High throughput method for detecting murine brain atrophy using a clinical 3T MRI

Multiple sclerosis (MS) is an inflammatory, demyelinating disease of the central nervous system (CNS) associated with disability progression in most patients. Although there are 23 FDA-approved therapies for MS which target inflammation, there are no medications that specifically address MS demyelination or disability progression, aspects of the disease which are poorly understood. Unlike lesional measures which do not correlate well with disability(“the clinico-radiologic paradox”) [1], CNS atrophy, measured by MRI, has had strong correlations with disability progression [2].

In order to learn more about CNS atrophy and progressive accrual of disability in MS, we utilized a mouse model of progressive neuroinflammation and neurodegeneration. Theiler’s murine encephalomyelitis virus-induced demyelinating disease (TMEV-IDD) is a virus-induced model that can recapitulate certain aspects of progressive MS, such as it is a chronic disease that features a slow accrual of disability. Cerebral atrophy in TMEV-IDD has recently been shown by one group to precede and predict disability progression [3, 4], which may serve as an important biomarker in this model. Further validation of this biomarker could assist in understanding the mechanisms of CNS damage in MS. We wished to confirm these findings and correlate cerebral atrophy, measured as ventricular enlargement by MRI, with immunological and neurobehavioral outcomes. To fully replicate the previously mentioned studies, we also focused our measurements on lateral ventricular enlargement because the lateral ventricles, in mice, contain most of the CSF space, by far. We have previously demonstrated that spinal cord MRI using diffusion tensor imaging (DTI) is very useful as a disability biomarker in TMEV-IDD [5]. However, ventricular volume assessment is easier, faster, and does not require a specialized animal MRI [6]. This technique would provide a more feasible option for many investigators working in MS animal models such as TMEV-IDD or experimental autoimmune encephalomyelitis (EAE).

As in previous TMEV-IDD experiments, sham mice slowly improved their Rotarod performance over time, while the Rotarod performance of the TMEV-IDD mice deteriorated over time. TMEV-IDD mice showed a range of diminished Rotarod performance, but as a group progressively worsened over time, with the median NFI of TMEV-IDD mice at 6 months p.i. being 62% that of the sham group. The five E-TVV mice had significantly lower NFI than the sham mice(p = 0.0032), while the N-TVV mice (Fig. 3) did not (p = 0.33).

The mechanisms of CNS destruction in MS leading to progressive neurological disability are unknown. Multiple beneficial and detrimental factors may be involved, among them inflammation, including both innate and adaptive immunity, neuronal injury, demyelination and remyelination, and plasticity. TMEV-IDD represents an excellent model of progressive disability associated with inflammatory demyelination. This allows teasing out of critical molecules involved in this complex process and identification of reliable CNS damage measures. In the experiments outlined above, we attempted to ascertain whether any of a range of immunological measures might correlate with enlarged ventricles, and to confirm previous findings that enlarged ventricular size in TMEV-IDD correlates with neurological disability [3].

Other groups have referred to their findings in rodent models as “brain atrophy”, and we agree that atrophy likely contributes to ventricular enlargement in this model. However, we choose to use the more precise term “enlarged total ventricular volume” (E-TVV) since we cannot rule out the possibility that other processes, e.g., hydrocephalus, may contribute to ventricular enlargement. In fact, given the considerable amount of CNS inflammation in TMEV-IDD [7, 16], hydrocephalus may be a contributing factor, especially considering the rapid development of E-TVV and its partial resolution over time, two features we observed that are less likely to occur if the cause was solely atrophy. This may indicate that TMEV-IDD may replicate disrupted CSF flow through the glymphatic pathways. A disruption to that flow could lead to the rapid ventricular enlargement and as that aspect of TMEV-IDD subsides the ventricles reduce in size but do not return to their baseline levels. This indicates some level of permanent damage and atrophy. Determining the precise cause(s) of E-TVV in this model will require larger MRI and pathology studies. This will provide the necessary data to determine the likely cause of the CNS injury that we observed.

Our data confirm that E-TVV does occur in the brains of TMEV-IDD mice, but this was a significant finding in only half of our mice. Our data also confirm that these E-TVV mice had worse neurological function as measured by Rotarod than those with normal total ventricular volumes(N-TVV). This indicates that E-TVV is a physiologically relevant biomarker for CNS injury in TMEV-IDD and those changes can be detected utilizing a clinical 3 T scanner. Another advantage of this MRI measure is that in 4 of 5 E-TVV mice, the TVV was already enlarged at one-month p.i., long before the development of disability, which usually begins around 75–90 days post infection. This finding raises the possibility that a clinical MRI could be used to identify mice that are more likely to develop weakness and are predisposed to the development and accumulation of disability.

