imaging measurements have emerged as the outcome of choice for translational research in pre-clinical studies based on their potential to objectively quantify change and their compatibility with modalities used in clinical trials (27
). Since MRI and micro-CT are widely accepted as the “gold standards” to assess soft tissue and bone volumes respectively, we aimed to adapt these methods as longitudinal outcome measures of inflammatory-erosive arthritis in mice. In the case of CE-MRI for the mouse knee, several innovations were required (). These included: i) the generation of a mouse knee specific coil that can interface with a clinical 3T MRI, ii) the establishment of pulse sequences that produce high-resolution images (105μm) with a minimal slice thickness (160μm) to reduce partial volume effects, iii) a method to normalize for the variability of Gd-DTPA administration using muscle contrast enhancement, and iv) standardization for thresholding and segmentation of biomarkers for longitudinal-quantitative 3D analyses. Although commercial small animal MRI instruments may become more popular in the future, we chose to use a clinical MRI to ensure that all of the biomarkers we identified could be used to study arthritis in humans using readily available pulse sequences. In addition, the quantification methods developed could easily be adapted to data collected with small animal scanners.
Several quantification methods have been developed to assess synovial volumes in humans using CE-MRI (28
). However, no quantification methods have been established for clinical trials, for which there is a great demand. The currently accepted evaluation determined by the Outcome Measures in Rheumatoid Arthritis Clinical Trails (OMERACT) task group is the RA MRI scoring system (RAMRIS) consisting of semi-quantitative global scoring system (0–3) based on synovial thickness and Gd-DTPA enhancement (30
). The current gold standard for quantification of synovitis is a manual segmentation technique (14
). Although this technique has been validated in intervention studies, and with a correlation to an aspirated volume of synovial fluid, the length of analysis (0.75–2 hours) and technical expertise necessary has limited its use in clinical trails. To address these limitations, automated segmentation methods have been evaluated (32
). This two-step segmentation process consists of a limited manual segmentation to remove enhancing vessels and skin (similar to the limit lines used in the current study), and application of a threshold on subtracted images. Methods to threshold enhancing synovial tissue remain controversial. Østergaard et al. evaluated several different thresholds based on percent enhancement of synovium compared to the manual segmentation technique, and determined that a 45% enhancement threshold was optimal (32
). Although this reduced the time of analysis to 20 minutes, there was increased inter-MRI variation versus manual segmentation, particularly when misalignment occurred between pre- and post-contrast scans. Palmer et al. used a threshold value based on the signal intensity differences between several regions of non-enhancing tissue (suppressed fat) and enhancing pannus (33
). This method has an advantage in that the threshold is determined after each individual scan, however the time of analysis is around 45 minutes. The threshold is also sensitive to misregistration artifacts and the consistency of fat suppression.
To our knowledge, the current study is the first to use adjacent muscle as a normalization tissue to determine the synovial threshold. This approach is attractive due to the fact that muscle is an enhancing tissue that is not implicated in the pathology of inflammatory arthritis and because its tissue properties do not change significantly between scans, variations in values in muscle will reflect the signal variations that are present between scanning sessions. In small animal studies this approach is especially warranted due to the inconsistencies in dosage delivery that are inherent during administration via intravenous injection. During our dosage study, we found that muscle linearly enhanced with increasing dosage of Gd-DTPA. Even more important, a direct relationship could be derived between muscle contrast enhancement and the synovial threshold used to maintain a constant synovial volume between doses. The use of this optimized threshold, combined with the ability to minimize motion artifacts with anesthesia and standardized positioning, has allowed reproducible measurements of synovial volume in mice (4.5% coefficient of variation) that far exceed those reported in humans (22% median relative variation as measured with the manual segmentation method (34
)). Whether a similar threshold approach based on muscle enhancement could be adapted to clinical studies warrants further investigation.
