To determine the relationship between changes in the extracellular matrix (ECM) and T1ρ and T2 values in vivo. The ECM is composed of proteoglycan (PG), collagen, and water. It has been unclear which of the ECM constituents affects T1ρ and T2 mapping in living human cartilage.
Materials and Methods:
Sagittal T1ρ and T2 maps were preoperatively obtained from 20 knee osteoarthritis patients. Osteochondral samples harvested from the resected tibial plateaus during total knee arthroplasty were consistent with the MRIs of the patients' knees. Parameters that included histological grading of cartilage degeneration, glycosaminoglycan (GAG) content (which constitutes PG), presence of collagen anisotropy and water content were evaluated along with T1ρ and T2 values, and statistical analysis was performed using multiple regression analysis.
T1ρ and T2 values were significantly correlated with the degree of cartilage degeneration (β = 0.397 and 0.357, respectively) and the GAG content (β = −0.340 and −0.244, respectively).
OSTEOARTHRITIS (OA) IS a degenerative disease that affects the hyaline cartilage at the articular surface. Cartilage matrix breakdown is characterized by changes in the content of glycosaminoglycan (GAG), type II collagen, and water (1). It is difficult to detect these changes in the cartilage matrix using plain radiography or conventional magnetic resonance imaging (MRI) techniques (2, 3). However, new MRI techniques that can detect changes in the extracellular matrix (ECM) have recently been developed. In particular, T1ρ and T2 mapping can quantify the ECM of the articular cartilage.
The T1ρ parameter describes the spin-lattice relaxation in the rotating frame (4). A “spin-lock” (SL) pulse is applied to the magnetization in the transverse plane at extremely low fields, and this spin-locked magnetization reflects the changes in the magnetic fields caused by water spin dynamics, such as the chemical exchange of protons. T1ρ mapping is most sensitive to changes in the GAG content in the ECM (5–7). A previous in vitro study using a model with selective degradation of GAG by trypsinization revealed that increases in T1ρ values correlated with decreases in the GAG content (6, 7). The chemical exchanges between bulk water and the hydroxyl and amine groups of proteoglycan (PG) are considered important for the T1ρ parameter's relaxation mechanism in articular cartilage (8). In addition, the chemical exchanges between the hydroxyl and amino groups of collagen and the water protons may affect the T1ρ parameter; a previous study demonstrated that T1ρ shows approximately exponential dependencies on molecular concentration using collagen solutions (9).
The T2 parameter is a spin–spin relaxation method related to the energy changes in the water protons in cartilage; consequently, the T2 relaxation time reflects the cartilage's water content. Additionally, the T2 values reflect the presence of collagen anisotropy due to dipolar interaction (10–12) because the two protons in water molecules are bound to and arranged in parallel with collagen fibers. Furthermore, an in vitro study using a model of PG depletion induced by hyaluronidase demonstrated a correlation between T2 values and the PG content (13).
Many in vitro studies have indicated that T1ρ values reflect GAG content (6, 7) and that T2 values reflect the presence of collagen anisotropy and water content (10, 12, 14). However, it has not been established whether T1ρ and T2 values reflect these components in the ECM of living human cartilage. In clinical studies, the T1ρ and T2 values in the articular cartilage of knee OA patients have been reported to correlate with the severity of their radiographic evaluation (15–17); however, it remains unclear whether in vivo T1ρ and T2 values reflect the changes that specifically result from cartilage degeneration.
The purpose of the present study was to determine whether the T1ρ and T2 values in living human cartilage reflect the ECM composition and whether they can reveal the relevant changes that result from cartilage degeneration. We hypothesized that T1ρ values in vivo would reflect the GAG content and that T2 values would reflect the presence of collagen anisotropy and water content. We performed preoperative T1ρ and T2 mapping for patients scheduled to undergo total knee arthroplasty (TKA) and calculated the T1ρ and T2 values from reconstructed maps. Subsequently, the degree of cartilage degeneration, the GAG content, the presence of collagen anisotropy and the water content of the osteochondral samples harvested during surgery were evaluated, and the correlations between these parameters and the T1ρ and T2 values were examined.
