Sensitivity of MRI to Proteoglycan Depletion in Cartilage: Comparison of Sodium and Proton MRI A

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Osteoarthritis and Cartilage (2000) 8, 288–293 © 2000 OsteoArthritis Research Society International 1063–4584/00/040288+06 $35.00/0 doi:10.1053/joca.1999.0303, available online at http://www.idealibrary.com on

Sensitivity of MRI to proteoglycan depletion in cartilage: comparison of sodium and proton MRI A. Borthakur, E. M. Shapiro*, J. Beers, S. Kudchodkar, J. B. Kneeland and R. Reddy Departments of Radiology and *Chemistry, University of Pennsylvania, Philadelphia, PA, U.S.A.

Summary

Objective: The purpose of this work was to evaluate the results from sodium and proton magnetic resonance imaging (MRI) in detecting small changes in proteoglycan (PG) content in bovine articular cartilage specimens. Design: Articular cartilage from 15 specimens of bovine patellae were subjected to partial PG depletion with different concentrations of trypsin for 30 min. Sodium and proton MR images of the intact specimen were obtained on a 4T GE clinical MRI system. Two custom-built 7 cm-diameter solenoid coils tuned to proton and sodium frequencies were employed. Fast gradient echo and spin echo imaging sequences

were used to determine sodium density, proton density and proton relaxation times (T1 and T2) of the specimens. Spectrophotometric assay was performed after MRI to determine PG concentrations of the cartilage specimens. Results: The sodium signal change correlated well with the observed PG loss (R 2=0.85, P<0.01) whereas the proton signal change was 2 2 inconsistent (R =0.10, P<0.8). The change in proton T1 and T2 between the two regions did not correlate with PG loss (R =0.07 and R2=0.06, respectively). Conclusions: Results from these studies demonstrate that sodium MRI is both sensitive and speciﬁc in detecting small changes in PG concentration, whereas proton density and relaxation properties are not sensitive to small changes in PG content. © 2000 OsteoArthritis Research Society International Key words: Sodium MRI, Proteoglycan, Articular cartilage, Osteoarthritis.

Introduction CURRENT TECHNIQUES IN PROTON MRI OF CARTILAGE Osteoarthritis (OA), a progressive disease of articular DEGENERATION cartilage, is associated with severe joint pain and results Cartilage contains a signiﬁcant amount of water (∼80% in in joint immobilization. Currently, there is no cure for OA healthy tissue) with a relatively short transverse relaxation but early detection followed by efficient therapy can slow time, T . This water is the ‘observable’ proton signal in an its detrimental effects. Articular cartilage mainly comprises 2 MR image of cartilage. Most techniques rely on short echo an extra-cellular matrix of collagen (type-II), proteoglycans time pulse sequences to image cartilage.3 Although (PG) and water. It has been observed that the early stage changes in cartilage thickness (measured from proton MR of OA is primarily associated with loss of PG with little images) has been attributed to changes in PG content, no change in collagen.1,2 Histology can be performed to consistent changes in proton density and relaxation times accurately assess PG loss in vitro. However, to diagnose were observed in the rabbit knee.4 Recent work has shown the early stages of OA and monitor therapy in vivo,a that there is no correlation between PG distribution and noninvasive technique that is sensitive to PG loss in proton T in articular cartilage.5 To date, proton MRI is cartilage is required. Magnetic resonance imaging (MRI), 2 inconclusive in measuring early degenerative changes.6,7 being a noninvasive technique, is an obvious choice for Contrast-enhanced techniques have been employed to studying cartilage. Numerous MRI methods have been study cartilage degeneration.8–11 Some of these contrast employed to study cartilage structure and degeneration agents could not be used in humans due to their toxicity. both in vitro and in vivo. These strategies can be broadly However, a promising new technique demonstrated that classiﬁed into two categories: (1) proton and (2) sodium Gd(DTPA)2−-enhanced proton MRI can be used to MRI techniques. measure the tissue PG content in vivo.12 The spatial distribution of PG was observed in both in vivo and in vitro images of the human knee joint. Although this method has all the advantages of proton MRI, it requires proton longi-

tudinal relaxation time, T1, maps before and after injection of Gd(DTPA)2−. Potential drawbacks of this technique are Received 30 August 1999; accepted 7 December 1999. that it is time intensive and it is difficult to obtain the same Address correspondence to: Arijitt Borthakur, Department of Radiology, University of Pennsylvania, B1 Stellar Chance Lab- slice pre- and post-injection. However, if the T1 of the oratories, 422 Curie Boulevard, Philadelphia, PA 19104-6100, cartilage before injection is assumed to be a constant U.S.A. Tel: (215) 898-9357; Fax: (215) 573-2113; E-mail: throughout the tissue and a T1 map is not obtained, then a [email protected] small error in PG measurement is incurred.13 Another