There were some notable differences between our studies and results and those of the Mayo group [4]. First, we used the BeAn strain of TMEV while the Mayo group used the DA strain. There are apparent differences between the disease induced by these two strains [17], with BeAN-induced disease being generally less severe. However, it has not been shown that one strain is better or should be used preferentially over the other strain. Each strain, BeAN and DA, has its uses, but it is definitely possible that slight dfferences in viral strain activity could have caused the disparity in our results. Second, we used a 3 T clinical MRI scanner, which is designed for human imaging, while the Mayo group used a 7 T small bore animal machine, which afforded greater resolution. Third, the time course of ventricular enlargement in our mice peaked early, i.e., at one-month p.i. in our mice, while the peak in the Mayo studies occurred at 3 months p.i.

A major finding of this study is that our use of the 3 T clinical scanner demonstrates that this type of analysis of ventricular volumes can be performed without a specialized animal magnet of 7 T or higher field. The use of a 3 T clinical scanner is a real advantage given that many research centers will not have access to a high field small animal scanner. A further advantage of the 3 T scanner is that the larger bore of the 3 T scanner allowed 4 mice to be scanned simultaneously, which substantially reduced the time required per experiment because multiple mice can be tested simultaneously. However, given the lower resolution and sensitivity of the 3 T relative to the 7 T system, ventricular volumes determined by 3 T cannot be compared directly to those determined by 7 T systems and require comparison to internal controls, such as the sham mice used in our study. The precision of measurements is necessarily impacted by partial volume effects and a lower signal to noise ratio compared to scans at 7 T. It is possible that precision could be increased by more sophisticated surface-based modeling of the ventricles. A more advanced coil, use of contrast agents, or the addition of pulse sequences could also improve acquisitions sufficiently to increase data quality or reproducibility. However, the novelty and relevance of this study is that the scanning that was performed utilized widely available clinical sequences. This provides a potential solution for imaging centers that would like to perform preclinical MRI studies but do not have access to an animal MRI scanner.

A limitation of this study is that the ventricle volumes were manually calculated. This can be an issue due to possible bias or variability between the individuals calculating ventricular volume. We attempted to overcome this limitation by ensuring that those who manually calculated the volumes were blinded to the health status of the mice and to the volumes calculated by the other investigator. Using these strategies, we were still able to obtain a high level of rater similarity. However, as we continue this work, we expect that sophisticated automated segmentation approaches could be feasible in this domain and would leverage our initial manually segmented training data. Future efforts will also be made to perform larger studies with a cross-sectional design that will help reduce the baseline variance.

We chose measures primarily relevant to B cells, antibodies, and complement, for our immunological analyses, because B cell depleting drugs are highly effective in MS [18]. B cell active chemokines are correlated with MS neuroinflammation [19], and intrathecal antibodies [20, 21] have been implicated in CNS injury in the disease. None of these immunological measures correlated with the development of enlarged ventricular volumes. This is consistent with our previous results in TMEV-IDD, where multiple immunosuppressive or immunomodulatory medications have downregulated neuroinflammation without ameliorating disability progression [7‐10]. It is also consistent with results in human MS, where most disability accrues in the secondary progressive phase of the disease. One of the most potent immunosuppressive MS drugs, natalizumab, did not reduce progression on the primary multicomponent disability endpoint used in a large, phase 3 randomized, double-blind, placebo-controlled trial in secondary progressive MS [22].

All these data indicate that we do not yet have an adequate understanding of the basic mechanisms of CNS injury in MS and its models, similar to our lack of understanding in other chronic neurological diseases such as Alzheimer’s disease, amyotrophic lateral sclerosis, and Parkinson’s disease. CNS atrophy occurs in numerous neurological disorders, and thus our finding has direct relevance to these other diseases. In addition, there is mounting evidence that there are immunological or viral elements to these other conditions as well. Therefore, understanding CNS atrophy in an immune-mediated, virally induced mouse model, like TMEV-IDD, could be useful to fields beyond MS. Magnetic resonance imaging in TMEV-IDD can thus provide a useful tool to address this formidable challenge facing the neuroscience community of identifying targets relevant to neurodegeneration in MS, and other conditions.