MATERIALS AND METHODS
This study received approval from our institution's Institutional Review Board and conformed to the Declaration of Helsinki. Twenty patients with knee OA (2 males and 18 females) scheduled for primary TKA participated in this study. The mean age at surgery was 77.0 years (range, 63 to 92 years). The mean interval between MR imaging and surgery was 18.5 days (range, 2 to 90 days). Using the Kellgren-Lawrence radiographic grading system, 19 OA patients were classified as grade 4 and 1 as grade 3. The clinical results were evaluated using the knee scoring system of the Japanese Orthopaedic Association (JOA). The subjects' mean JOA score was 50.8 out of the maximum 100 points (pain on walking, 13.3/30 points; pain on ascending or descending stairs, 6.5/25 points; range of motion, 24.3/35 points; and joint effusion, 6.8/10 points).
MRI was performed preoperatively, and all MR examinations were performed using a 3-Tesla MR Scanner (Achieva 3 Tesla, Philips, Best, the Netherlands) and an eight-channel SENSE knee coil (Philips, Best, the Netherlands). When installing a knee coil for MR imaging, the affected limb was fixed at 15 degrees flexion and in a neutral position of rotation. The MR images were set to record sagittal slices orthogonal to the line connecting the edges of the medial and lateral posterior tibial condyles by the coronal images, and the images were acquired with slices at 4-mm intervals from the outer edge of the lateral condyle. The protocol included three sequences, and the acquisition parameters were as follows: sagittal T1-weighted 3D water selective excitation technique (WATS) imaging (repetition time/echo time [TR/TE] = 4.7 ms/10 ms, flip angle = 20 degrees, field of view = 14 cm, matrix = 400 × 400, section thickness = 1.5 mm); T2-weighted turbo spin-echo imaging by multiecho spin-echo sequence (Carr Purcell Meiboom Gill sequence) (TR/TE = 16 × n (n = 1–7) ms/4,613 ms, flip angle = 90 degrees, field of view = 14 cm, matrix = 320 × 320, section thickness = 4 mm); and 3D T1ρ-weighted fast field echo imaging (spin-lock pulse amplitude = 440 Hz [ω1 = γB1/2π], time of the spin-lock pulse (TSL) = 1, 10, 20, 30, 40 ms, TR/TE = 2.3 ms/4.7 ms, flip angle = 35 degrees, field of view = 14 × 14 cm, matrix = 256 × 256, section thickness = 4 mm, bandwidth = 0.7 kHz/pixel, NEX = 1). T1ρ and T2 mappings were reconstructed by fitting the image intensity pixel-by-pixel to the equation below using the mono-exponential fitting algorithm:
S (TSL) = S0 × exp (-TSL/T1ρ), where TSL is the time of spin-lock, S is the signal intensity of the T1ρ-weighted image with a given TSL, and S (TE) = S0 × exp (-TE/T2), where S is the signal intensity of the T2-weighted image with a given TE.
T1ρ and T2 mappings were constructed using the PRIDE software (Philips, Best, the Netherlands) written in an Interactive Data Language (RSI, Inc, Boulder, CO). The image analyses were performed with Image J software (NIH, Bethesda, MD). To determine the precise setting for the regions of interest (ROIs) on the reconstructed map, we evaluated the interobserver reproducibility in a preliminary test. The coefficient of variation (CV, %) for the interobserver reproducibility test, which was calculated for the T1ρ and T2 values of each ROI, showed good reproducibility (CV < 6.0%) for all measurements. The ROIs of full-thickness cartilage were set at 12 mm (third slice) and 16 mm (fourth slice) of the lateral tibial plateau and at 56 mm (fourteenth slice) of the medial tibial plateau from the lateral edge of the lateral condyle. The anterior–posterior (AP) length and its midpoint in the cartilage of the lateral tibial plateau were determined for each slice. The center of each 10-mm-long ROI was positioned in each slice to match the midpoint on the lateral tibial plateau. Two other ROIs, each 10 mm in length, were placed anterior and posterior to the central ROI on each slice of the lateral tibial plateau (Fig. 1a). Similar to the lateral tibial plateau, an ROI was established on the remaining cartilage area of the posterior medial tibial plateau (Fig. 1b). The relaxation times for T1ρ and T2 mapping at each ROI were measured 3 times by a single examiner, and the mean values were calculated. We demonstrated high reproducibility of measurements in the T1ρ and T2 mapping (intraclass coefficients of 0.98 and 0.96, respectively).