288 Osteoarthritis and Cartilage Vol. 8 No. 4 289

detecting small changes in PG concentration in the tissue. We achieved this by comparing sodium density images of PG depleted cartilage specimens to their corresponding proton density images. Proton density images are a direct map of the water content of the tissue and will show changes in water content (if any) as a function of PG

depletion. We also determined proton T1 and T2 maps to measure regional variations in relaxation times across cartilage.

Methods Patellae were obtained from freshly slaughtered cows (Bierig Bros., NJ). A groove was cut on the articular surface of 15 specimens to divide the surface into two halves (Fig. 1). Each specimen was placed in a tray containing a nonpermeable divider such that the groove on the articular surface was wedged onto the divider. One side of each specimen was equilibrated in 137 mM phosphate buffered saline (PBS). The other half was immersed in a solution of PBS and three different concentrations of trypsin: 10 g/ml Fig. 1. Axial proton MR image of a specimen of bovine patella. A groove was cut into the articular surface to divide the cartilage in (N=4), 50 g/ml (N=4) and 100 g/ml (N=5) of trypsin. The half. remaining two specimens served as controls and were not subjected to trypsin treatment. The samples were left to equilibrate for 30 min. Then they were rinsed with PBS and complication is that the changes in proton signal intensity kept in PBS till the time of the imaging experiments. In this due to the contrast agent diffusing in and out of cartilage way, each patella served as a self-control specimen for the tissue while MRI is performed potentially may lead to errors imaging experiments. in PG measurement.13 After the MR, the entire cartilage from each specimen was removed with a scalpel and weighed to determine the wet weight of the cartilage. Each sample was then digested 19 CURRENT TECHNIQUES IN SODIUM MRI OF CARTILAGE with papain and subjected to spectrophotometric assay. DEGENERATION The amount of PG extruded during degradation with trypsin was determined by assaying the degradation media of During the early stage of OA, the tissue primarily loses each specimen as well. The total PG of the specimen was PG molecules. Negatively charged PG plays an important simply the sum of the cartilage and the degradation role in determining tissue integrity by maintaining a ﬁxed medium PG contents. The percent PG loss is the PG charge density (FCD) in cartilage.14 Positively charged content of the degradation medium normalized by the sodium ions are attracted by this FCD. It has been shown total PG. that the sodium content of cartilage can be correlated with All images were obtained on a 4T whole body GE Signa its FCD.15 Sodium MRI has been proposed as a direct MRI scanner (General Electric, WI). We used two (1H and method to determine PG loss in cartilage.14,16,17 Further- 23Na) 7 cm diameter solenoid radio frequency (RF) coils. more, it has been shown that sodium T1 and T2 increase 3D sodium MR images were obtained with a fast gradient with PG depletion.18 echo sequence with TE/TR=2/13 ms resulting in a total It is widely believed that PG loss leads to changes in imaging time of 13 min for 16 slices of FOV=16×16, 4 mm water content in cartilage.4 However, to date there has slice thickness with 256×64 matrix size and 60 NEX. been no systematic proton MRI study to determine its Proton density images were obtained using a standard 3D sensitivity (or speciﬁcity) to changes in water content as a fast spin echo (FSE) sequence with the following par- function of PG depletion. Since sodium density weighted ameters: TE/TR=13/6000 ms, FOV=7 cm×7 cm, 1 mm imaging provides quantitative spatial maps of PG, in this slice thickness, 512×256 matrix size. The echo time of paper we used it to validate the sensitivity of proton MRI in 13 ms was the minimum TE achievable with the available

Table I Percentage change in pixel intensity of the PG depleted to nondepleted regions from a representative slice from

the sodium images and the corresponding slice from the proton images. Also shown are average proton T1 (and T2) obtained from T1 (and T2) maps of each region Average PG loss 0% 4.5±1.1% 15.5±0.5% 18.4±2.3% MR parameter (N=2) (N=4) (N=4) (N=5)

Change in pixel intensity Sodium images 0±1 4±4 15±4 20±4 (% of nondepleted signal)H Proton images 0±5 4±12 14±17 3±4

