|Home | About | Journals | Submit | Contact Us | Français|
The purpose of this work is to demonstrate the feasibility of T1ρ-weighted magnetic resonance imaging (MRI) to quantitatively measure changes in proteoglycan content in cartilage. The T1ρ MRI technique was implemented in an in vivo porcine animal model with rapidly induced cytokine-mediated cartilage degeneration. Six pigs were given an intra-articular injection of recombinant porcine interleukin-1β (IL-1β) into the knee joint before imaging to induce changes in cartilage via matrix metalloproteinase (MMP) induction. The induction of MMPs by IL-1 was used since it has been extensively studied in many systems and is known to create conditions that mimic in part characteristics similar to those of osteoarthritis. The contralateral knee joint was given a saline injection to serve as an internal control. T1ρ-weighted MRI was performed on a 4 T whole-body clinical scanner employing a 2D fast spin-echo-based T1ρ imaging sequence. T1ρ relaxation parameter maps were computed from the T1ρ-weighted image series. The average T1ρ relaxation rate, R1ρ (1/T1ρ) of the IL-1β-treated patellae was measured to be on average 25% lower than that of saline- injected patellae indicating a loss of proteoglycan. There was an average reduction of 49% in fixed charge density, measured via sodium MRI, of the IL-1β-treated patellae relative to control corroborating the loss of proteoglycan. The effects of IL-1β, primarily loss of PG, were confirmed by histological and immunochemical findings. The results from this study demonstrate that R1ρ is able to track proteoglycan content in vivo.
The early stages of osteoarthritis (OA) are hallmarked by molecular and biochemical changes in the joint, including loss of proteoglycan (PG) from the extracellular matrix (ECM) of articular cartilage [17,25,36]. While the pathology of OA is a result of multiple cellular and molecular changes, the early changes in PG, in part, can greatly affect the integrity of the tissue and cause profound changes in its molecular and biochemical characteristics. Articular cartilage is composed primarily of collagen (type II), negatively charged PG molecules, chondrocytes, and water [15,26,34]. Biochemical changes in cartilage PG content are associated with the early stages of OA [17,25,36] and are known precede changes in collagen content [4,20] although small structural changes in collagen have been observed during early OA .
Magnetic resonance imaging (MRI) is a commonly used modality to noninvasively ascertain the condition of cartilage. Magnetic resonance imaging can readily be used to detect morphological changes in cartilage thickness and structure including defects such as gross fibrillation and fissures [14,24]. However, standard MRI methods have been shown to be inconsistent in detecting biochemical changes associated with the early stages of OA . Several PG-sensitive MRI techniques have been developed to image articular cartilage. Contrast- enhanced proton MRI, in which Gd-DTPA2− contrast medium is permitted to diffuse into the cartilage, has been used as a means to measure PG content . The contrast-enhanced approach has been previously reported to track degenerative changes in several animal models of osteoarthritis [21,22,40]. However, this technique requires careful calibration of the post-contrast injection delay and joint exercise to allow for proper diffusion of the contrast medium . In addition, T2 relaxation mapping using a spin–echo-based sequence has been shown to be sensitive to early degenerative changes in human subjects in vivo . However, T2 has been found to correlate poorly with PG content in controlled in vitro studies . Sodium MRI has been reported as a noninvasive technique to quantitatively measure PG content in cartilage [7,19,23]. The method has been previously validated as an accurate measure of cartilage PG content as corroborated by independent measurements via spectrophotometric assay .
Another promising technique of PG quantification is T1ρ-weighted MRI. The T1ρ parameter describes the spin–lattice relaxation in the rotating frame. It is sensitive to the slow-motion interactions between motionally restricted water molecules and their local macromolecular environment. The ECM in articular cartilage provides a motionally restricted environment to water molecules that can be detected by T1ρ-weighted MRI. Changes to the ECM, such as PG loss, are reflected in quantitative changes in the T1ρ relaxation rate, R1ρ (1/T1ρ). R1ρ has been shown to decrease linearly with decreasing PG content in controlled degradation experiments performed on ex vivo bovine patellae and osteoarthritic human specimens [2,13,31,43].
To investigate the use of the T1ρ technique as a means to study the pathology of OA, we carried out experiments using an animal model with a known mediator of cartilage catabolism: interleukin-1 β (IL-1 β). Introduction of IL-1 β into an in vivo system was hypothesized to result in the degradation of PG molecules by increasing expression and subsequent activation of matrix metalloproteinases (MMPs) in chondrocytes of the joint and surrounding ECM [11,28,32]. Since the effect of IL-1 β rapidly produces biochemical changes in chondrocytes in cartilage that closely mimic the molecular events and pathology associated with OA it was the mediator of choice [10,32,35]. The delivery of IL-1 β has been previously used in a variety of OA animal models [22,37–39].