Tibial plateau specimens were resected during surgery and stored on ice. To exclude the influence of sample irregularities, the osteochondral samples were digitalized (two dimensionally) after the spur and soft tissue were removed. As with the sagittal MRI slices, the slices based on osteochondral sampling sites were set orthogonally to a line connecting the edge of the medial and lateral posterior tibial condyles (Fig. 2a). The slices were set 12 mm, 16 mm, and 56 mm from the lateral edge of the lateral tibial condyle. The sampling sites on the lateral tibial plateau were first determined by the AP length and its midpoint on the cartilage; next, the centers of 10-mm-long sampling sites were placed in each slice to match the midpoint, and two other 10-mm-long sites were placed anterior and posterior to the center site. The samples were harvested from three sites on each slice of the lateral tibial plateau and from one site on the medial tibial plateau, so that seven samples were harvested from each tibial plateau. Each sample was 4 mm in width × 10 mm in length and was divided into segments for evaluating the amount of sulfated GAG (sGAG), the histological grade and the water content (Fig. 2b). Cartilage samples for quantitative sGAG measurement were dissolved in a digestion buffer that included papain immediately after harvesting, and the solution was frozen at −80°C. Half of each sample was fixed in 4% paraformaldehyde and was decalcified, defatted, and paraffin-embedded for histological evaluation. The remaining cartilage quadrants were used for water content analysis.
Quantitative Measurement of GAG
Frozen cartilage solutions containing GAG were thawed and used in an Alcian blue binding assay (18) based on the specific interaction between negatively charged GAG and the tetravalent cationic dye, Alcian blue. The sGAG quantitative kit from Wieslab (Euro-Diagnostica AB, Medeon, Sweden) containing stock Alcian blue solution and 0.4 M guanidine HCl was used for the Alcian blue binding assay. The GAG content was quantitatively analyzed by measuring the sample absorption at 620 nm using a spectrophotometer.
Paraffin-embedded specimens were cut into 4-μm-thick slices, and sections were stained with hematoxylin-eosin and safranin O. The histological results were classified according to the Osteoarthritis Research Society International (OARSI) OA cartilage histopathology assessment system (OARSI grade) (19). After the removal of paraffin and GAG according to Kiraly's methods (20), the collagen fibril anisotropy was graded from 1 to 4 under a polarized light microscope (PLM), according to David-Vaudey's method (21). The histological evaluation was performed by another examiner, who was blinded to the MRI findings.
The wet weight of the cartilage was measured after the surface moisture was removed, and the dry weights of cartilage were recorded after they had been incubated at 60°C for 3 days. The water content was calculated as the wet-to-dry ratio (W/D ratio).
We analyzed 140 samples from 20 subjects. This sample size was considered sufficient based on pairwise comparisons of the sample correlation coefficients, using the following cutoff values for statistical significance; an α of 0.05, a β of 0.20 and a power (1-β) of 0.80.
The correlations between the T1ρ and T2 values and the GAG content or water content were determined using Pearson's correlation analysis. A Spearman's rank correlation test was performed to study the relationships between the histological evaluation (OARSI grade and PLM grade) and the T1ρ and T2 values, the GAG content and the water content. Analysis of variance (ANOVA) and post hoc comparisons using Scheffe's test were used to assess the differences in T1ρ and T2 values among the different OARSI and PLM grades, and the relationship between the OARSI grade and the PLM grade was analyzed using the chi-squared test. Moreover, multivariate analysis was performed using multiple regression analysis to determine the relationships between the T1ρ and T2 values and the OARSI grade and ECM composition. Additionally, we analyzed the relationship between the ECM and T1ρ and T2 using the structural equation modeling (SEM) program. Differences were considered statistically significant for P values < 0.05. All statistical tests were performed using the PASW Statistics 18 software package and SPSS Amos, Version 19.0 (SPSS Inc, Chicago, IL).