Proton T1 (s) PG depleted 1.1±0.1 1.1±0.2 1.1±0.2 1.1±0.2 H Nondepleted 1.1±0.2 1.1±0.1 1.1±0.1 1.2±0.3

Proton T2 (ms) PG depleted 44±5 49±5 44±8 37±7 H Nondepleted 35±6 41±6 40±11 42±6 290 A. Borthakur et al.: Sensitivity of MRI to PG depletion

Fig. 2. Typical sodium density MR image of a specimen of bovine patella following partial degradation in 100 g/ml (A), 50 g/ml (B) and 10 g/ml (C) of trypsin for 30 min. Pixel size is 0.625 mm×2.5 mm in-plane and 4 mm thick.

Fig. 3. Sodium density MR image from Fig. 2 (A) and the corresponding high-resolution proton density image (B). The proton image was obtained with a 3D FSE sequence with the following parameters: TE/TR=13/6000 ms, FOV=7 cm×7 cm, 1 mm slice thickness, 512×256 matrix size and 1NEX. Pixel size is 137 m×273 m in-plane.

gradient strength. Since we are comparing two regions (PG depleted vs nondepleted) within the same specimen, the Fig. 4. Percentage change in pixel intensity (measured from MR regions will have equal T2 weighting. Proton T maps were obtained using a standard 2D images) vs PG extruded (measured by PG assay) for each 1 specimen. Sodium intensity shows a high correlation (R 2=0.85, inversion-recovery FSE sequence with the following par- 2 ameters: TE/TR= 25/6000 ms, FOV=7 cm×7 cm, 2 mm P<0.01) with PG extruded, whereas proton has a R =0.10 (P<0.8). thick slice, 512×256 matrix size. The inversion time was varied from 70 ms to 6 s. Proton T2 maps were obtained with a standard 2D-FSE sequence with 7 different Data was processed using image-processing routines echo times ranging from 15 ms to 200 ms and the same written in IDL (RSI, Boulder, CO) programming language. resolution as the T1 maps. The T1 maps were calculated by pixel-to-pixel linear ﬁtto Regions of interest in the PG-depleted and nondepleted the equation: areas of the image were manually selected. The same slice location in the sodium and proton MR images was chosen by visual inspection. The percent change in pixel intensities from the two regions of interest was determined. Average

T1 and T2 from the two regions were recorded from the Where S(TI)=pixel intensity as a function of inversion time corresponding T1 and T2 maps respectively. (TI), So=pixel intensity for a long TR (6 s), C=intercept

Fig. 5. Axial proton density MR image (A) and corresponding T1 (B) and T2 maps (C) of a specimen of bovine patella following partial degradation in 100 g/ml trypsin for 30 min. Dotted line in A indicates the location of the T1 and T2 proﬁle in Fig. 6. Osteoarthritis and Cartilage Vol. 8 No. 4 291

Fig. 6. Proﬁle of proton T and T across the cartilage of the 1 2 Fig. 7. Change in proton T s between the PG depleted and control specimen of bovine patella shown in Fig. 5A. 1 regions from the bovine articular cartilage specimens as a function

of PG loss. The change in T1 values did not correlate with PG loss (R 2=0.07). (equal to ln(2), if an IR sequence is used to measure T1). T1 is estimated from the slope of the linear fit. T2 maps were calculated by fitting the pixel intensities to an exponential decay function. Statistical analysis was performed with JMP (SAS Institute, Cary, NC) software package. Analysis of variance was used to determine if the mean of relaxation by dividing the difference between the average pixel inten- rates from PG depleted and control regions were signifi- sities in the nondepleted and PG-depleted regions by the cantly different. Similar analysis was performed to obtain average intensity of the nondepleted region. The percent P-values for sodium (and proton) intensity change between change in pixel intensities obtained from the sodium the two regions (as a percentage of nondepleted region) vs images strongly correlate with the observed PG loss (R2=0.85, P<0.01), whereas similar correlation in the pro- PG extruded. The coefficients of correlation and P-values 2 were calculated. ton intensity ratios are rather poor (R =0.10, P<0.8). The large deviations of the proton ratios indicate that changes in water content are not specific to PG concentration Results and discussion changes. Proton T1 and T2 maps from a specimen partially Data from the experiments is summarized in Table I. degraded in 100 g/ml-trypsin for 30 min are shown in Fig. ∼ Figure 2 shows typical sodium density images of speci- 5. T2 values ranged between 35–45 ms and T1 was 1 s for mens of bovine patellae following partial degradation in all samples regardless of the amount of trypsin used for

