Normal-appearing White Matter Quantitative Magnetization Transfer Imaging correlates with Disability in Multiple Sclerosis

Introduction Multiple sclerosis (MS) is an inflammatory disease of the central nervous system, whose neuroradiological hallmark are multifocal white matter (WM) lesions. Several pathological and MRI studies based on quantitative techniques have clearly demonstrated that diffuse abnormalities are present in the normal appearing WM (NAWM) and in lesions, and that they partially explain symptom severity. Nevertheless, the correlation between lesion loads, indexes of NAWM damage, and measures of disability still remain largely unsatisfactory. Magnetisation transfer imaging (MTI) is based on the exchange of magnetisation existing between protons in tissue water and those bound to the macromolecules. This technique has already been used to investigate MS and it was found to correlate with myelination and axonal count in a post-mortem study [1]. Quantitative MTI (qMTI) [2] is an extension of MTI which allows the binary spin bath model parameters to be estimated. The application of qMTI to the study of MS is promising although it has been limited so far, due to the relatively long scan times required. Aims of this study are to use a clinically feasible whole-brain qMTI protocol to investigate the different pathological substrates of NAWM and lesions in MS, and to assess whether qMTI parameters may explain the clinical disability of relapsing remitting (RR) MS patients. Methods We recruited 13 patients with RRMS [F/M ratio=8/5; mean (SD) age=39.38 (9.6) years; median (SD) expanded disability status scale [EDSS] score 2 (range 1.0-4.0)] who underwent a neurological examination and an MRI acquisition at 3.0T. Fourteen agematched healthy subjects were recruited as controls [F/M ratio=6/8; mean (SD) age=35.14 (8.37) years]. The MRI session included for every subject: 1) a dual-echo turbo spin echo (TSE); 2) a fluid attenuated inversion recovery (FLAIR) scan; 3) a Magnetization Prepared Rapid Gradient Echo (MPrage) for tissue segmentation; 4) a series of 12 MT-weighted 3D FLASH sequences, with various combinations of amplitude and offset frequency of the MT pulse, optimised according to [3]; 5) three 3D FLASH sequences with variable flip angle for T1 mapping [4], and 6) three 3D FLASH sequences with near-180° flip angles for B1 mapping [5]. White matter lesions were outlined on proton density weighted scans using a semi-automated contouring technique (http://www.xinapse.com/), using the FLAIR and the T2-weighted scans as reference. T1 and B1 maps were obtained as described in [4] and [5], respectively. We fitted Ramani's model [6] of MT to the data to yield RA, F, T2B, and RM0B (where R is the exchange rate, F= M0B/M0A, with M0A and M0B being the fully relaxed values of longitudinal magnetisation for the two pools, and RB=1s). The MPrage scans were segmented in SPM5 (http://www.fil.ion.ucl.ac.uk/spm/) to produce a map of WM. An image of the NAWM was obtained by subtracting from this WM image the voxels corresponding to lesions. After image coregistration, NAWM masks were applied to the RM0B, T2B, F, and RA maps to compute the mean NAWM parameters. T tests were performed to compare mean NAWM parameters of patients with, respectively, the mean WM parameters of controls, and with patients lesions. Correlation between lesion load values and qMT parameters were calculated. Two models were investigated using ordered logistic regressions, with the EDSS as the dependent variable. In the first one, mean NAWM MT parameters (RM0B, T2B, F, RA) and lesion load were entered as explanatory variables; in the second one, mean MT variables within lesions and lesion load were entered as explanatory variables. In both models age was added as a covariate of no interest. Results The mean T2 lesion load in patients was 5.95 (SD=4.96)mL (Table 1). Mean comparisons showed significantly lower values of F and RM0B, and approaching significance for RA in patients NAWM compared to controls WM (Table 1). Significantly higher values including RM0B, T2B, F, and RA were also observed when comparing patients NAWM and lesions (Table 2). Significant inverse correlation was found between T2 lesion load (p=0.01) and both F (r=-.64) and RA (r=-65) in patients NAWM. Within the NAWM, the best variable predicting EDSS score was T2B (adjusted for age), which was inversely correlated with EDSS. The model was less significant when adding remaining MT parameters and lesion load values as additional explanatory variables. Conversely, in patients lesion model, disability scores were the best predicted by T2B, together with RM0B, F, RA, and lesion load (again, adjusted for age). Discussion F and RM0B values were significantly lower in patients NAWM than in controls WM. In the NAWM, F (which is thought to correlate with myelin content) and RA were the only parameters associated with lesion load, suggesting the presence of Wallerian degeneration, at least to some extent. RM0B, T2B, F, and RA were significantly lower in lesions than in NAWM, confirming that qMT parameters reflect structural properties of tissue. Unexpectedly, despite its minimal variation across individuals, T2B was the most significantly associated with EDSS in NAWM. This parameter is likely to represent a weighted average of the relaxation times of spins with different molecular environments, and therefore its reduction in MS lesions could be interpreted as a change in the balance between subpopulations of macromolecular spins. Conversely, in lesions, RM0B, T2B, F, and RA and lesion load significantly predicted disability only when combined together. This might reflect the complex interaction between demyelination, remyelination, gliosis, inflammation and axonal loss taking place within lesions. References 1. Van Waesberghe JH, et al. Ann Neurol 1999; 46: 747-54. 2. Henkelman RM, et al. Magn Reson Med 1993; 29: 759-66 3. Cercignani M, et al., Magn Reson Med 2006; 56: 803-10. 4. Venkatesan R, et al. Magn Reson Med 1998; 40: 592-602. 5. Dowell N, et al. Magn Reson Med 2007; 58: 622-30. 6. Ramani A, et al., Magn Reson Imaging 2002; 20; 721-31. Parameter Control WM Mean (SD) NAWM Mean (SD) P value