|Home | About | Journals | Submit | Contact Us | Français|
Changes in T1ρ and T2 magnetic resonance relaxation times have been associated with articular cartilage degeneration, but similar relationships for meniscal tissue have not been extensively investigated. This work examined relationships between T1ρ and T2 measurements and biochemical and mechanical properties across regions of degenerate human menisci.
Average T1ρ and T2 relaxation times were determined for nine regions each of seven medial and thirteen lateral menisci from fourteen total knee replacement patients. Sulfated glycosaminoglycan (sGAG), collagen and water contents were measured for each region. Biomechanical measurements of equilibrium compressive, dynamic compressive and dynamic shear moduli were made for anterior, central and posterior regions.
T1ρ and T2 times showed similar regional patterns, with longer relaxation times in the middle region compared to the inner and outer region. Pooled over all regions, T1ρ and T2 times showed strong correlations both with one another and with water content. Correlations with biochemical content varied depending on normalization to wet or dry mass, and both imaging parameters showed stronger correlations with collagen compared to sGAG content. Mechanical properties displayed moderate inverse correlations with increasing T1ρ and T2 times and water content.
Both T1ρ and T2 relaxation times correlated strongly with water content and moderately with mechanical properties in osteoarthritic menisci, but not as strongly with sGAG or collagen contents alone. While the ability of MRI to detect early osteoarthritic changes remains the subject of investigation, these results suggest that T1ρ and T2 relaxation times have limited ability to detect compositional variations in degenerate menisci.
The knee menisci distribute and transmit load through the joint by increasing the contact area of the tibiofemoral articulation, performing essential biomechanical functions1, 2. The semilunar, fibrocartilaginous meniscus is mainly composed of water, multiple collagen types and proteoglycans (PGs)1, 3. As in articular cartilage, negatively charged sulfated glycosaminoglycan (sGAG) side chains attached to the PG core proteins osmotically attract water, contributing to the tissue's resistance to compression3. In early-stage osteoarthritis (OA), the meniscal extracellular matrix (ECM) exhibits increased sGAG and collagen contents, accompanied by increased swelling4. However, with further degeneration, the water content increases while PG and collagen contents decrease, resulting in inferior mechanical properties5, 6. Such degenerative changes predispose the meniscus to injury and often precede cartilage degeneration during OA progression7, 8. Thus, methods to detect changes that occur within the meniscus in a quantitative manner could provide powerful diagnostic and research measurements for early-stage OA. Recently, quantitative magnetic resonance imaging (MRI) parameters such as T1ρ and T2 relaxation times have been proposed to perform such functions and used to study compositional and structural changes in soft tissues.
T1ρ describes the relaxation time due to spin-lattice interactions in the rotating frame and has been shown to be sensitive to hydrogen exchange between slow macromolecules and bulk water9, 10. In cartilage, many have proposed that T1ρ can detect PG loss in cartilage by probing the interactions between the amine and hydroxyl groups on the GAG chains and water molecules10-12. Less energy exchange occurring between these macromolecules and water will result in longer T1ρ relaxation times. T2, on the other hand, measures the decay caused by spin-spin interactions and has been associated with hydration and collagen content and organization13-15. Collagen matrix degradation increases water mobility, decreasing the spin-spin interactions and increasing the T2 relaxation time. Mechanical properties are also affected by compositional and structural changes in the ECM, and in cartilage, T1ρ and T2 have shown negative correlations with various mechanical measures16, 17.
As a result of their association with ECM changes, T1ρ and T2 changes in cartilage and meniscus have been associated with OA. In cartilage, both imaging parameters were found to increase with progressive stages of OA18, 19, and significant correlations have been found between T2 and clinical symptoms20. In the meniscus, average T1ρ and T2 values were significantly elevated in mild and severe OA groups relative to healthy subjects21, T1ρ was observed to be greater in the lateral meniscus of ACL-injured patients compared to control patients22 and T2 was found to correlate with WOMAC scores of OA severity23. T2* imaging, which reflects relaxation due to magnetic field inhomogeneities in addition to the spin-spin interactions, has been explored as an approach to detect subclinical meniscal damage that manifests as changes in collagen structure24. While these quantitative imaging parameters have potential to improve clinical OA diagnosis, the sensitivity to regional variations remains unclear and the underlying mechanisms that cause changes in meniscal relaxation times are not completely understood.