100 g/ml (A), 50 g/ml (B) and 10 g/ml (C) of trypsin for degradation. The T1 and T2 profile across the cartilage 30 min. These trypsin concentrations are chosen to change sample in Fig. 5A is shown in Fig. 6. It is interesting to note the PG content by less than 20% of the total. The images that the T1 along the surface of the cartilage is consistently represent a cross section of bovine patella where the red greater than the interior regions of the cartilage. The regions (in the sodium image) indicate articular cartilage. change in proton T1 (Fig. 7) and T2 (Fig. 8) did not correlate The difference in the signal intensity between the PG with PG loss (R2=0.07 and R2=0.06, respectively). The depleted and nondepleted regions is obvious in the sodium small variations in T1 over the volume of cartilage are images. On average, enzymatic degradation with 10 g/ml consistent with previously published results.20 of trypsin for 30 min resulted in 4±4% sodium signal loss in The lack of specificity to PG degradation of proton the PG depleted region. Degradation with 50 g/ml and imaging may be explained in two ways. Firstly, there may 100 g/ml-trypsin decreased the sodium signal by 15±4% be little or no change in water concentration with relatively and 20±4%, respectively. In the same regions, proton small change (<20%) in PG concentration. Another contrib- density images showed a signal change of 4±12%, uting factor may be the subtle changes in the proton signal 14±17% and 3±4% for the 10, 50 and 100 g/ml-trypsin resulting from inhomogeneities, background gradients etc. degraded regions, respectively. The effect of PG depletion are greater than the changes resulting from PG loss. Since on sodium and proton images is compared in Figure 3. these errors scale with the gyromagnetic ratio, , they From this image pair, clear changes in the sodium image would be four times greater for proton (H=42.57 MHz/T) and almost no change in the proton image is evident. The than for sodium (Na=11.26 MHz/T). Further experiments controls did not show an appreciable signal change in both are in progress to conclusively determine this. We have the sodium and proton images. The large standard devi- previously shown17 that it is possible to obtain in vivo ations in the proton data compared to the sodium data sodium MR images of the human knee cartilage with good should be noted. signal-to-noise ratio and resolution (Fig. 9). Given the Figure 4 is a plot of percent change in pixel intensity as sensitivity and specificity of sodium MRI to PG changes, it a function of PG extruded. Percent change was calculated could be used to monitor PG changes in vivo. 292 A. Borthakur et al.: Sensitivity of MRI to PG depletion

Acknowledgments This work was performed at a NIH supported resource center (NIH RR02305) with support from NIH grants R01-AR45242 and RO1-AR45404. We thank Professors Felix W. Wehrli and John S. Leigh for their encouragement and help.

