tissue, respectively. The slides were examined under an SV11 stereo microscope (Carl Zeiss Inc., Thornwood, NY) and images were captured into the computer using SCION Image software (Scion Corp., Frederick, MD).
NMR spectra were processed using spectral-processing routines written in IDL (RSI, Boulder, CO) data processing language. An exponential-weighted filter to smooth the data was applied to the raw free induction decay signals. Then the corresponding spectrum was obtained by a Fourier transform. A graphical-user-interface-based program was written in IDL to fit DQFMA and TQF spectra. Spectra for each τ time were phased and then fitted to the Fourier transform of either Eq.  or  using a nonlinear, Levenberg-Marquardt-based algorithm. Statistical analysis was performed with the JMP software package (SAS Institute, Cary, NC) to determine the P-values of the correlation coefficients (R2) between the observed values and the values predicted by the model.
The DQFMA spectra were used to calculate sodium T2f and σ. The T2f and σ values for each plug were obtained by averaging the T2f and σ obtained by fitting the eight spectra (of the 15 spectra per plug) with the highest SNR. With these two numbers as input, the TQF spectra were used to calculate sodium T2s. The T2s for a plug was obtained by averaging the T2s of the 10 (of 18 spectra) highest-amplitude spectra.
Sodium T1 was calculated by fitting the IR spectral amplitudes (S(TI)) as a function of inversion time (TI) to the equation:
Where So is the signal intensity of the equilibrium magnetization and C is the baseline offset. A nonlinear, least-squares algorithm was used to determine the best fit.
Although IL-1β-induced macromolecular degradation resulted in a loss of collagen, we found that the percentage of collagen lost (as measured by collagen assay) was significantly less than the percentage of PG loss (as measured by PG assay) from each plug (Fig. 2). Furthermore, there is no significant change in the amount of collagen loss as a function of concentration of IL-1β used in our degradation method. However, we clearly observed an increased PG loss with increasing IL-1β degradation. As a result, the sodium NMR parameters (T2s, T2f, T1, and σ) are plotted as functions of PG loss by the specimen and not the percentage of collagen loss in the subsequent graphs.
Figure 3 shows typical sodium TQF spectra (dotted line) and fit (solid line) obtained at τ = 5 msec from bovine articular cartilage plugs degraded with IL-1β. The corresponding DQFMA spectra for τ = 800 μs for the same two specimens are shown in Figure 4. The fits were obtained with Eqs.  and . The control specimen in A lost 6% of its total PG (probably from autodigestion of PG) resulting in σ = 177 Hz, T2f = 3.35 msec, and T2s = 12.7 msec, as calculated from the fits. The IL-1β-degraded specimen in B lost more than half of its total PG; σ = 120 Hz, T2f = 2.48 msec, and T2s = 14.8 msec were the parameters for the fits. The TQF spectra are normalized to their corresponding single-quantum integrated spectral areas.
The difference between the two samples is more obvious when we plot the spectral amplitudes as a function of preparation time, τ (Fig. 5). The DQFMA spectral amplitudes (A) for the IL-1β-degraded specimen reach a minimum value at τ = 1.2 msec, whereas for the control sample the minimum is reached 400 μs later. The longer T2s for the IL-1β-degraded sample is evident by the slower decay of its TQF spectral amplitudes (B).
Histological sections from a control and an IL-1β-degraded specimen are shown in Figure 6. Normal cartilage (with higher PG content) appears red, whereas PG-depleted cartilage is blue with the staining used in this study. The slides illustrate that PG loss is nonuniform across the cartilage. There is more loss at the articular and chondral surfaces and less PG loss in the middle of the tissue.
The variation in relaxation times as a function of PG loss from the cartilage plugs is shown in Figure 7. A spectrophotometric assay to determine PG content was performed only on nine specimens. The other six samples were used for histology. The first three data points (corresponding to a PG loss of 6%, 6%, and 8%) are all control specimens. The PG loss in the controls is a result of autodigestion of cartilage macromolecules while in the culture medium. Figure 7 shows that there is a significant increase in T1 with PG loss (P < 0.01). T1 ranged from 16 msec for a 6% PG loss to 21 msec for a 52% PG loss. T2s was 12.3 msec for a 6% PG-depleted specimen and increased with PG depletion, resulting in a time of 14.9 msec for the specimen with 50% PG loss. Statistical analysis determined this trend to be significant (P < 0.01). The short component of the transverse relaxation time, T2f, decreases significantly (P < 0.01) with increasing PG depletion (Fig. 8). The longest T2f was 3.50 msec for the 8% PG loss specimen and decreased to 2.26 msec for the specimen with 50% PG loss. As explained in the Data Processing section, for every plug σ, T2f, and T2s values were obtained by averaging these parameters from several fits of spectra obtained at different τ times. Measurements of relative error of the fitting procedures, which can be obtained from the standard deviations of these values, were less than 5% of their average value in each plug.
Figure 9 plots σ as a function of PG loss by the specimen. The quadrupolar interaction frequency remains fairly constant until a 25% loss of PG. Beyond this point, σ decreases by approximately 20% with increasing PG loss. Therefore, there is a substantial decrease in order with PG loss.
It was previously shown (15) that collagen alone is responsible for σ changes in cartilage. This was determined from the fact that a collagen suspension produced a DQF spectrum, whereas a suspension of PG did not. This may be the case for isolated ECM constituents. However, as shown here in intact cartilage, PG loss also leads to change in the quadrupolar coupling constant, and hence the order of macromolecules in the tissue. This may be due to a decrease in the repulsive forces between negatively charged PG molecules in cartilage that influence the orientation of the collagen fibrils. Therefore, in intact cartilage, it is possible that the PG loss may be affecting the order either by influencing the collagen ordering or by changing the organization of both collagen and PG.
In conclusion, we have shown that in intact cartilage, cytokine-induced degradation changes not only sodium relaxation characteristics but also ωQ, and hence the order of macromolecules in the tissue. The results showed that T2f decreased whereas T2s and T1 increased with increasing PG loss from cartilage. There was a significant decrease in order (σ) with PG loss. Contrary to the results obtained from isolated model systems of ECM of cartilage, in intact cartilage PG also contributes to the ωQ and hence the order of macromolecules in the tissue. These results also demonstrate that MQF spectroscopy is uniquely suited for determining the macromolecular order in tissues in a nondestructive fashion. This measurement provides indirect information about structural changes in the tissue in response to artificially induced or natural pathology. To the best of our knowledge, no other method provides this information in a noninvasive, nondestructive manner. In conjunction with currently available MQF-MRI techniques, it should be possible to quantify changes in sodium MR parameters in cartilage pathology in vivo.
Contract grant sponsor: NIH; Contract grant number: RR02305; Contract grant sponsor: National Institutes of Arthritis, Musculoskeletal, and Skin Diseases (NIAMS); Contract grant numbers: R01-AR45242; R01-AR45404.
We thank Professor John S. Leigh for his encouragement and support.