Numerous studies have examined relationships between cartilage composition and T1ρ and T2 relaxation times, but similar in-depth studies on the meniscus have not yet been reported. Compared to cartilage, the meniscus is difficult to image due to its short T2 relaxation time, on the order of 8-18ms23, 25, which causes rapid transverse signal decay. The complex, heterogeneous regional patterns of tissue composition and mechanical function pose another challenge in the meniscus, requiring higher resolution to detect these differences. The radially inner, avascular region experiences high compressive stresses under functional loading and has a relatively high PG content26. The radially outer, vascularized region, experiences high circumferential tension and has a relatively low PG content with the collagen fiber bundles oriented primarily in the circumferential direction1. In addition, differences in the shape of the meniscus and position within the knee between medial and lateral menisci and anterior and posterior horns contribute to regional differences in loading27. The heterogeneous nature of the meniscus adds another layer of complexity in understanding OA progression, as the disease can affect one region differently than another.
In order to evaluate the potential of T1ρ and T2 measurements in the meniscus as potential diagnostic tools for OA, it is critical to understand what these relaxation times can detect about the degenerative state of the meniscus. The purpose of this study was to determine regional variations of T1ρ and T2 relaxation times in the radial and circumferential directions in OA human menisci, and to investigate their relations to biochemical content (sGAG, collagen and water) and mechanical properties (equilibrium compressive, dynamic compressive and dynamic shear moduli). We hypothesized that regional variations in T1ρ and T2 times would reflect regional variations in tissue composition and functional properties, with increased T1ρ and T2 in regions with greater water content and lower sGAG content, collagen content and mechanical properties.
Twenty isolated menisci (13 lateral, 7 medial) were obtained as incidental surgical waste from 14 patients (6 males, 8 females, age 64.6 (57.4, 71.7) years) undergoing total knee replacement (TKR) procedures at Stanford University Hospital. The specimens were stored until imaging at 4°C in 1X phosphate buffered saline (PBS, Mediatech, Manasass, VA) with protease inhibitors (PI, Protease Inhibitor Cocktail Set I; EMD Biosciences, San Diego, CA).
Within 24 hours of surgery, each specimen was secured in a container filled with perfluorooctyl bromide (Exfluor Research Corporation, Round Rock, TX) to minimize image artifacts from the air-tissue interface. T1ρ- and T2-weighted images were acquired using a Magnetization-Prepared Angle-Modulated Partitioned k-Space Spoiled Gradient Echo Snapshots (3D MAPSS) sequence that has a magnetization preparation followed by an immediate RF-spoiled gradient recalled echo (SPGR) acquisition during transient signal evolution28. To minimize magic angle effects, the specimens were oriented as if a patient were being imaged in a feet first, supine position, with the main circumferential collagen fibers oriented perpendicular to the main magnetic field. Coronal images were acquired with a 3T GE scanner (MR750, GE Healthcare, Waukesha, WI) in an 8-channel wrist coil. Imaging parameters were TR = 7.5 ms, FOV = 10cm, matrix = 256x256, BW = ±31.25kHz, number of excitations (NEX) = 1, and 1mm slice thickness. To tailor the sequence for the short T2 meniscus, the pulse sequence was designed to achieve TE of 3.6ms for SPGR readout. T1ρ-weighted images were collected at 6 spin-lock durations (TSL) (0, 4, 12, 20, 40, and 72ms) with spin-lock frequency 500Hz, and T2-weighted images were collected at 6 T2-preparation times (0, 6.4, 12.8, 19.3, 38.5, and 64.2ms) (Fig. 1a). Each scan took about 30 minutes.