References 1. Mankin HJ, Dorfman H, Zarins L. Biochemical and metabolic abnormalities in articular cartilage from osteoarthritic human hips II. Correlation of mor- phology with biochemical and metabolic data. J Bone Joint Surg 1971;53A:523–37. 2. van de Loo AA, Arntz OJ, Otterness IG, van den Berg WB. Proteoglycan loss and subsequent replenish- ment in articular cartilage after a mild arthritic insult by IL-1 in mice: impaired proteoglycan turnover in the recovery phase. Agents Actions 1994;41:200–8. Fig. 8. Change in proton T2s between the PG depleted and control 3. Brossmann J, Frank LR, Pauly JM, Boutin RD, regions as a function of PG loss. The change in T values did not 2 Pedowitz RA, Haghighi P, et al. Short echo time correlate with PG loss (R 2=0.06). projection reconstruction MR imaging of cartilage: comparison with fat-suppressed spoiled GRASS and magnetization transfer contrast MR imaging. Radi- ology 1997;203:501–7. 4. Paul PK, O’Byrne E, Blancuzzi V, Wilson D, Gunson D, Douglas FL, et al. Magnetic resonance imaging reflects cartilage proteoglycan degradation in the rabbit knee. Skeletal Radiol 1991;20:31–6. 5. Fragonas E, Mlynarik V, Jellus V, Micali F, Piras A, Toffanin R, et al. Correlation between biochemical composition and magnetic resonance appearance of articular cartilage. Osteoarthritis Cartilage 1998;6:24–32. 6. Recht MP, Resnick D. Magnetic resonance imaging of Fig. 9. In vivo sodium (A) and proton (B) axial MR image of the articular cartilage: an overview. [Review] [69 refs]. patellar-femoral joint of a healthy human volunteer. Both MR data Top Magn Reson Imaging 1998;9:328–36. sets were obtained on a 4T GE clinical scanner with custom-built 7. Peterfy CG. MR imaging. [Review] [53 refs]. Baillieres 7 cm diameter surface coils. Sodium image was obtained with the Clin Rheumatol 1996;10:635–78. same imaging parameters detailed in the Methods section. The 8. Bacic G, Liu KJ, Goda F, Hoopes PJ, Rosen GM, proton image was obtained with a 3D SPGR sequence with the following parameters: TE/TR=5/22 ms, FOV=7 cm×7 cm, Swartz HM. MRI contrast enhanced study of cartilage 1 mm slice thickness, 512×128 matrix size and 1 NEX. proteoglycan degradation in the rabbit knee. Magn Reson Med 1997;37:764–8. 9. Chandnani VP, Ho C, Chu P, Trudell D, Resnick D. Knee hyaline cartilage evaluated with MR Conclusion imaging—a cadaveric study involving multiple imaging sequences and Intraarticular Injection of We compared sodium density to both proton density and gadolinium and saline solution. Radiology proton relaxation time maps of PG depleted and non- 1991;178:557–61. depleted cartilage. Sodium MRI is both sensitive and 10. Kusaka Y, Grunder W, Rumpel H, Dannhauer KH, specific in detecting small changes in PG concentration. Gersonde K. MR microimaging of articular cartilage Proton density and relaxation properties are neither and contrast enhancement by manganese ions. specific nor sensitive to these changes. The sodium MR Magn Reson Med 1992;24:137–48. signal can measure as small as 5–10% changes in PG 11. Grunder W, Biesold M, Wagner M, Werner A. Improved concentration. Future work will focus on application of nuclear magnetic resonance microscopic visual- sodium MRI to monitor PG loss in vivo. Sodium MRI has ization of joint cartilage using liposome entrapped the potential to be a noninvasive diagnostic tool to measure contrast agents. Invest Radiol 1998;33:193–202. early cartilage degeneration that is associated with OA 12. Bashir A, Gray ML, Boutin RD, Burstein D. Glyco- in vivo. Despite its lack of specificity to PG content in saminoglycan in articular cartilage: in vivo assess- cartilage, proton MRI can provide high-resolution images of ment with delayed Gd(DTPA)2−-enhanced MR imag- physical defects and advanced degeneration in cartilage ing. Radiology 1997;205:551–8. associated with OA. Excellent anatomical details of proton 13. Bashir A, Gray ML, Hartke J, Burstein D. Nondestruc- MRI and high sensitivity and specificity of sodium MR to tive imaging of human cartilage glycosaminoglycan PG changes can be exploited to develop a powerful concentration by MRI. Magn Reson Med noninvasive diagnostic tool for early OA. 1999;41:857–65. Osteoarthritis and Cartilage Vol. 8 No. 4 293

14. Lesperance LM, Gray ML, Burstein D. Determination lar cartilage in vivo. Magn Reson Med 1998;39:697– of ﬁxed charge-density in cartilage using nuclear- 701. magnetic-resonance. J Orthop Res 1992;10:1–13. 18. Insko EK, Kaufman JH, Leigh JS, Reddy R. Sodium 15. Maroudas A, Muir H, Wingham J. The correlation of NMR evaluation of articular cartilage degradation. ﬁxed negative charge with glycosaminoglycan con- Magn Reson Med 1999;41:30–4. tent of human articular cartilage. Biochim Biophys 19. Farndale RW, Sayers CA, Barrett AJ. A direct spectro- Acta 1969;177:492–500. photometric microassay for sulfated glycosamino- 16. Jelicks LA, Paul PK, O’Byrne E, Gupta RK. H-1, Na-23, glycans in cartilage cultures. Conn Tiss Res and C-13 MR spectroscopy of cartilage degradation 1982;9:247–8. in vitro. J Magn Reson Imag 1993;3:565–8. 20. Henkelman RM, Stanisz GJ, Kim JK, Bronskill MJ. 17. Reddy R, Insko EK, Noyszewski EA, Dandora R, Anisotropy of NMR properties of tissues. Magn Kneeland JB, Leigh JS. Sodium MRI of human articu- Reson Med 1994;32:592–601.