The study included six three- to five-month-old Yorkshire pigs weighing 40–45 kg (Animal Biotech Industries, Danboro, PA). The study was approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania. The six pigs were administered an injection of 100 ng of recombinant porcine IL-1 β (R&D Systems, Minneapolis, MN) dissolved in 1 ml of saline in the joint space of the right knee and a control injection of 1 ml of saline in the left knee 6 h before imaging. Preliminary data were acquired on three separate pigs using delays of 1, 6, and 12 h between injection of IL-1 β and imaging to analyze the rapidity of the effect of IL-1 β. The injection was delivered by a board-certified veterinarian using a 20-gauge, 1-inch needle through the suprapatellar ligament into the center of the intra-articular joint space. The animal was sedated (20 mg/kg ketamine) for the period while it received the injections. Within 30 min of the injections, each animal was moving freely about its pen while being monitored by a veterinary technician. The animal was allowed to exercise the joint in order to aid the diffusion of the IL-1 β within the cartilage of the joint space. Immediately before imaging, each animal was anesthetized with ketamine (20 mg/kg) and acetylpromazine (0.15 mg/kg) administered intramuscularly. The animal was then intubated endotracheally and maintained under isoflurane (1–3%) anesthesia by veterinary technicians for the duration of the imaging session (approximately 50 min). Immediately after imaging, the animal was euthanized by a veterinary technician using thiopental (60 mg/kg), and the patellae were harvested surgically. In our analysis of the animal model, we chose to focus on the patella largely because of its ease of assessment. The patella is proximal to the surface of the skin and therefore is readily identifiable in surface coil generated MR images. Furthermore, the patella is easy to dissect from the knee joint thereby simplifying the ex vivo analyses.
T1ρ modulation can be incorporated into an MR imaging sequence by including a pre-encoded long duration low power “spin-lock” pulse. The resultant signal becomes exponentially weighted by the T1ρ parameter according to Eq. (1) where TSL is the spin-lock pulse duration and S is the signal intensity as a function of TSL. By acquiring a series of T1ρ-weighted images at various TSL durations and using Eq. (1), the T1ρ parameter can be quantified on a pixel-by-pixel basis using linear regression to create a spatial map of T1ρ values.
The spin-lock pulse is generally included as part of a spin-lock pulse cluster consisting of a 90°+x square pulse followed by a self-compensating spin-lock pulse (with spin-lock frequency of 500 Hz) and a 90°−x square pulse subsequently followed by a crusher gradient to suppress any residual transverse magnetization . The nonselective nature of the pulses used in the T1ρ-weighted pulse sequence restricts the image acquisition to a single slice at a time. In this work, we pre-encoded the spin-lock pulse cluster onto a 2D fast spin-echo pulse sequence .
Each animal was positioned rear-first and prone in the magnet bore of a 4 Tesla Signa clinical MR scanner (General Electric Medical Systems, Waukesha, WI). The animals underwent T1ρ MRI with imaging parameters of FOV = 8 cm × 8 cm, slice thickness = 3 mm, acquisition matrix = 256 × 128, and TE/TR = 17/4000 ms for a total imaging time of 2 min for a single slice. The images were obtained using a linear transmit/receive radiofrequency (rf) surface coil to provide adequate signal-to-noise ratio in the region of the patella (~60:1). The band of 90° excitation produced by the surface coil, and hence the region of full spin-lock effect, was calibrated for each image session to lie in the region of the patella of the specimen by increasing rf transmit power. We previously validated this approach of T1ρ mapping using a surface coil on a phantom . For each knee, five images of the same 3-mm thick slice were acquired using evenly spaced TSL times from 10 to 50 ms in order to provide data for T1ρ mapping. The nonselective nature of the spin-lock pulse cluster employed in the T1ρ-weighted pulse sequence restricted our acquisition to only a single slice rather than the entire patella. Thus, a slice from the central region of the patella was chosen to represent the changes in T1ρ in the cartilage. The harvested patellae were imaged with a quadrature birdcage coil using the same scanner equipment with similar imaging parameters as the in vivo situation with the exception of using a 4 cm × 4 cm FOV. Initially, a multi-slice image set was acquired on the ex vivo patella. The location of the ex vivo slice in which to measure T1ρ was chosen from the multi-slice set as the slice that most closely matched the in vivo slice upon visual inspection.