The menisci were additionally scanned using a 3D gradient echo (GRE) sequence to obtain an image in the transverse plane that displayed the entire meniscus, showing its curvature (Fig. 1c). The imaging parameters for this sequence were TR = 7.2ms, TE = 3.4ms, FA = 30°, FOV = 14cm, matrix = 256X256, BW = ±31.25kHz, NEX = 3 and 1.5mm slice thickness. This image was used to define 9 different regions in the meniscus (Fig. 1d). First, the center of a circle was defined by the curvature of the meniscus using the Taubin circle fit method29 (MATLAB, MathWorks, Natick, MA). The angle that encompassed the meniscus from this center was split into circumferential thirds, defining the boundaries of the anterior (A), central (C) and posterior (P) regions. To divide the meniscus in the radial direction, ten equally spaced radial lines were drawn from the center of the Taubin circle to the outer edge of the meniscus. For each line, the intersection of the radial line with the inner and outer edges of the meniscus were identified, and the distance between these two points was divided into thirds marking the inner/middle and middle/outer boundaries. A polynomial fit was performed through corresponding inner and outer points to define continuous boundaries of the inner (I), middle (M) and outer (O) regions (Fig. 1d).
From the 3D MAPSS sequence, each coronal slice was manually segmented and pixel intensities vs. time for different TSLs and TEs were fit with mono-exponential decay equations (MATLAB) to obtain the T1ρ and T2 time constants for each pixel from each region. T1ρ and T2 maps were generated for each slice (Fig. 1b), excluding boundary pixels to avoid partial volume effects. The regions identified in the 3D GRE images were mapped onto the cross-sectional T1ρ- and T2-weighted images and average relaxation times were calculated for each of these regions.
After imaging, the menisci were stored at -20°C in PBS with proteinase inhibitors (EMD Millipore, Billerica, MA) and later thawed for biochemical and biomechanical analysis. 4mm-diameter cores were taken from the radially middle portions of the anterior, central and posterior regions of each meniscus and were trimmed to 2mm thickness, discarding the surface layers. Photographs of the meniscus were taken after the samples were cored in order to register the locations of the cores with the 3D GRE images. Briefly, the photograph and 3D GRE image were converted to binary images and the photograph image was iteratively transformed to minimize the error between the two images (MATLAB).
The samples were first tested at 37°C in oscillatory torsional shear using an AR-2000ex torsional rheometer (TA Instruments, New Castle, DE). Samples were compressed at 0.1%/s to 10%, allowed to stress relax for 20 minutes, and subjected to ±0.5% nominal shear strain at 0.1Hz to determine the dynamic shear modulus (G*). After unloading and recovery in PBS for 30 minutes, the samples were tested at room temperature in unconfined compression using an Instron 5848 MicroTester (Instron, Norwood, MA). Sequential compressive offsets of 5, 10, 15, and 20% were applied at 0.1%/s followed by stress relaxation for 20 minutes in order to obtain the equilibrium compressive modulus (Eeq), which was determined via linear regression of the equilibrium stress vs. applied strain as described previously30. At the 10% offset, ±1.5% compressive strain was applied at 0.1Hz to determine the dynamic compressive modulus (E*) based on Fast Fourier Transform analysis of the force data (MATLAB).
Radial cross-sections with 1-2mm thickness adjacent to the mechanical test samples were taken from the anterior, central, and posterior regions of the menisci and further divided into inner, middle and outer portions in roughly equal radial distances to obtain samples corresponding to all 9 regions analyzed via MRI. Water content was calculated using the wet and dry masses of the samples measured before and after lyophilization. The wet masses ranged from 10 to 50mg, with a mean value of 32.5mg (30.6, 34.4). The samples were then digested in 3mg/ml Proteinase-K solution (Invitrogen, Carlsbad, CA) at 60°C overnight, with 1mg of Proteinase-K for every 20mg of tissue. sGAG contents were measured with the 1,9-dimethylmethylene blue (DMMB) spectrophotometric assay at 525nm31, using chondroitin suflate C (Sigma Aldrich, St. Louis, MO) standards. Collagen content was estimated from hydroxyproline content using the p-dimethylaminobezaldehyde assay with trans-4-hydroxy-L-proline (Sigma Aldrich) standards32 and a collagen to hydroxyproline mass ratio of 8.
Additional intact radial cross-sections from the anterior, central and posterior regions were fixed in 10% neutral buffered formalin at room temperature for 12 hours, dehydrated through graded alcohols, cleared with a xylene substitute (Histo-Clear, Electron Microscopy Sciences, Hatfield, PA), and embedded in paraffin. Sections were cut at 5 m using a sliding microtome and mounted on slides. The slides were stained for sGAG with Safranin-O (Sigma Aldrich) and Fast Green counterstain (EMS, Hatfield, PA).
