While much is known about MS-associated 1
H-MRS changes, their time progression is still unclear due to the limited utility of comparing cross-sectional studies using different patient cohorts and techniques. Despite the reiterated need for serial studies,1,3,4
practical constraints impede regular follow-up over considerable time periods. To our knowledge, the data here represent the most frequent 1
H-MRS follow-up for the longest duration. In addition, most previous studies employed metabolite ratios which decrease the specificity of 1
H-MRS, and used single voxels or region-of-interest analysis. In contrast to previous serial 1
H-MRS, this study assesses metabolism of a large brain volume, accounts for partial volume effects, and investigates widespread diffuse involvement.
High sensitivity without partial volume bias is achieved by 3D 1
H-MRSI in a large (360 cm3
) VOI using each voxel's signal and tissue composition to estimate each metabolite's global concentration per tissue type,5
thereby boosting the sensitivity, i.e. the precision.8
It also renders changes in T2-visible lesions insignificant, since they comprise less than 1% of the VOI's volume, as reflected by a lack of correlation between their load and metabolic levels. Therefore, abnormalities detected with this approach must be diffuse throughout the VOI.
Our specific goals were to assign the glial abnormalities previously found in these patients6
to tissue type and follow this cohort in order to establish the temporal dynamics of their GM and WM metabolism vs controls.
Previously, we showed diffusely elevated Cr, Cho, and mI in the entire VOI of these patients, but could not assign them to a specific tissue type. Given the VOI's ~3:2 WM:GM ratio, however, we conjectured that they represent WM status.6
This hypothesis is supported by the results here indicating diffuse glial abnormalities in patients' WM relatively early in their RRMS course. The Cr, Cho, and mI concentrations are consistently higher in patients' WM than in controls (on average by 8%, 12%, and 11%, ). Importantly, the increased sensitivity due to separating tissue types helped identify the patients' smaller (6% on average) WM NAA deficits, indicating that axonal impairment accompanies the glial pathology.
Note that the differences in concentrations were on the order of the coefficients of variation, suggesting that they might not always be statistically identifiable due to biological and instrumental “noise.” Indeed, higher standard deviations in patients' WM, despite larger sample size, may reflect heterogeneity of disease course or immune or medication responses, suggesting caution in drawing general conclusions from cross-sectional studies, where they can mask subtle pathologic effect. This point may be most pertinent for NAA, which showed the smallest differences, explaining the conflicting reports of both unchanged9,10
levels in cohorts of similar disease duration. This has important implications for timing and monitoring the neurodegenerative phase of RRMS. In this study, we conclusively show diffuse changes in Cho, Cr, mI, and NAA in the WM of patients with RRMS within ~3 to 6 years from diagnosis.
The single time point GM metabolic differences between patients and controls preclude us from concluding definite GM dysfunction. Previous 1
H-MRSI in RRMS has ascribed GM changes to effects of axonal transection associated with WM lesions,13
or even only to their inflammation.14
Given the paucity of lesions and relapses in our cohort, it is plausible that GM injury is below the detection threshold since it may still be focal; is still confined to specific structures, e.g., basal ganglia but not the cortex15
; or is spatially heterogeneous. Such hypotheses, testable in the original 3D data, are beyond the scope here. We conclude, therefore, that at this stage, MS is not characterized by diffuse 1
H-MRSI-detectable GM damage.
None of the intercohort rates were significantly different, but patients as a group had significant rates of change. In GM, these were decreases of Cho and mI, and although they may reflect normalizations from a previous elevation, they seem to be driven mainly by a sharp decrease at the last time point, a behavior mimicked by the controls, who have almost identical rates (Cho: −0.03 mM/year for patients and −0.02 mM/year for controls, mI: −0.1 mM/year for both), as seen in . We are therefore uncertain whether these changes are meaningful or resulted from a systematic bias, also present in controls, where rates were not significant due to their smaller sample size.
In WM, there were significant increases in patients' Cr, Cho, and NAA with much larger intercohort rate differences: fivefold faster Cr and ~ twofold Cho, as well as NAA increases in patients compared to controls. We infer the Cr and Cho as progression of glial pathology and the NAA as concurrent axonal recovery or sparing, possibly due to treatment. This is the first serial report of increasing WM Cr and Cho, and likely reflects progressing glial pathology. In contrast, stable or recovering WM NAA is reported in the majority (10 of 14) of serial studies (e.g., references 16-18), where most patients were on interferons or glatiramer acetate, the main medications of our cohort and a possible reason for the observed partial NAA recovery. Indeed, WM NAA/Cr increased in interferon-treated19
and glatiramer acetate–treated20
patients and declined in nontreated ones. However, as studies with untreated patients are rare, extrapolating these effects to medication per se is speculative. In addition, since most serial studies employed NAA/Cr ratios, Cr elevation, as shown here, may confound concurrent NAA recovery.