The mean GAG level was 252.1 μg/mL (range, 41.3 to 601.3 μg/mL), and the OARSI grade evaluation revealed a decrease from 339.7 μg/mL for OARSI grade 1 to 125.1 μg/mL for OARSI grade 4 samples. There was a negative correlation between GAG level and OARSI grade (ρ = −0.58; P < 0.001, Table 1). The PLM grade showed the breakdown of birefringence according to OA progression (Fig. 3). The mean PLM grade increased from 1.6 for OARSI grade 1 samples to 4.0 for OARSI grade 4 samples, and the PLM and OARSI grades were highly correlated (ρ = 0.77; P < 0.001, Table 1). The mean W/D ratio was 3.9 ± 1.2 and exhibited a significant increase, from 3.7 for the OARSI grade 1 samples to 4.5 for the OARSI grade 4 samples (ρ = 0.35; P < 0.001; Table 1).
Table 1. The relationships between OARSI grade and GAG content, PLM grade, W/D ratio and T1ρ or T2 values*
n = 17
n = 41
n = 35
n = 12
The values represent the mean ± S.D.
Spearman's rank correlations test.
Versus grade 1.
Versus grade 2.
Versus grade 3, determined by ANOVA with Scheffe's test.
As the histological degeneration progressed by OARSI grade, T1ρ mapping showed an elevated signal intensity (Fig. 3). The mean T1ρ value was 42.3 ms (range, 33.1 to 62.4 ms), and the values were positively correlated with the OARSI grade (ρ = 0.63; P < 0.001, Table 1). A comparison of all of the OARSI grades revealed significant differences between the T1ρ and OARSI grades. The univariable relationships between the T1ρ values and the ECM compositions showed a significant correlation between GAG content (r = −0.64; P < 0.001, Fig. 4a) and PLM grade (r = 0.53; P < 0.001, Fig. 4b); however, there was no significant correlation with the water content (r = 0.21; P = 0.071, Fig. 4c). In a multivariate analysis using multiple regression analysis, the T1ρ values were significantly correlated with the OARSI grade (β = 0.397, P = 0.006) and GAG content (β = −0.340; P < 0.001) (Table 2).
Table 2. Results of Univariate and Multivariate Analyses for T1ρ Values*
Standardized partial regression coefficient (β)
Univariate analysis: correlation coefficient and P values. Multivariate analysis: β (standardized partial regression coefficient) and P values, with 95% confidence intervals in parentheses.
0.397, (0.630, 3.578)
−0.340, (−0.022, −0.007)
0.087, (−0.911, 1.851)
0.142, (−0.201, 2.591)
The T2 values showed elevated signal intensity with the progression of histological degeneration as the OARSI grade increased (Fig. 3). The mean T2 values were 37.0 ms (range, 30.2 to 54.8 ms) and were positively correlated with the OARSI grade (ρ = 0.43, P < 0.001, Table 1). There were differences in the T2 values between grade 1 and all of the other grades, as well as between grades 2 and 4. Similar to the results for the T1ρ values, the univariate analysis revealed significant correlations between the T2 values and the GAG content (r = −0.48; P < 0.001, Fig. 5a) and the PLM grade (r = 0.40; P < 0.001, Fig. 5b), but no correlation with water content (r = 0.18; P = 0.123, Fig. 5c). A multiple regression analysis showed that the T2 values were significantly correlated with the OARSI grade (β = 0.357, P = 0.038) and GAG content (β = −0.244, P = 0.030) (Table 3). In a multivariate regression analysis using SEM, the standardized partial regression coefficient (β) for T1ρ was higher than that for T2. Furthermore, the coefficients of determination (R2) for T1ρ and T2 were 0.53 and 0.29, respectively (goodness of fit index = 1.0; comparative fit index = 1.0).
