Sodium Intensity Changes Differ Between Relaxation- and Density-Weighted MRI in Multiple Sclerosis

Introduction: The source of Tissue Sodium Concentration (TSC) increase in Multiple Sclerosis (MS) remains unclear, and could be attributed to altered intracellular sodium concentration or tissue microstructure. This paper investigates sodium in MS using three new MRI sequences. Methods: Three sodium scans were acquired at 4.7 T from 30 patients (11 relapsing-remitting, 10 secondary-progressive, 9 primary-progressive) and 9 healthy controls including: Density-Weighted (NaDW), with very short 30° excitation for more accurate TSC measurement; Projection Acquisition with Coherent MAgNetization (NaPACMAN), designed for enhanced relaxation-based contrast; and Soft Inversion Recovery FLuid Attenuation (NaSIRFLA), developed to reduce fluid space contribution. Signal was measured in both lesions (n = 397) and normal appearing white matter (NAWM) relative to controls in the splenium of corpus callosum and the anterior and posterior limbs of internal capsule. Correlations with clinical and cognitive evaluations were tested over all MS patients. Results: Sodium intensity in MS lesions was elevated over control WM by a greater amount for NaPACMAN (75%) than NaDW (35%), the latter representing TSC. In contrast, NaSIRFLA exhibited lower intensity, but only for region specific analysis in the SCC (−7%). Sodium intensity in average MS NAWM was not significantly different than control WM for either of the three scans. NaSIRFLA in the average NAWM and specifically the posterior limb of internal capsules positively correlated with the Paced Auditory Serial Addition Test (PASAT). Discussion: Lower NaSIRFLA signal in lesions and ~2× greater NaPACMAN signal elevation over control WM than NaDW can be explained with a demyelination model that also includes edema. A NAWM demyelination model that includes tissue atrophy suggests no signal change for NaSIRFLA, and only slightly greater NAWM signal than control WM for both NaDW and NaPACMAN, reflecting experimental results. Models were derived from previous total and myelin water fraction study in MS with T2-relaxometry, and for the first time include sodium within the myelin water space. Reduced auditory processing association with lower signal on NaSIRFLA cannot be explained by greater demyelination and its modeled impact on the three sodium MRI sequences. Alternative explanations include intra- or extracellular sodium concentration change. Relaxation-weighted sodium MRI in combination with sodium-density MRI may help elucidate microstructural and metabolic changes in MS.


Healthy Control White Matter
White matter tissue volume fractions were calculated for healthy tissue using the framework presented in the study of myelin water fraction (MWF) with T2 relaxometry by Laule et. al. (Laule et al. 2004). Using values of myelin water content = 0.369 (g H2O)/(g myelin) and non-myelin water content = 0.82 (g H2O)/(g non-myelin), Laule et. al. calculated water and dry macromolecule fractions of tissue by weight for a given MWF. These weight fractions (in grams per gram of tissue) are given in Figure 6a of reference (Laule et al. 2004) for healthy control white matter with MWF = 0.114. Here myelin water accounts for 0.082 g, while non-myelin water accounts for 0.638 g.
Myelin and non-myelin volumes can then be calculated for 1.0 gram of healthy control white matter using the density values of 1.08 for dry myelin and 1.33 for dry non-myelin (also from (Laule et al. 2004)). These tissue volumes are given in Figure 7A1, where non-myelin volume is the combination of both intra-and extracellular space (the splitting of which is described below).
Given that the density of water is 1.0, the volumes of myelin and non-myelin water are 0.082 mL and 0.638 mL respectively, while the total tissue volume is 0.955 mL (also listed in Figure 6a of reference (Laule et al. 2004)). Thus, myelin water accounts for 9% of healthy control white matter volume ( Figure 7A2). Non-myelin water (78% of the tissue volume) was then split into intra-and extracellular components such that an extracellular volume fraction (ECVF) of 20% was achieved (Sykova and Nicholson 2008;Nicholson and Hrabetova 2017), where this 20% includes both the extracellular water (17%) and the dry macromolecules (3%) in the ratio of 86% water and 14% dry macromolecules for non-myelin space (volume ratios calculated from the non-myelin water content and molecule densities given above). Subtracting the extracellular volume fraction from the non-myelin water volume fraction yields an intracellular volume fraction of 58%. Splitting this into water and dry macromolecule components using the same non-myelin volume ratios yields an intracellular water fraction of 50%. Summing all of the water spaces yields a total water fraction of 75%. (Figure 7A2).

NAWM with Demyelination
Using multi-exponential T2 'myelin water imaging', Laule et. al. experimentally measured 16% lower MWF and 2.2% greater tissue water content in NAWM compared to controls, and suggested a demyelination model (Figure 6c in reference (Laule et al. 2004) and Figure 7B) explained this result better than either edema (Figure 6b in reference (Laule et al. 2004)) or cellular infiltrates ( Figure 6d in reference (Laule et al. 2004)). However, a simple demyelination model can not simultaneously describe 16% lower MWF and 2.2% greater tissue water content, and thus Laule's model included a 17.7% lower MWF = 0.094, through which the 2.2% tissue water increase was attained. In this case, the loss of myelin yields a tissue volume of only 0.914 mL, describing tissue atrophy of 4.3% (also given in Figure 6c of reference (Laule et al. 2004)), a value reflecting previous study (Ge et al. 2001). As a result, the relative WM volume fraction associated with intraand extracellular space is increased (Figure 7B2), even if the volumes of these spaces themselves remain the same (Figure 7B1).

Lesions with Demyelination and Edema
A model which includes both demyelination and edema explains both the lower MWF and greater water content in lesions. In Figure 7C, lesions are modelled using values extrapolated from (Laule et al. 2004). Note that while a mean lesion MWF = 0.046 was experimentally measured in (Laule et al. 2004), mean healthy control white matter does not represent the same directly comparable tissue. The lesion water increase of 8.6% (relative to control white matter) is derived from the measured 6.3% increase over contralateral NAWM, and the average 2.2% water increase in NAWM of specific structures compared to control white matter (i.e 1.086 = 1.063 * 1.022).
Nevertheless, these values provide initial estimates for a model which includes both demyelination and edema to explain both lower MWF and greater water content in lesions.

23 Na Relaxation Models Used to Estimate Sequence Weighting
The same (total tissue based) model is used for both intra-and extracellular space, and is derived from the biexponential T2 fitting described in (Stobbe and Beaulieu 2016) and the T1 fitting described in (Stobbe and Beaulieu 2006) is used for both intra-and extracellular space. For the myelin water space, zero-mean Gaussian distributed residual quadrupole splitting with standard deviation of 625 Hz is added to the previous model, a data fitting estimate from reference (Stobbe and Beaulieu 2016). Finally a model for edema was also derived from relaxation measurements of blood plasma (Perman et al. 1986).
Spectral density parameters of edema were taken as J0 = 0.074, J1 = 0.0094, J0 = 0.0094, for T2fast = 12 ms, and a T1 = 53 ms similar to saline at 4.7T. Note that for each environment, 23 Na relaxation parameters remain to be directly measured. This may require new regression methods, such as described in (Kordzadeh et al. 2020). Nevertheless, 'best estimate' models are used to see if they might help explain experimental image contrast.