Regional differences were evaluated using general linear models (GLMs) with P<0.05 and Tukey's test for pairwise comparisons, treating the donor as a random variable (Minitab, Minitab Inc., State College, PA). Data were checked for normality using an Anderson-Darling test. T1ρ, T2, sGAG content, and mechanical properties data showed non-normal distribution and were log-transformed prior to GLM analysis. Imaging and biochemical data were analyzed with side (medial vs. lateral meniscus), radial location (inner/middle/outer) and circumferential location (anterior/central/posterior) as factors, while mechanical properties were analyzed with side and circumferential location as factors. For each outcome, an initial multifactor model with all interactions was sequentially reduced via elimination of non-significant terms (p>0.05) while maintaining a hierarchical model.
Association across all pooled samples among imaging parameters, biochemical contents and mechanical properties were examined via Spearman's correlation (SPSS v19, IBM Corp, Armonk, NY). Correlation analyses of mechanical properties with imaging parameters used the average T1ρ and T2 relaxation times in the registered core regions from which the mechanical test samples were taken (Fig. 1d). The mechanical properties were additionally divided into upper and lower halves, and their correlations to T1ρ and T2 relaxation times were separately analyzed for the two subgroups.
T1ρ and T2 relaxation times varied depending on region in both the radial and circumferential direction. The central region showed shorter T2 relaxation times than either the anterior (P<0.0001) or the posterior region (P=0.0021) (Fig. 2c). T1ρ displayed similar patterns, with the lateral menisci showing lower relaxation times in its central vs. anterior region (P=0.0009) and the medial menisci in its central vs. posterior region (P=0.0007) (Fig. 2a). In the radial direction, both relaxation times were lower in the middle region than in the inner region (T1ρ: P=0.0043, T2: P=0.0001) (Fig. 2b, d). Small mediolateral differences were observed in T2, which showed higher relaxation times in the medial meniscus (P=0.0075).
On the other hand, water content, sGAG per dry mass (dm) and collagen per dry mass did not show much variation among anterior, central, and posterior regions (sGAG: P=0.52, collagen: P=0.18, water: P=0.56) but varied in the radial direction. The outer and middle regions showed higher sGAG/dm than the inner region (O>I: P=0.0056, M>I: P<0.0001), and the outer region had higher collagen/dm than the other regions (I<O: P=0.0017, M<O: P=0.0152) (Fig. 3b,d). Water content was lowest in the middle region (I>M: P=0.031, O>M: P=0.013) (Fig. 3a). Mediolateral differences in composition were also detected, with medial menisci showing higher sGAG/dm (P<0.0001) and lower collagen/dm (P=0.0041) and water content (P=0.047).
sGAG and collagen content normalized by wet mass (wm) showed slightly different patterns. Collagen/wm showed little regional variation (medial vs. lateral: P=0.66, circumferential: P=0.24, radial: P=0.073). sGAG/wm was higher in the medial meniscus than in the lateral meniscus for the anterior and central regions (A: P<0.0001, C: P<0.0001) but not the posterior region (P=0.11). However, it did not exhibit significant circumferential variation in either medial (P≥0.12) or lateral (P≥0.63) mensci. sGAG/wm was highest in the radially middle region, followed by the outer and inner regions (M>O: P=0.001, O>I: P=0.023, M>I P<0.0001) (Fig. 3c). Safranin-O staining for sGAGs was consistent with the DMMB assay, with the greatest intensity in the middle region and the least in the inner regions (Fig. 4).
The average values for Eeq, E* and G* were 52.2 (39.9, 72.6) kPa, 212.3 (157.8, 282.3) kPa, and 15.5 (12.0, 20.1) kPa, shown as mean (95% upper limit, 95% lower limit) respectively. Overall, mechanical properties exhibited little regional variation. No differences were detected for Eeq (medial vs. lateral: P=0.51, circumferential: P=0.24). The GLM indicated a significant interaction between side and circumferential location for E* (P=0.042) and G* (P=0.031), but none of the pairwise comparisons revealed differences (E*: P≥0.25, G*: P≥0.20).