For each data set, the T1ρ image series was used to compute a least squares regression estimation on a pixel-by-pixel basis of the T1ρ parameter according to Eq. (1). In this manner, we created a spatial map in which each pixel represented the estimation of the T1ρ parameter at that position. In order to minimize the effect of image noise in the T1ρ mapping process, the image data were smoothed with a 3 pixel-wide Gaussian filter prior to the fitting process. Using a visualization software program custom-written in IDL (Research Systems Inc, Boulder, CO), we manually segmented the cartilage of each patella by drawing a region-of-interest covering the entire patella on the 2D T1ρ map. Each data point was measured as the mean and standard deviation of the T1ρ data within the segmented region. In this way, the global value of T1ρ in the entire central region of the patellae was measured. The T1ρ measurements were converted into R1ρ data by taking the reciprocal of T1ρ.
The in vivo measurement of FCD via sodium MRI was carried out on the same pigs in a concurrent study . This approach to FCD measurement was employed in order to noninvasively measure in vivo FCD in the same region as the T1ρ measurement. The method of FCD measurement via sodium MRI is summarized in brief. Immediately following the T1ρ measurement, in vivo sodium MR images were acquired on the same 4 Tesla clinical scanner using a 3D gradient-echo sequence with the following imaging parameters: FOV = 15.2 cm × 15.2 cm, slice thickness = 5.7 mm, acquisition matrix = 256 × 64, TE/TR = 2.4/40 ms. Due to the low natural abundance and gyromagnetic ratio of the sodium nucleus, sodium MR images were acquired at a lower resolution than the T1ρ-weighted images in order to produce data with sufficient signal-to-noise for FCD quantification. The proton surface coil used to acquire the T1ρ data was carefully replaced within the scanner by a sodium surface coil tuned to the resonant frequency of sodium at 4 T (45 MHz) so that the knee of the animal remained in the same position as the T1ρ images. In this manner, sodium MR images were acquired using the same spatial positioning as the T1ρ experiments in order to capture the central region of the patella corresponding to the location of the T1ρ-weighted images. Sodium content maps were calculated from the sodium MR data using a previously established methodology . Fixed charge density data were calculated from the sodium maps assuming ideal Donnan equilibrium.
In an effort to assess the accuracy of the in vivo T1ρ measurement, the in vivo and ex vivo measurements on each sample were compared using a paired Student’s t-test. Statistical significance was evaluated using the statistical software package JMPIN (SAS Institute, Cary, NC). To analyze the relationship between changes in PG content with changes in T1ρ, the R1ρ data were compared with the FCD data measured via sodium MRI conducted on the same pig specimens . The sodium MRI approach has been previously shown to be an accurate method of noninvasive FCD quantification as validated by spectrophotometric assay . The significance of the relationship between FCD and R1ρ was analyzed via analysis of variance (ANOVA) performed using JMPIN. The linearity of the relationship was measured via the correlation coefficient of the regression between FCD and R1ρ data.
The histological and immunochemical analyses were performed as an independent confirmation of the loss of cartilage PG content and the effects of IL-1 β. The synovial fluid was removed when the animals were euthanized and samples were analyzed for the presence of effects due to IL-1 β including changes in PG and PG-degrading enzymes. Routine smears were prepared and Geimsa stained and the cellular characteristics analyzed. The ex vivo patellae were processed for routine histology and stained for PG content using Safranin-O employing standard laboratory procedures. Histological slides taken from the central region of the patellae, corresponding to the location of the MR images, were microscopically examined using a Nikon E-600 with Spot digital camera or scanned using a Nikon Super Coolscan 4000 ED digital slide scanner (Nikon USA, Melville, NY).