Table 3. The Results of Univariate and Multivariate Analyses of T2 Values*
Standardized partial regression coefficient (β)
Univariate analysis: correlation coefficient and P values. Multivariate analysis: β (standardized partial regression coefficient) and P values, with 95% confidence intervals in parentheses.
0.357, (0.113, 3.783)
−0.244, (−0.020, −0.001)
0.120, (−1.056, 2.384)
W / D ratio
0.001, (−1.726, 1.749)
The present study showed that the T1ρ and T2 values in living human cartilage were increased and correlated with the histological degeneration of cartilage and that these values showed significant correlations, even in a multivariate analysis. Univariate analysis showed that the T1ρ and T2 values were correlated with the GAG content and PLM grade. In the multivariate analysis, only the GAG content was shown to correlate with both T1ρ and T2 values, but not with collagen anisotropy or water content. To our knowledge, this study is the first to elucidate the relationship between in vivo T1ρ and T2 values and the degree of cartilage degeneration or ECM composition.
Our study confirmed that the T1ρ values that are measured in living tissue reflect the GAG content. The ECM is primarily composed of PG, which contains negatively charged GAG and type II collagen, facilitating the binding of water molecules. The breakdown of GAG results in an increase in the density of mobile protons in the bulk water and causes the T1ρ values to increase (6, 7, 22, 23). Therefore, the T1ρ values are sensitive to changes in GAG content. In vitro studies using chondroitin sulfate phantoms (22) and PG degenerative samples (trypsinized bovine cartilage) (6–8, 22) have also revealed that T1ρ values were correlated with the GAG content. Regarding collagen, Menezes et al. (9) have reported a relationship between T1ρ values and the collagen molecule concentration. However, the T1ρ values showed no increase in collagen degeneration in a veal patella model treated with collagenase (7). Because the T1ρ changes within the collagen concentration range in articular cartilage were small (24), the results of collagen anisotropy changes showed no correlation with T1ρ values in vivo.
The T2 values correlated with histological degeneration and GAG content in multivariate analysis, but not with PLM grade or water content. Because PLM is easily distinguished in the sequence of human articular cartilage layers divided into three distinct structural zones (25), we used PLM to evaluate collagen anisotropy in this study. The T2 values were related to energy changes resulting from the dipolar interaction, which stems from the interaction of water protons with the oriented collagen fibers (10, 25, 26). An increase in T2 values reflects a considerable loss of the birefringence signal during PLM evaluation of collagenase-treated cartilage (10). In contrast, the model of PG depletion induced by chondroitinase ABC showed no abnormality in collagen anisotropy and had no significant effect on the T2 values (10). According to these in vitro studies, T2 values reflect changes in collagen anisotropy compared with the PG content. In contrast, Watrin-Pinzano et al. (13) have reported that T2 values correlated with the PG content in an in vitro study using hyaluronidase-mediated degeneration of rat cartilage. Furthermore, Keenan et al. (27) revealed a relationship between T2 values and GAG content in the multivariate analysis of an in vitro study using patella cartilage from the knees of human cadavers. The relationship between T2 values and PG content has remained unclear. T2 values reflect the collagen anisotropy, which is affected by the interactions of water protons. The water protons need the presence of PG to arrange them parallel to the collagen fibers. Thus, changes in T2 values have been thought to reflect not only collagen anisotropy but also GAG content (12). Our in vivo results demonstrated a correlation between the T2 relaxation time and the PLM grade in a univariate analysis, similar to the in vitro findings of previous studies. However, the results of the multivariate analysis indicated that T2 relaxation time reflects GAG content, which is consistent with the previous in vitro studies by Watrin-Pinzano et al. (13) and Keenan et al. (27). The lack of a correlation between T2 values and collagen anisotropy in the multivariate analysis will need to be confirmed by additional studies using electron microscopy or other methods.