Considering that micro-CT has been used to examine focal erosions in murine arthritis models for several years (35
), it is somewhat surprising that this approach has yet to evolve into a longitudinal-quantitative outcome measure. Based on our experience, we found that this is likely due to the difficulty in registering the baseline and outcome 3D-CT images so that the erosion as a negative change in volume could be accurately assessed. It is also clear from our work (), that segmentation can only be readily performed on small bones that are clearly defined by soft-tissue boundaries (e.g. patella and talus). This is because minor imperfections in the registration of large bones, and subjective segmentation of bone parts (e.g. distal femur and proximal tibia) can result in significant measurement errors.
In this study we focused our attention on the synovial and LN volume as the primary outcome measures of inflammation, based on their facile segmentation and quantification from CE-MRI. Interestingly, our findings indicated that these tissues behave differently during the onset and amelioration of inflammatory arthritis following effective therapy. We found that the popliteal LN volume was the most sensitive biomarker of lower limb arthritis. This conclusion comes from the observation that LN volumes significantly increase when TNF-Tg mice are 2.5 months-old (), which correlates with the point at which increased TNF serum levels and changes in peripheral blood mononuclear cell (PBMC) populations are first detected in this model (36
). Our finding that increases in LN volume precede knee synovitis is consistent with the function of popliteal lymph nodes in draining both the knee and ankle joints, and the fact that arthritis first occurs in the ankle in this model (23
). This conclusion was corroborated by similar results using the K/BxN serum transfer model (37
), in which the popliteal LN volume increased concurrently with ankle inflammation in the absence of knee synovitis (Supplemental Figure 2
). Thus, the introduction of pro-inflammatory mediators into the joint alone is not sufficient for the initiation of pannus formation. Our future efforts will focus on understanding the mechanism that initiates this destructive process. This issue is paramount since synovitis correlates with joint destruction in inflammatory arthritis (). Additionally, many groups have generated transgenic animals that can be used to dissect these pathologies (35
), which now can be assessed with our novel longitudinal outcome measures.
By using a validated drug therapy in our model we were able to address the major concern with all pre-clinical intervention studies of arthritis that are performed with a single cross-sectional endpoint, namely, whether or not the differences observed are due to drug effects or variability between the initiation and/or progression of disease in the individual mice. Using our previous study of anti-TNF treatment of TNF-Tg mice as an example (3
), we were unable to make definitive conclusions regarding the healing effects on bone and cartilage lesions since there were no baseline assessments. When we repeated this experiment using age as the enrollment criterion (), these concerns regarding inter-animal variability were realized based on the broad range of baseline synovitis (3.00 to 9.85 mm3
). Furthermore, during this study we saw a dramatic decrease in synovial volume in the placebo group from 12 to16 weeks that was due to pannus tissue fibrosis. Thus, by modification of the study design to include entrance criterion based on synovial volume (), we were able to make several remarkable observations including: i) demonstration that the treatment effect of anti-TNF therapy on synovial volume and LN volume is significantly greater than when entrance criterion is by age, ii) synovial volumes were significantly reduced to WT levels after only 4 weeks of anti-TNF therapy, iii) initiation of the study earlier in the disease process combined with an 8-week time course avoided the pannus tissue fibrosis effects previously observed, iv) a healing response was noted in which the mean lymph node volumes of the TNF-Tg mice treated with anti-TNF therapy were reduced but were still 2.5x higher than WT levels at the end of the study, and v) bone erosion was arrested as demonstrated by a significant difference in change in talus volume between anti-TNF and placebo animals.
In summary we found that CE-MRI and micro CT are very useful longitudinal outcome measures of the severity of inflammatory arthritis, as they demonstrated significant results with as few as four mice per group. We are also pursuing additional outcomes with the CE-MRI, including quantifications of bone marrow edema, which appears as a bright signal with this MR pulse sequence in the bones of arthritic animals versus the dark signal in WT controls (Supplemental Figure 3
). This biomarker has previously been demonstrated to be a faithful predictor of focal erosions in RA (15
), and its nature has recently been demonstrated to be osteitis (42
). Moreover, bone marrow edema measured by CE-MRI has been effectively used to assess disease severity and response to therapy in patients with ankylosing spondylitis (43
). Thus, the hope is that an edema outcome measure could be developed as the long sought after surrogate of arthritic pain in animal models and/or a predictor of joint destruction.