As an indicator of the action of IL-1β in the joint we examined the joint fluid for the up-regulation of MMP. We chose as an indicator MMP-13 of which aggrecan is a substrate . The presence of MMP-13 was detected by immunoprecipitation and Western blotting using specific antibodies. Saline- and IL-1 β-treated joints were processed together. Immunoprecipitations were carried out as previously described  and according to the manufacturer’s (Santa Cruz Biotechnology) instructions with the following exceptions. Each synovial fluid was immunoprecipitated followed by Western blotting with the same antibody to identify MMP-13. Fifty microliters of synovial fluid was precleared overnight with 1 μg of control secondary anti-mouse antibody followed by 20μl of protein A/G (Santa Cruz Biotechnology, Santa Cruz, CA) and a 30 min incubation. One microgram of anti-pro- MMP-13 (Cat. #MAB913, R&D Systems) was added for 1.5 h and the preceding steps were followed as outlined by the manufacturer. The resultant immunoprecipitated MMP was then subjected to SDS-PAGE and Western blotting using the same antibodies. The SDS-PAGE and Western blots were performed by routine methods as previously described . Bands of corresponding molecular sizes were identified as pro-MMP-13 on Western blots, and bands that resulted from carryover immunoglobulins from the immunoprecipitation were clearly identified with lanes from precipitates from incubations without synovial fluid.
Preliminary studies performed on three pigs given injections of IL-1 β at 1, 6, and 12 h before imaging produced changes in PG content (14%, 50%, and 61% loss of FCD in IL-1 β-treated patellae vs. saline-treated, respectively). The observed change between the 1 and 6 h data points was substantially greater than the change observed between the 6 and 12 h data points. Hence we conjectured that the cascade of events initiated by the introduction of IL-1β in our model stabilized between 6 and 12 h. This result is in agreement with our hypothesis of cytokine effects on cartilage and joint pathology in vivo. As a result, we chose to inject the animals 6 h before imaging to provide an easily identifiable change in PG content that is characteristic of degenerative joint disease in cartilage thereby providing a physiologically relevant test model.
A set of in vivo T1ρ maps from a representative animal are displayed in Fig. 1. Table 1 contains a summary of the in vivo T1ρ and FCD data from all six pigs. There was a significant elevation in T1ρ in the IL-1β-treated patellae compared to the saline-treated patellae according to a two-tailed paired Student’s t-test (p < 0.001). The average T1ρ values of the in vivo saline- and IL-1β-treated patellae were 54 ± 5 and 74 ± 8 ms (mean ± standard error), respectively. In vivo measurements of R1ρ yielded averages of 18.5 ± 1.0 and 13.8 ± 0.9 s−1, respectively. The average intra-specimen decrease in R1ρ between the saline- and IL-1β-treated patellae was 25 ± 4%. The FCD data yielded an average FCD of the IL-1β-treated patellae (−95 ± 11 mM) that was measured to be 49% lower than that of control patellae (−185 ± 9 mM), indicating a loss of PG content. The relationship between R1ρ and FCD was statistically significant (p = 0.001) according to an ANOVA test. There was also a significant correlation in the linear regression of R1ρ and FCD data (R2 > 0.77, p < 0.001).
The accuracy of the in vivo T1ρ measurements was validated by comparison with measurements obtained from high resolution ex vivo data. A comparison of in vivo and ex vivo T1ρ maps of an IL-1β-treated patella from a representative specimen is displayed in Fig. 2. The in vivo T1ρ maps exhibit a global elevation in T1ρ but lack the resolution to clearly show spatial variations within the cartilage. The high resolution ex vivo T1ρ maps more faithfully display the heterogeneity of T1ρ within the tissue. We typically observed higher T1ρ values in the middle zone of the cartilage like those in Fig. 2. The in vivo and ex vivo data were not statistically different according to a paired two-tailed Student’s t-test (p > 0.89). In addition, the in vivo and ex vivo data were highly correlated (R2 = 0.86, p < 0.001).
The cartilage of all specimens was observed to be smooth and devoid of fibrillation upon histological inspection thereby substantiating the understanding that IL-1β produces biochemical changes in the tissue with little change in morphology. The effects of IL-1β in the joint were confirmed by examining the synovial fluid and cartilage via immunochemistry and histology, respectively. There was a reproducible reduction in histological staining for PG by Safranin-O for IL-1β-treated cartilage as compared with control cartilage. Loss of proteoglycan was generally noted in the middle zone of the cartilage as indicated in Fig. 3. There was a marked increase in the cellularity of the synovial fluid (polymorphonuclear leukocytes and monocytes) in the IL-1β-treated joint. In addition, a marked increase in one of the aggrecan-degrading enzymes, MMP-13, was detected by immunoprecipitation and Western blot in the synovial fluid of the IL-1β-treated joint confirming the predicted action of IL-1β. Fig. 4 shows the increase in pro-MMP-13 in the fluid of the IL-1β-treated joints and is representative of all samples tested.