T1ρ and T2 values showed no significant correlation with water content, as evaluated using the W/D ratio based on previous reports (14, 28). Consistent with the present study, a previous study reported that T1ρ values showed no relationship with water content (29). Lusse et al. (14) reported a correlation between T2 values and water content in an in vitro study. These authors used a perpendicular orientation to the magnetic field of the collagen fibrils in the radial zone (90° orientation). In living human cartilage, however, an evaluation of all layers is necessary, and it is difficult to obtain an accurate assessment because of the mismatch between the magnetic fields and the direction of the collagen fibers of each cartilage layer. Our study showed no correlation between T2 values and water content, which was likely due to the differences in collagen fibers between the different layers, because the water content was evaluated for all layers.
Comparing the correlation to the OARSI grade of each MRI parameter (T1ρ and T2 values), the correlation coefficient of T1ρ (β = 0.397, ρ = 0.63) was higher than that of T2 (β = 0.357, ρ = 0.43). Our statistical analysis of the relationship between the T1ρ and T2 values and each factor of the ECM and OARSI grades using the SEM indicated that T1ρ is more sensitive for detecting cartilage degeneration than T2. Li et al. reported that the T1ρ and T2 in OA patients were higher than in healthy subjects. These authors observed a 19.1% increase in the T1ρ relaxation time, which was almost double that of the T2 relaxation time (9.6%). In addition, the T2 values in vivo are different from those of excised samples because T2 mapping is affected by imaging directions, such as the magic angle effect (30). In contrast, T1ρ values are not considerably affected by such artifacts and are therefore more effective than T2 values for providing an accurate evaluation of the living human knee (30).
There were several limitations to this study. First, there were inconsistencies between the ROI settings and the tissue sampling sites. We established this region to match the MR images and tissue sampling sites in sagittal slices based on a certain distance from the outer edge of the lateral condyle of the tibia; it was set orthogonal to the line connecting the edges of the medial and lateral posterior tibial condyles. The MR images of analytic sites were two-dimensional; however, the ECM evaluation was performed using three-dimensional samples because the GAG content and water content were measured in 2-mm-wide segments of cartilage. These discrepancies between MR imaging and the ECM evaluations may be disregarded because the histological degree of degeneration of each side was consistent with that of the opposite side. Second, we analyzed the entire cartilage for all of the evaluations reported in this study. Because T2 values are affected by the differences in the collagen anisotropy in different magnetic fields (11, 31, 32), the relationships between the T2 values and the PLM grade or water content in this study may not be exact. It is necessary for the T2 mapping values to be evaluated for relationships with the PLM grade or water content in each layer. The third limitation is our relatively small sample size and the single target site, the tibia, that was evaluated. We selected the cartilage of the tibial plateau because it was resectable and could be removed intact during surgery for use in further evaluations. Because the sagittal view of the tibia is best for evaluating the morphological changes that reflect OA (21), both T1ρ and T2 mapping were performed at this site in the present study. However, it is necessary to perform a more comprehensive assessment with a larger sample size and to study another condyle. The final limitation of this study is related to the collection of some samples from the same knees. The results of two-factor ANOVA analyses comparing intrasubject and intersubject variation indicated that different sites in the same subject showed significant differences (data not shown). However, there were no significant differences among the subjects at the same site for each sample. Our study design represented a reasonable approach for evaluating the MRI parameters and the ECM composition according to the different degrees of cartilage degeneration.
In conclusion, the present study demonstrated that the T1ρ and T2 values of living human knee correlated with histological degeneration. Furthermore, both T1ρ and T2 values correlated with the GAG content in living tissue, in contrast to our hypothesis that there would be correlations between T1ρ values and the GAG content and between T2 values and collagen anisotropy or water content. In clinical evaluations, previous studies have demonstrated that the T1ρ (16, 33) and T2 (15, 34) relaxation times increased with the progression of OA. Our results support these relationships between OA progression, as evaluated radiographically, and T1ρ and T2 values that reflect cartilage degeneration (evidenced by GAG depletion). Using both imaging methods could provide an evaluation of cartilage constituents, detect qualitative changes in the ECM, and quantitatively measure cartilage degeneration. These MR imaging methods are useful for understanding OA pathogenesis and determining the effects of treatment.