Diffusion tensor imaging in metachromatic leukodystrophy

Metachromatic leukodystrophy (MLD, OMIM 250100) is an autosomal recessive lysosomal disorder caused by mutations in the ARSA gene. This results in deficiency of the enzyme arylsulfatase A (ASA), essential for sulfatide metabolism [1]. Sulfatides are major myelin lipids; their accumulation, mainly in membranes, leads to demyelination and subsequently storage in macrophages that cannot digest them [1]. MLD is a devastating disease: without treatment, eventually all acquired skills are lost and patients die.

MLD has three clinical subtypes, based on age of onset. The late-infantile form starts before 30 months, usually presenting with motor deterioration. The juvenile form presents with a combination of motor and cognitive decline before 16 years. The adult form begins with cognitive decline and psychiatric symptoms thereafter [2]. When performed early, hematopoietic cell therapy (HCT) has promising results, especially for juvenile and adult patients [3, 4].

Brain magnetic resonance imaging (MRI) in MLD is characterized by bilateral symmetric T2 signal hyperintensities, starting in the corpus callosum and subsequently involving periventricular, central and subcortical white matter (WM), in addition to projection fibers such as the corticospinal tract, and eventually cerebellar WM [5]. Thalamic volume and signal intensity on T2-weighted images in the thalamus are decreased already at diagnosis [6, 7]. Typical for MLD are stripes of low signal intensity throughout the hyperintense signal on T2-weighted images in the cerebral WM, related both to the accumulation of macrophages bursting with undigested lipids and to better preserved perivascular myelin [8].

Brain diffusion tensor imaging (DTI) is based on the motion of water molecules, which is more restricted perpendicular to than along WM fibers, a feature termed diffusion anisotropy [9]. Magnitude and direction of diffusivity are determined by molecules, membranes and microtubules, and provide information about tissue composition and microstructure and its architectural organization [10, 11].

The tensor model is a relatively simple model using diffusion-weighted images (DWI) obtained with one b value. It results in axial diffusivity (AD), radial diffusivity (RD) and the derived fractional anisotropy (FA) and mean diffusivity (MD) [11]. Often, increased RD is thought to be correlated with myelin degradation and changes in AD to axonal degeneration or inflammation and gliosis [12‐17], but it is difficult to unequivocally associate the interpretation of diffusivity variations with specific biophysical changes [18].

The precise pathomechanisms involved in MLD, such as importance of inflammation or how accumulated sulfatides lead to demyelination, are not completely understood. DTI is, taking into account its recognized limitations, a valuable tool to gain more insight into changes in tissue properties in MLD. We, therefore, compared diffusion measures (FA and the three diffusivities) between patients who were eligible for HCT, patients not eligible at time of diagnosis, and controls. HCT-eligible patients are typically in an early disease stage, while patients not eligible for HCT have more advanced disease with extensive demyelination of the WM. We also studied the longitudinal behavior of diffusion measures of both treated and untreated patients.

Using diffusion-weighted MRI, we compared magnitude and direction of diffusion between MLD patients and controls to gain insight into the microstructure of affected brain tissue. At baseline, FA was decreased and MD and RD were increased in MLD patients compared to controls throughout the WM, not only in the corpus callosum (affected early in the disease), but also in NAWM. FA measures of the thalamus did not differ between groups, but its components AD and RD were both increased in patients compared to controls. Whereas AD was increased in thalamus, it was unchanged in the pyramidal tracts and decreased in the corpus callosum and, to a lesser degree, in NAWM. All differences were most pronounced between controls and HCT-non-eligible patients.

Longitudinally, in treated HCT-eligible patients, diffusion measures remained stable or showed only minor changes. FA remained constant or even tended to increase in NAWM, pyramidal tracts, and thalamus, whereas it slightly decreased in the corpus callosum. HCT-non-eligible patients had less follow-up examinations than eligible patients, but those available showed clear increases of RD and AD, causing a small FA reduction in all investigated regions. The treatment effect of HCT most likely influenced the longitudinal differences between treated eligible patients (approaching control values in NAWM and pyramidal tracts), and untreated non-eligible patients (increasingly abnormal values). This is in line with our observation that metabolite concentrations observed with magnetic resonance spectroscopy partially normalized in successfully transplanted patients, whilst concentrations for non-treated patients further deteriorated [30].