In addition to exhibiting similar regional patterns, T1ρ and T2 relaxation times showed comparable correlations with biochemical contents and mechanical properties, and a high correlation was observed between the two imaging parameters (rho=0.92, P<0.001, Fig. 5). Both exhibited moderately strong correlations with water content and with sGAG and collagen content normalized to wet mass (Fig. 6, ,7).7). However, collagen/dm showed a weak correlation with T2 relaxation time (Fig. 7d) and no correlation with T1ρ (Fig. 6d), and sGAG/dm did not correlate with either T1ρ or T2 time. As expected, water content showed strong inverse correlation with sGAG/wm and collagen/wm but did not correlate with sGAG or collagen normalized to dry mass (Table 1). Overall, these results indicate that water content is a dominant determinant of T1ρ and T2 times in OA menisci.
All mechanical properties exhibited significant, negative correlations with both T1ρ and T2 relaxation times (Fig. 8). Interestingly, most of the variation in imaging parameters was seen among samples with lower moduli; when samples with low and high moduli were separately analyzed, correlation significance was either completely lost or substantially reduced. Similar to the imaging parameters, mechanical properties exhibited moderately strong correlations with water content, negative correlations with collagen/wm but none with collagen/dm (Table 2). No correlations of mechanical properties with sGAG content were observed other than a moderately positive correlation between sGAG/wm and G* (Table 2). As with the imaging parameters, these results suggest that water content is a strong determinant of compressive and shear moduli in OA menisci.
In this study, we explored regional variations of T1ρ and T2 relaxation times in degenerate human menisci and their relationships to biochemical contents and mechanical properties. We found that T1ρ and T2 times overall exhibited very similar regional patterns and were correlated with biochemical content and mechanical properties. However, the results suggested that variations in water content, rather than the sGAG or collagen content, determined much of the variation in T1ρ and T2 times in these samples.
The T1ρ and T2 times observed in this study were constantly higher than previously reported in-vivo values for the meniscus25, 33. This difference could be due to the fact that we used severely degenerated menisci from TKR patients, as diseased tissue has been reported to have higher relaxation times21-23. Additionally, the measurements were taken ex-vivo, necessitating storage in PBS prior to imaging, and changes in tissue hydration during this period may also contribute to the prolonged signal. However, as the purpose of this study was to examine the regional patterns within the meniscus and the relationships between the imaging and physical parameters, the absolute values of the time constants are not crucial in the interpretation of our results. All specimens were subject to a constant experimental procedure, and the relationships among the variables examined are expected to remain valid in this context.
Few studies have investigated regional variations of T1ρ or T2 relaxation times in the meniscus. Tsai et al. found that T2 relaxation times gradually increased from the inner to the outer regions in healthy human menisci25. We found an opposite pattern in OA menisci, with lower time constants in the middle region compared to the inner region. This altered pattern is consistent with the altered matrix composition in highly degenerated menisci. Although the inner region of the normal meniscus has the highest PG content26, 34, the PG content of the middle region increases with age and the onset of OA35, 36. We found low sGAG contents in the inner regions of our meniscal samples, suggesting that substantial PG degradation had occurred by late stage OA. As shown by the DMMB assay and Safranin-O stained sections, PG content was highest in the middle region, which also showed the shortest T1ρ and T2 relaxation times.
In cartilage, many studies have reported specific associations of T1ρ with sGAG content and of T2 with collagen content and organization12, 13. However, most of these studies did not separately report the biochemical contents per wet and dry mass. In our study, we saw substantial differences between the two normalizations. sGAG/wm showed relatively stronger correlation with T1ρ than with T2, similar to the findings in cartilage that T1ρ was more sensitive to proteoglycan content than was T2. However, we found no correlations between the imaging parameters and sGAG/dm. Additionally, we found stronger correlations with collagen/wm compared to sGAG/wm for both T1ρ and T2 (Fig. 6, ,7).7). As with sGAG, correlations between imaging parameters and collagen content were weaker or absent when normalized to dry mass. Combined with the direct correlations of both imaging parameters with water content, these results strongly suggest that water content may be the dominant factor in determining T1ρ and T2 relaxation times in highly degenerate menisci.