The specificity of 1
H-MRS can differentiate glial from neuronal injury. In vitro and ex vivo studies have established Cr, Cho, and mI as markers of the former and NAA of the latter.21,22
The total Cr resonance comprises free creatine and phosphocreatine,2
found in all cell types, but at higher concentrations in glia than neurons.21
In MS, high WM phosphocreatine levels have been associated with dysfunctional astrocytes,23
while recent absolute quantification showed that high Cr levels are due to both free and phosphocreatine and therefore represent gliosis rather than change in energy metabolism.24
Presence of gliosis in WM has been documented ex vivo4,25
and corroborated here by the elevated mI, an osmolyte involved in signal transduction in astrocytes26
and thought to represent astrogliosis.27
Further glial involvement is suggested by high levels of Cho, which indicate abnormal membrane turnover from demyelination and remyelination.27
Since Cho is present in all cell walls, however, it may also reflect an ongoing gliosis, as reported in a combined 1
H-MRS histopathology study.28
In contrast, NAA is almost exclusive to neurons and is therefore considered their marker. NAA levels decline not only with the number of neurons,28
but also under metabolic stress, so by itself a decrease is nonspecific. In this study, however, we observe concurrent increasing NAA and decreasing WMf
, suggesting that at this stage of RRMS, neuronal dysfunction dominates. Recently, energy failure due to mitochondrial dysfunction has been proposed as a mechanism for neuronal degeneration in MS.23,29–31
Since NAA is synthesized in the mitochondria, lower levels may reflect reduced mitochondrial activity.29
If that is driven by inflammation,31
it is possible that the increasing NAA reflects mitochondrial recovery in response to anti-inflammatory medication. The fact that in the progressive phase of MS neuronal death is independent of inflammation, however, may be related to the steady increase in glial abnormalities, which are more pronounced than NAA loss in early RRMS, and which, unlike NAA, do not seem to abate.
was higher than controls', and increased significantly, reflecting progressive enlargement of lateral and third ventricle and sulci around the medial part of the longitudinal fissure. It was the only rate that showed any evidence of intercohort difference (p
= 0.06). This type of deep central atrophy characterizes the RRMS phenotype, as it starts early and progresses faster than global atrophy.32
Patients and controls had statistically indistinguishable WMf
, but there was evidence of loss of both tissue types in patients. First, their WMf
had a significant rate of decrease, suggesting accumulating WM atrophy. Second, they had consistently lower GMf
medians and distributions (), suggesting greater GM contribution to atrophy. Indeed, GM loss may be apparent as early as clinical onset with involvement of the cortex, basal ganglia, and thalamus.33–35
It is possible that some WM atrophy is masked by inflammatory changes, such as astrogliosis and remyelination, which are less pronounced in GM,36,37
as demonstrated by the high WM Cr, Cho, and mI but mostly normal GM metabolite levels.
The postprocessing choice of obtaining global tissue values was to investigate diffuse changes only. Since the original 3D data are available, exploring the contribution of specific structures or regions is feasible. The advantage of our approach, however, was in assessing the largest tissue volumes possible in order to increase the sensitivity, which might otherwise be insufficient to distinguish subtle changes in smaller regions of interest. Second, since our goal was to assess the overall disease load by tissue type, the reported WM changes represent the combined contribution of normal-appearing and “dirty” WM (), with unknown contribution of each to the signal. Third, not accounting for lesions in segmentation can result in volume underestimation or overestimation depending on their intensity.38
Given their low loads in our patients (~4 cm3
), in the most common scenario of misclassification as GM (), volume errors would be under 0.7%.38
Finally, while the reported findings are from a relatively large volume (~40% of the brain's WM and 20% of its GM), they are pertinent only to the regions inside the VOI () and may not be representative of other domains, such as the rest of the cortex.
Glial abnormalities in WM, consistent with myelin breakdown and astrogliosis, are more pronounced than the axonal deficits attributed to dysfunction, such as mitochondrial impairment. In the context of increasing central atrophy and WM tissue loss, the glial changes progressed, while the axons showed partial recovery, presumably in response to treatment. Absolute quantification allowed us to assign metabolic profiles by cell class, which revealed a different time course of glial and neuronal metabolism. This has implications for understanding the natural history of MS as well as monitoring response to treatment paradigms.