The diffusion tensor model used in this study reflects the underlying structural characteristics in a simplified manner, hampered by partial volume effects and crossing fibers [18]. Advanced multi-compartment diffusion models, such as the composite hindered and restricted model of diffusion (CHARMED), are more sensitive than the conventional ones [31‐33]. However, since our study, ongoing since 2007, concerns a rare disease, application of these advanced diffusion models was not feasible.

This implies that we can merely hypothesize about the precise mechanism responsible for the observed differences rather than draw general conclusions because different cellular processes may lead to identical changes [9‐11]. Since both animal [13, 34, 35] and human studies [36] have shown an increased RD parallel to myelin loss, our results of increased RD in WM suggest myelin loss in patients, in line with histopathological findings [37]. This is also supported by our observation that high RD and low FA in the pyramidal tracts at baseline indicate poor motor function at follow-up.

With regard to AD, animal and human studies provide discrepant results, correlating axonal damage with either an AD decrease [13, 17] or increase, [38, 39], respectively. This reflects the difficulty in relating AD to underlying pathological processes. Our observation of opposing AD patterns in MLD suggests that different pathological mechanisms can cause either a decrease or an increase in AD, with the overall balance between these effects depending on brain region and disease stage. MLD is characterized by accumulation of sulfatides, major myelin lipids mainly synthesized by oligodendrocytes. Analytical studies of MLD patients’ brain tissue showed that metachromatic deposits are mainly present in the WM, with a sulfatide content up to eight times higher than normal, with relatively minor chemical GM changes [37]. Regarding WM, we assume that the massive intracellular sulfatide accumulation in swollen macrophages, in vain trying to digest these lipids, causes overall diffusion restriction and thereby AD reduction. Using conventional DWI, restricted diffusion in the outermost part of the demyelinated WM has indeed been described for single cases in relatively early disease stage [40, 41]. However, as the disease progresses, axons are increasingly damaged. Based on the previous observations in human studies, we expect that loss of both myelin and axons will lead to an AD increase [38, 39]. Our results suggest that, particularly in the corpus callosum of non-eligible patients at baseline, diffusion restriction due to sulfatide accumulation in macrophages contributes more to the severely reduced AD values than increased diffusion due to myelin and axonal loss. Our longitudinal observation of an increase, and thereby a pseudo-normalization of AD, in the corpus callosum suggests that myelin and axonal loss likely becomes more prominent in progressive disease. The corpus callosum is, of the investigated ROIs, least hindered by limitations of the tensor model, suggesting that the interpretation of AD values is not much influenced by the presence of crossing fibers. Whether inflammation and gliosis also contribute to AD changes in this setting is uncertain.

The thalamus is a DGM structure, in which sulfatide accumulation is much more limited than in WM [37]. In addition, the thalamus has a different microstructure than WM as it largely consists of neurons and contains few axons. In control subjects, this is mirrored by low thalamic FA, as the difference between AD and RD is much smaller than for a WM structure like the corpus callosum. In the thalamus of patients, we observed an increase of AD and RD of similar relative magnitude, which had hardly any effect on FA. This increase of both AD and RD probably implies an increase in extracellular space due to neuronal loss, which apparently dominates reductions in diffusivity due to storage material.

Limitations of this study were its retrospective character, a large age range of patients (inherent to the inclusion of patients with all disease types), and a limited age range of controls at 3 T. The combination of 1.5 and 3 T data also introduced some variability, although these differences were typically smaller than differences between controls and patients. In fact, the ROI-based analysis of baseline data at 1.5 and at 3 T separately (as shown in the supplementary material) showed identical patterns as those observed in the combined analysis. We noticed the same main group effects, although in these smaller groups fewer post hoc pairwise comparisons were significant.

Overall, the observed changes of FA, RD, and especially AD indicate that MLD alters certain aspects of brain microstructure. These changes most likely reflect a multitude of pathological processes such as accumulation of metachromatic material, followed by myelin and axonal loss. The differences between untreated and treated patients indicate that diffusion measures are positively affected by HCT, further emphasizing the beneficial effects of this intervention on WM and supporting the findings of other quantitative MR measures as proton MR spectroscopy [30]. Altogether, quantitative MR measures provide more insight into time-dependent disease mechanisms and might in the future aid in determining the right window for intervention.