While T1ρ and T2 describe different magnetic resonance relaxation mechanisms, they both probe for the motion of water molecules, which are greatly influenced by changes in sGAG and collagen content. The similar regional patterns and high correlation between the two imaging parameters are congruent with previous reports in cartilage. Mlynarik et al. found almost identical T1ρ and T2 map distributions in cartilage specimens obtained from total joint replacement surgeries37, and Majumdar et al. observed a significant, direct correlation between the two parameters in-vivo38. These results indicate that, by this advanced stage of OA, any specific sensitivity T1ρ or T2 might have to a biochemical constituent in the meniscus is overwhelmed by their sensitivity to changes in water content. It should be noted that the meniscus has a much lower proteoglycan concentration than cartilage39, and the changes in proteoglycan content that may be possible to detect in cartilage by the T1ρ relaxation mechanism may not be detectable in the meniscus.
The influence of water on T1ρ and T2 relaxation times was also reflected in the imaging parameters’ moderate correlation with Eeq, E* and G*, which were in turn negatively correlated with water (Fig. 8, Table 2). While the correlations between the imaging parameters and mechanical properties were moderately strong, this relationship did not hold in narrow ranges when samples with only high or low moduli were considered, indicating that T1ρ and T2 may not have high sensitivity to mechanical function. None of the measured mechanical properties showed much regional dependence, consistent with previous studies in older patients40. However, some other studies have shown regional variation in mechanical properties in relatively younger and healthier patients6, 41. It will be interesting to investigate whether T1ρ and T2 reflect those regional patterns in that population, as it would indicate that these imaging parameters could differentiate between regions of weak and strong mechanical properties and perhaps detect changes in regional patterns earlier in OA development. It should be noted here that the measured mechanical properties may not be an accurate average of the entire region as the core sample preparation procedure itself disrupts the interaction between the collagen fibers, which contributes to the biomechanics of the tissue. While this is inevitable in mechanical testing of this nature, future studies may explore different methods such as indentation testing of an intact meniscus in order to avoid such limitations.
In this study, we found that the regional variation in T1ρ and T2 relaxation times reflect changes in biochemical composition, but are mostly associated with differences in water content. The association of imaging parameters with the composition and mechanical properties of the tissue indicate that T1ρ and T2 have the potential to be no n-invasive markers of meniscal degeneration. However, we only examined menisci from late-stage OA patients in this study, and future work should include healthy, age-controlled samples in order to establish a baseline to properly interpret changes in T1ρ and T2 relaxation times. We need to investigate whether the correlations found in this study hold in earlier stages of OA, and whether these relationships are sensitive enough to detect subtle changes. Thus, understanding how the compositional and mechanical heterogeneity of the meniscus changes from a healthy, normal state to different stages of OA and investigating how these changes are reflected in the imaging parameters will be essential in developing protocols for detecting early degenerative changes in the meniscus. This study is a starting point in evaluating T1ρ and T2 relaxation times as a potential marker of meniscal degeneration and the first to investigate the relationship between these imaging parameters and the biochemical and mechanical properties of the meniscus.
The authors thank Dr. Emily McWalter for her support in conducting experiments and interpreting data.
Role of the Funding Source
This research was supported by NIH R01AR052861, NIH R01EB002524, NIH K24AR062068, a Bio-X Graduate Fellowship, the Siebel Scholars Foundation, GE Healthcare, the Arthritis Foundation and SCBT-MR. None of the sponsors had any involvement in the design, execution or interpretation of the study or in the writing or submission of the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Min-Sun Son contributed to the concept and design of the study, acquisition of data, analysis and interpretation of data and drafted the article. Dr. Chen provided the MAPSS sequence and technical support to run the study. Dr. Goodman provided the meniscal specimens used in the study and contributed to study design. Drs. Levenston, Hargreaves and Gold contributed to the concept and design of the study and analysis and interpretation of data. Drs. Levenston and Gold additionally provided funding for the study. All authors critically revised the article and approved the final version to be submitted. Min-Sun Son and Marc Levenston take responsibility for the integrity of this work as a whole and can be contacted at minsunson/at/gmail.com and levenston/at/stanford.edu.
Weitian Chen is employed by GE Healthcare.