The strong correlation coefficient of the regression between in vivo and ex vivo T1ρ data establishes the accuracy of the in vivo measurement of T1ρ using a surface coil. The close relationship between the in vivo and ex vivo data (p > 0.89) supports the assumption that the in vivo T1ρ data are devoid of contamination from contributions from adjacent fluid and possible errors resulting from an imperfect 90° flip angle in the region of the patella. The close agreement of the in vivo and ex vivo images of the patella in Fig. 2 illustrates how the location of the ex vivo slice was carefully chosen to correspond to the position of in vivo slice.
Due to the single-slice restriction imposed by the pulse sequence, T1ρ and FCD data were measured in only a representative section of each patella rather than in the whole cartilage. The measurements of T1ρ and FCD were gathered from the entire cartilage area on each slice in a global sense by using a region-of-interest approach. Due in part to the discrepancy in resolution of each method, we did not attempt a point-to-point comparison of the T1ρ, FCD, and histology data. Despite the lower resolution due to experimental conditions, the in vivo T1ρ maps exhibited a global elevation of T1ρ in IL-1β-treated patellae. The heterogeneity of T1ρ in the cartilage was more clearly evident in the high resolution T1ρ maps (Fig. 2) and is represented in part by the standard deviation of the T1ρ data (Table 1). The increase in T1ρ middle layer of the cartilage that is evident on the high resolution ex vivo maps is consistent with the loss of PG in the mid-zone of the cartilage in the histological images. In future experiments with optimized experimental equipment and parameters, in vivo resolution of both sodium and T1ρ data could be increased while maintaining sufficient signal-to-noise thereby enabling a possible point-to-point analysis.
Several experimental considerations need to be taken into account in the implementation of the T1ρ MRI technique. The use of long duration spin-lock pulses in the T1ρ-weighted pulse sequence substantially adds to the specific absorption rate (SAR) of rf energy by the sample which is regulated by the US Food & Drug Administration. Possible concerns about SAR can be alleviated by using an approach based on partial k-space acquisition in which SAR is greatly reduced while nearly full T1ρ-weighted contrast is maintained in the images . Also, the use of B1-sensitive spin-lock pulses necessitates the use of rf hardware with good B1 homogeneity. Potential image artifacts arising from any B1 inhomogeneity can be mitigated by using a phase-alternating self-compensating spin-lock pulse  as evidenced in this work and other reports using a surface coil to generate artifact-free T1ρ-weighted images .
In order to detect changes in PG in a biologically relevant system, we chose to use IL-1β since its effects on PG are well-established. As measured by a reduction in R1ρ, there was a marked decrease in PG content in the IL-1β-treated patella in comparison with the saline- treated patella. The occurrence of PG loss in the IL-1β tissue is corroborated by a reduction in histological staining for PG and decreased sodium concentration as measured by sodium MRI . The highest loss was seen in the mid zone as a result of diffusion of IL-1β on responsive chondrocytes. Some degradation could also arise from effects of IL-1β on cells of the synovial membrane and in the synovial fluid, presumably stimulating MMP biosynthesis. The identification of MMP in synovial fluid along with histochemical findings substantiates IL-1β activity as a plausible pathway of PG loss in cartilage.
In this work, we demonstrated that R1ρ tracks changes in PG content rapidly induced using the physiologically important IL-1β. There have been numerous reports of similar cytokine-induced models in the literature [5,22] but none have been analyzed in the manner presented in this report. The salient contribution of this work lies in the investigation of the feasibility of the T1ρ MRI technique to detect minor degenerative changes in PG. The T1ρ technique is flexible enough to be easily applied to other established in vivo or in vitro models of OA. Several experimental advantages led us to choose the particular cytokine-induced porcine model used in this report. By using a large animal model of nearly the same size and weight as a human subject, we were able to design our study to be readily transferable to human disease assessment and clinical application using currently available technology. The knee joint of the pig provided a large load-bearing joint that has similar anatomy to that of a human. The size of the joint allowed us to use identical rf coils, experimental parameters, image resolution, and whole-body clinical MR system as would be used in the case of a human subject. Our study provides the basis for expanding this approach for use in the assessment of human degenerative joint disease.
This work was performed at the MMRRCC, an NIH-supported resource center (NIH RR02305), with support from NIAMS-NIH grant R01-AR45404, a grant from the Arthritis Foundation, The Whitaker Foundation Graduate Research Fellowship, and the Nemours Foundation. The authors would like to thank Christine Gardiner for technical assistance with the immunochemistry, Theresa Michel for editorial assistance, and the veterinary technicians from the University Laboratory Animal Resources facility at the University of Pennsylvania.